American Journal of Epidemiology

Chahal, Harpreet S; Gelaye, Bizu; Mostofsky, Elizabeth; Sanchez, Sixto E; Mittleman, Murray A; Maclure, Malcolm; Pacora, Percy; Torres, Jose A; Romero, Roberto; Ananth, Cande V; Williams, Michelle A

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy138pmid: 29992226

Abstract While there is consistent evidence that episodes of physical exertion are associated with an immediately higher risk of acute ischemic vascular events, the risk of placental abruption immediately following episodes of physical exertion has not been studied. In a multicenter case-crossover study, we interviewed 663 women with placental abruption at 7 Peruvian hospitals between January 2013 and August 2015. We asked women about physical exertion in the hour before symptom onset and compared this with their frequency of physical exertion over the prior week. Compared with times with light or no exertion, the risk of placental abruption was 7.8 (95% confidence interval (CI): 5.5, 11.0) times greater in the hour following moderate or heavy physical exertion. The instantaneous incidence rate ratio of placental abruption within an hour of moderate or heavy physical exertion was lower for women who habitually engaged in moderate or heavy physical activity more than 3 times per week in the year before pregnancy (rate ratio (RR) = 3.0, 95% CI: 1.6, 5.9) compared with more sedentary women (RR = 17.3, 95% CI: 11.3, 26.7; P for homogeneity < 0.001), and the rate ratio was higher among women with preeclampsia/eclampsia (RR = 13.6, 95% CI: 7.0, 26.2) than among women without (RR = 6.7, 95% CI: 4.4, 10.0; P for homogeneity = 0.07). abruptio placentae, case-crossover, exercise, physical exertion, placental abruption Editor’s note: An invited commentary on this article appears on page 2080, and the authors’ response appears on page 2083. Placental abruption—the premature separation of the implanted placenta—is an obstetrical complication that affects roughly 1%–2% of all pregnancies (1). Potential sequelae of placental abruption include hemorrhagic shock, coagulopathy, disseminated intravascular coagulation, uterine rupture, and renal failure. In more and less developed countries alike, placental abruption, along with infection and hypertensive disorders, continues to appear among the top 3 causes of maternal mortality. In addition to its impact on women, placental abruption is also a significant cause of infant morbidity and mortality (2–4). Despite its acute clinical presentation, placental abruption is characterized by both acute and chronic pathophysiologic features. The specific etiology of placental abruption is still unknown, but its underlying mechanisms include uteroplacental underperfusion (5, 6), ischemia (7, 8), and chronic hypoxemia (9). While there is a substantial body of evidence showing an acutely heightened risk of ischemic cardiovascular events immediately following physical exertion (10), the association between physical exertion and the acute risk of placental abruption—“an ischemic placental disorder” (7, 11–13)—has not yet been investigated. Similar to cardiovascular disorders, habitual physical activity among pregnant women is associated with several health benefits over the long-term, including a lower baseline risk of several maternal (14–17) and fetal (18) outcomes. We used a case-crossover design to compare each woman’s risk of placental abruption in the hour following episodes of physical exertion with that same woman’s risk at other times. We examined whether moderate or heavy physical exertion was associated with an immediately higher risk of placental abruption and whether the risk was different for moderate versus heavy physical exertion. We also evaluated whether the association varied according to habitual physical activity or with comorbid hypertensive disorders of pregnancy (i.e., preeclampsia or eclampsia). METHODS Source population The Placental Abruption Genetic Epidemiology and Triggers (PAGE) study was conducted in Lima, Peru, in South America, in the following hospitals: Instituto Nacional Materno Perinatal, Hospital Rebagliati, Hospital San Bartolome, Hospital Hipolito, Hospital Loayza, Hospital Dos de Mayo, and Hospital Maria Auxiliadrosa. During the study period, there were 218,420 deliveries in the participating hospitals. Between January 2013 and August 2015, 883 women were identified as eligible for interview during the in-patient labor and delivery period, 663 (75%) of whom elected to participate. Research staff identified eligible women by reviewing admission logbooks for the emergency room, labor and delivery, and surgery. Study participants were recruited during their hospital stay. Placental abruption cases were identified using daily monitoring of all new admissions to antepartum wards, emergency room wards (intensive care unit), and labor and delivery wards of the study hospitals. Study personnel made periodic visits to specific wards in a fixed order for the purposes of identifying potential placental abruption cases for the present study. Hospital medical records were reviewed to ensure that cases met the definition of placental abruption based on preestablished diagnostic criteria (19), by fulfilling one or more of the following 4 eligibility criteria noted in the women’s medical records: 1) antepartum hemorrhage after 20 weeks of gestation; 2) uterine pain or tenderness (localized or diffuse); 3) fetal distress or death; and 4) retroplacental blood clot. The data-collection protocol also included ascertainment of differential diagnoses for placental abruption, including all causes of abdominal pain and bleeding, such as placenta previa, appendicitis, urinary tract infection, preterm labor, fibroid degeneration, ovarian pathology, and muscular pain. We excluded women who had a multifetal pregnancy, who were not residents of Lima (i.e., women transferred from remote areas of Peru), or whose medical records were insufficient to determine the presence or absence of placental abruption. If a woman met the criteria, she was invited to participate in a 30-minute in-person interview, and she signed a consent form if she agreed. Interviews occurred a median of 2 days (interquartile range, 1–3 days) after placental abruption. The study procedures used in this study were in agreement with the protocols approved by the institutional review board at each participating institution. Case-crossover design Instead of selecting a separate control group, we employed a case-crossover design whereby the cases serve as their own controls. The method was developed to estimate the transient effect of an intermittent exposure on events with an identifiable, abrupt onset (20). Because the approach involves a comparison of the same person at different times, there is no confounding by fixed or slowly varying characteristics such as age, prepregnancy body mass index, socioeconomic level, or history of previous placental abruption. This design involves collecting information on exposure (e.g., participation in moderate or heavy physical exertion) immediately preceding the event (e.g., placental abruption) and comparing this with the expected frequency of exposure based on women’s exposure patterns during similar control time periods. A priori, based on the presumed pathophysiologic process, we defined the case window as the hour before onset of placental abruption and the control window as the rest of the week before onset of placental abruption. Interview Interviews were completed in Spanish by trained, female interviewers using a standardized, structured questionnaire that covered a variety of potential triggers as well as sociodemographic, medical, reproductive, and lifestyle information. We also conducted a physical exam to obtain anthropometric measurements. To help standardize reporting of the intensity of physical exertion, women were shown a 15-point visual analogue Borg Scale (21) (scores ranging from 6 to 20). Then we provided examples of physical exertion at each level of intensity to help women classify episodes as light (e.g., mopping), moderate (e.g., dancing), or heavy (e.g., sprinting). We started by asking each woman her usual frequency of exertion during 3 periods: the year before pregnancy, the first 3–6 months of pregnancy, and months 6–9 of pregnancy. Then we asked about her frequency of light exertion during the week before placental abruption onset, and asked when she last engaged in light exertion before placental abruption onset (e.g., abdominal/back pain with or without vaginal bleeding). Her responses to the latter question were classified as: never, at the time of event onset, 1/2 hour before, 1 hour before, 2 hours before, 3–6 hours before, 6–24 hours before, 1–2 days before, 3–4 days before, or ≥5 days before. Then we asked about her frequency of moderate exertion in the week before, and asked when she last engaged in moderate exertion before placental abruption onset. Last, we asked about her frequency and last time of heavy exertion. Statistical analysis In a case-crossover analysis, data are stratified by individual persons (20, 22). In the analysis of this self-matched design, each woman forms her own stratum. She is considered an exposed case if she was exposed to moderate or heavy physical exertion in the hour before symptom onset and unexposed otherwise. For each woman, the number of exposed person-hours in the week prior to placental abruption was calculated by multiplying the number of self-reported episodes of moderate or heavy physical exertion by the hypothesized duration of its physiologic effect (1 hour). Her unexposed person-time was calculated by subtracting the number of exposed person-hours from the 168 hours in a week. Using methods for sparse data analysis, we calculated the Mantel-Haenszel incidence rate ratio for person-time and 95% confidence intervals (23). We conducted analyses to assess whether the association between physical exertion and placental abruption differed according to the intensity of physical exertion (heavy vs. moderate). In addition, we conducted subgroup analyses to assess whether the association varied according to self-reported history of habitual physical activity (≤3 vs. >3 times per week in the year before pregnancy) and preeclampsia/eclampsia (yes vs. no) by means of a Wald χ2 test of homogeneity (23). To evaluate whether other potential triggers could account for the observed association, we conducted a sensitivity analysis excluding women who engaged in other potentially triggering activities in the hour before placental abruption (sexual intercourse; cigarette smoking; consumption of alcohol, coffee, tea, or cola; or use of cocaine, marijuana, or terokal (an adhesive that is inhaled recreationally)). To explore the extent to which results might vary due to circadian rhythm, we conducted an ad hoc analysis in which we compared the proportion of exertion- and non-exertion-associated placental abruptions that occurred between 6 pm and 12 am. All reported P values are 2-sided. RESULTS The characteristics of women who experienced placental abruption are described in Table 1. Among 663 women who had placental abruption, 7 provided no information on usual physical exertion so they were excluded from the analyses. Of the 656 included, 352 (54%) reported that they engaged in moderate or heavy physical exertion in the week before placental abruption. Among the 352 who engaged in moderate or heavy physical exertion, 263 (75%) reported engaging in such exertion once that week, and 42 (12%) reported more than 3 times that week. The mean frequency of habitual physical activity decreased after conception and over the course of pregnancy, from 3.8 times/week in the year before pregnancy, to 2.7 times/week in the first 3–6 months of pregnancy, to 1.6 times/week in months 6–9 of pregnancy, to 0.9 times/week in the week prior to placental abruption (Figure 1). Table 1. Characteristics of Women With Placental Abruption (n = 656) in Peru, 2013–2015 Characteristic . Moderate or Heavy Physical Exertion in Prior Week . Yes (n = 352) . No (n = 304) . No.a . % . No.a . % . Maternal age, years  <20 19 5.4 22 7.4  20–24 89 25.5 82 27.4  25–29 92 26.4 66 22.1  30–34 64 18.3 68 22.7  35–39 66 18.9 42 14.0  ≥40 19 5.4 19 6.4 Prepregnancy BMIb  Underweight (<18.5) 9 3.2 5 2.2  Normal (18.5–24.9) 142 51.1 123 54.4  Overweight (25.0–29.9) 101 36.3 82 36.3  Obese (≥30.0) 26 9.4 16 7.1 Education  Less than high school 36 10.4 18 6.0  High school 205 59.2 174 58.0  More than high school 105 30.3 108 36.0 Gravidac  ≤1 109 31.5 100 33.3  2 102 29.5 86 28.7  ≥3 135 39.0 114 38.0 Parityc  0 19 5.5 16 5.3  1 135 39.1 127 42.3  ≥2 191 55.4 157 52.3 Substance use during pregnancy  Cigarette 14 4.0 8 2.6  Alcohol 88 25.0 41 13.5  Cocaine 3 0.9 5 1.7  Marijuana 3 0.9 2 0.7  Terokal 4 1.1 1 0.3 Employed during pregnancy 192 55.0 165 54.6 Preeclampsia/eclampsia 72 20.5 73 24.6 Chorioamnionitis 14 4.0 14 4.7 Premature rupture of membranes 54 15.4 44 14.8 Prior placental abruption 14 4.0 12 4.0 Chronic hypertension 8 2.3 0 0.0 Anemiad 146 44.9 111 41.6 Prenatal care 329 94.3 286 94.1 Prenatal vitamin use 293 84.4 262 86.8 Gestational age, weekse,f 35.1 (4.1) 34.1 (4.4) Characteristic . Moderate or Heavy Physical Exertion in Prior Week . Yes (n = 352) . No (n = 304) . No.a . % . No.a . % . Maternal age, years  <20 19 5.4 22 7.4  20–24 89 25.5 82 27.4  25–29 92 26.4 66 22.1  30–34 64 18.3 68 22.7  35–39 66 18.9 42 14.0  ≥40 19 5.4 19 6.4 Prepregnancy BMIb  Underweight (<18.5) 9 3.2 5 2.2  Normal (18.5–24.9) 142 51.1 123 54.4  Overweight (25.0–29.9) 101 36.3 82 36.3  Obese (≥30.0) 26 9.4 16 7.1 Education  Less than high school 36 10.4 18 6.0  High school 205 59.2 174 58.0  More than high school 105 30.3 108 36.0 Gravidac  ≤1 109 31.5 100 33.3  2 102 29.5 86 28.7  ≥3 135 39.0 114 38.0 Parityc  0 19 5.5 16 5.3  1 135 39.1 127 42.3  ≥2 191 55.4 157 52.3 Substance use during pregnancy  Cigarette 14 4.0 8 2.6  Alcohol 88 25.0 41 13.5  Cocaine 3 0.9 5 1.7  Marijuana 3 0.9 2 0.7  Terokal 4 1.1 1 0.3 Employed during pregnancy 192 55.0 165 54.6 Preeclampsia/eclampsia 72 20.5 73 24.6 Chorioamnionitis 14 4.0 14 4.7 Premature rupture of membranes 54 15.4 44 14.8 Prior placental abruption 14 4.0 12 4.0 Chronic hypertension 8 2.3 0 0.0 Anemiad 146 44.9 111 41.6 Prenatal care 329 94.3 286 94.1 Prenatal vitamin use 293 84.4 262 86.8 Gestational age, weekse,f 35.1 (4.1) 34.1 (4.4) Abbreviation: BMI, body mass index. a Frequencies in subgroups may not sum to column total due to missing data. b Weight (kg)/height (m)2. c Including current pregnancy. d Hemoglobin value of <11 g/dL in second or third trimester. e Estimated from last menstrual period. f Values are expressed as mean (standard deviation). Open in new tab Table 1. Characteristics of Women With Placental Abruption (n = 656) in Peru, 2013–2015 Characteristic . Moderate or Heavy Physical Exertion in Prior Week . Yes (n = 352) . No (n = 304) . No.a . % . No.a . % . Maternal age, years  <20 19 5.4 22 7.4  20–24 89 25.5 82 27.4  25–29 92 26.4 66 22.1  30–34 64 18.3 68 22.7  35–39 66 18.9 42 14.0  ≥40 19 5.4 19 6.4 Prepregnancy BMIb  Underweight (<18.5) 9 3.2 5 2.2  Normal (18.5–24.9) 142 51.1 123 54.4  Overweight (25.0–29.9) 101 36.3 82 36.3  Obese (≥30.0) 26 9.4 16 7.1 Education  Less than high school 36 10.4 18 6.0  High school 205 59.2 174 58.0  More than high school 105 30.3 108 36.0 Gravidac  ≤1 109 31.5 100 33.3  2 102 29.5 86 28.7  ≥3 135 39.0 114 38.0 Parityc  0 19 5.5 16 5.3  1 135 39.1 127 42.3  ≥2 191 55.4 157 52.3 Substance use during pregnancy  Cigarette 14 4.0 8 2.6  Alcohol 88 25.0 41 13.5  Cocaine 3 0.9 5 1.7  Marijuana 3 0.9 2 0.7  Terokal 4 1.1 1 0.3 Employed during pregnancy 192 55.0 165 54.6 Preeclampsia/eclampsia 72 20.5 73 24.6 Chorioamnionitis 14 4.0 14 4.7 Premature rupture of membranes 54 15.4 44 14.8 Prior placental abruption 14 4.0 12 4.0 Chronic hypertension 8 2.3 0 0.0 Anemiad 146 44.9 111 41.6 Prenatal care 329 94.3 286 94.1 Prenatal vitamin use 293 84.4 262 86.8 Gestational age, weekse,f 35.1 (4.1) 34.1 (4.4) Characteristic . Moderate or Heavy Physical Exertion in Prior Week . Yes (n = 352) . No (n = 304) . No.a . % . No.a . % . Maternal age, years  <20 19 5.4 22 7.4  20–24 89 25.5 82 27.4  25–29 92 26.4 66 22.1  30–34 64 18.3 68 22.7  35–39 66 18.9 42 14.0  ≥40 19 5.4 19 6.4 Prepregnancy BMIb  Underweight (<18.5) 9 3.2 5 2.2  Normal (18.5–24.9) 142 51.1 123 54.4  Overweight (25.0–29.9) 101 36.3 82 36.3  Obese (≥30.0) 26 9.4 16 7.1 Education  Less than high school 36 10.4 18 6.0  High school 205 59.2 174 58.0  More than high school 105 30.3 108 36.0 Gravidac  ≤1 109 31.5 100 33.3  2 102 29.5 86 28.7  ≥3 135 39.0 114 38.0 Parityc  0 19 5.5 16 5.3  1 135 39.1 127 42.3  ≥2 191 55.4 157 52.3 Substance use during pregnancy  Cigarette 14 4.0 8 2.6  Alcohol 88 25.0 41 13.5  Cocaine 3 0.9 5 1.7  Marijuana 3 0.9 2 0.7  Terokal 4 1.1 1 0.3 Employed during pregnancy 192 55.0 165 54.6 Preeclampsia/eclampsia 72 20.5 73 24.6 Chorioamnionitis 14 4.0 14 4.7 Premature rupture of membranes 54 15.4 44 14.8 Prior placental abruption 14 4.0 12 4.0 Chronic hypertension 8 2.3 0 0.0 Anemiad 146 44.9 111 41.6 Prenatal care 329 94.3 286 94.1 Prenatal vitamin use 293 84.4 262 86.8 Gestational age, weekse,f 35.1 (4.1) 34.1 (4.4) Abbreviation: BMI, body mass index. a Frequencies in subgroups may not sum to column total due to missing data. b Weight (kg)/height (m)2. c Including current pregnancy. d Hemoglobin value of <11 g/dL in second or third trimester. e Estimated from last menstrual period. f Values are expressed as mean (standard deviation). Open in new tab Figure 1. Open in new tabDownload slide Mean frequency (standard deviation (SD)) of self-reported habitual physical activity per week according period relative to pregnancy among women with placental abruption in Peru (n = 656), 2013–2015. The mean frequency of habitual physical activity decreased from before conception and over the course of pregnancy: 3.8 (SD, 4.9) times/week in the year before pregnancy, 2.7 (SD, 4.5) times/week in the first 3–6 months of pregnancy, 1.6 (SD, 3.3) times/week in months 6–9 of pregnancy, and 0.9 (SD, 2.9) times/week in the week prior to placental abruption. Among 352 women who engaged in moderate or heavy physical exertion in the week before placental abruption, 34 women reported having engaged in moderate or heavy physical exertion in the hour before the onset of placental abruption. The immediate risk of placental abruption was 7.8-fold higher (95% confidence interval (CI): 5.5, 11.0) within an hour of moderate or heavy physical exertion compared with periods of lower exertion or rest. The rate ratio of placental abruption within an hour of physical exertion was higher following heavy-intensity physical exertion (rate ratio (RR) = 13.7, 95% CI: 7.0, 26.5) than after moderate physical exertion (RR = 6.0, 95% CI: 4.0, 9.0; P for homogeneity = 0.04; Figure 2). Figure 2. Open in new tabDownload slide Rate ratio (RR) of placental abruption within an hour of heavy-intensity physical exertion and moderate-intensity physical exertion, among women with placental abruption in Peru (n = 656), 2013–2015. The RR of placental abruption within an hour of physical exertion was higher following heavy physical exertion (RR = 13.7, 95% CI: 7.0, 26.5; n = 10 women exposed in the hour before placental abruption) than after moderate physical exertion (RR = 6.0, 95% CI: 4.0, 6.0; n = 24 women exposed in the hour before placental abruption; P for homogeneity = 0.04). The y-axis is in log scale. The error bars indicate 95% confidence intervals. The rate ratio of placental abruption was lower for women who habitually engaged in physical activity more than 3 times per week in the year before pregnancy (RR = 3.0, 95% CI: 1.6, 5.9) compared with women who engaged in such activity 3 or fewer times per week (RR = 17.3, 95% CI: 11.3, 26.7; P for homogeneity < 0.001; Figure 3A). Additionally, the rate ratio of placental abruption appeared to be higher among women with preeclampsia/eclampsia (RR = 13.6, 95% CI: 7.0, 26.2) than among normotensive women (RR = 6.7, 95% CI: 4.4, 10.0; P for homogeneity = 0.07) (Figure 3B). Figure 3. Open in new tabDownload slide Rate ratio (RR) of placental abruption within an hour of moderate or heavy physical exertion, stratified by habitual physical activity in the year before pregnancy or preeclampsia/eclampsia, among women with placental abruption in Peru (n = 656), 2013–2015. A) Among women who habitually engaged in physical activity more than 3 times per week before pregnancy, the RR of placental abruption was lower (RR = 3.0, 95% confidence interval (CI): 1.6, 5.9; n = 25 women exposed in the hour before placental abruption) compared with women who engaged in physical activity 3 or fewer times per week (RR = 17.3, 95% CI: 11.3, 26.7; n = 9 women exposed in the hour before placental abruption; P for homogeneity < 0.001. B) Among women with preeclampsia/eclampsia, the RR of placental abruption appeared to be higher (RR = 13.6, 95% CI: 7.0, 26.2; n = 10 women exposed in the hour before placental abruption) than among normotensive women (RR = 6.7, 95% CI: 4.4, 10.0; n = 24 women exposed in the hour before placental abruption; P for homogeneity = 0.07). The y-axis is in log scale. The error bars indicate 95% confidence intervals. In a sensitivity analysis excluding women who were exposed to other potential triggers (sexual intercourse, cigarette smoking, illicit drugs, or caffeinated or alcoholic beverages), the rate ratio of placental abruption associated with moderate or heavy physical exertion was marginally attenuated but remained strong (RR = 7.0, 95% CI: 4.9, 10.2). Between 6 pm and 12 am, there were 10 (31%) exertion-related placental abruptions and 160 (27%) non-exertion-related abruptions. DISCUSSION In this study, episodes of moderate or heavy physical exertion were associated with an immediately heightened risk of placental abruption that was 7.8-fold higher in the subsequent hour compared with periods of lower exertion or rest. The risk of placental abruption within an hour of physical exertion was higher for exertion at heavy intensity than it was for moderate levels of exertion. The risk of placental abruption within an hour of moderate or heavy physical exertion was lower among women who habitually engaged in physical activity more than 3 times a week prior to pregnancy than for those who were more sedentary, and it appeared to be higher for women with preeclampsia or eclampsia than it was for normotensive women, although the association among women with preeclampsia or eclampsia was statistically imprecise. To our knowledge, no prior study has assessed the association between episodes of physical exertion and the acute risk of placental abruption. While considerable progress has been made in understanding traditional chronic risk factors of placental abruption, acute triggers have been the subject of much less inquiry. More than two decades ago, placental abruption was linked to acute sympathetic activation in a case report (24), and it was linked to labor-intensive employment in a case-control study (25). Change in the risk of acute placental abruption was recently described in a different context; Michikawa et al. (26) used a case-crossover approach to demonstrate an increased risk of placental abruption within 3 days of heightened nitrogen dioxide air pollution in Japan, an isolated finding that parallels the association between pollutants and acute myocardial infarction (27) and ischemic stroke (28). Moreover, effect modification from habitual physical activity in this study of placental abruption and in other studies of acute ischemic cardiovascular events (29–31) was similar, providing support for the possibility that the short-term effect of physical exertion might contribute to the pathophysiology of both types of disorders through similar mechanisms. Our finding that habitual physical activity decreases after conception and over the course of pregnancy is also consistent with prior work on activity trends during pregnancy in the United States (32). Given that physical activity during pregnancy is also associated with a lower risk of gestational diabetes (15, 17), preeclampsia (16), and cesarean delivery (14) and that sedentary behavior is associated with a higher risk of venous thromboembolism (33), it is nonetheless important for women who are pregnant to remain active at levels that are safe for them. Further, because habitual physical activity before pregnancy may mitigate the acute risk of placental abruption associated with each episode of moderate or heavy physical activity, the period before conception may represent the ideal time for enrollment into a graduated exercise program for those at highest risk of placental abruption. In the case of physical exertion and acute myocardial infarction, the transient effect of strenuous exertion is more than offset by the cumulative protective effect of regular moderate exertion. It is possible that the same is true for physical exertion and placental abruption. However, without comparison with a traditional control group, we cannot assess whether there is an overall protective effect of regular physical exertion on occurrence of placental abruption. However, based on the lower rate ratio for women who were more physically active, we believe it is plausible that regular moderate exertion provides a protective effect that more than offsets the very transient hazardous association between moderate and heavy exertion and placental abruption risk. Most cases of placental abruption are caused by both acute and chronic pathophysiological processes (1). Studies of its acute etiology report factors such as shear stress, often a result of trauma and/or sudden uterine decompression from hydramnios or delivery of a first twin (1). Other acute factors include vasoconstriction secondary to cocaine use and endothelial dysfunction secondary to intense inflammatory responses at the placental-decidual interface (34). Suggested pathophysiologic triggers that may be involved in the acute effect of episodes of physical exertion include hypoxia-induced changes in the maternal-fetal circulation, uteroplacental vascular insufficiency and ischemia, infarctions, and heightened sympathetic activity, which provide support for the concept that placental abruption is in part an ischemic placental disorder (7, 11, 13). There are some limitations to our study. By using each patient as her own control, the case-crossover design eliminates between-person confounding by all fixed and slow-varying characteristics. However, within-person confounding by factors that change over time can occur. Therefore, we conducted a sensitivity analysis excluding individuals exposed to other potential triggering exposures in the hour before placental abruption; the results were not materially different. In an effort to minimize recall bias, women were not informed of the duration of the hypothesized hazard period and were presented with many options for the time period in which they were last exposed, including the entire control period in the week preceding the placental abruption. In this study, we had limited statistical power to test whether risk associated with physical exertion varied according to vaginal bleeding during pregnancy, hypertensive disorders of pregnancy, and prior placental abruption, due to the small number of exposed cases in these categories. A larger study that also collects information on the type of physical exertion performed before placental abruption would help inform physical activity recommendations among women at highest risk. Also, we did not collect information on the timing of typical physical exertion in the reference period, but future research to examine the association according to time of day might provide insight into whether there are times when women are particularly susceptible. In conclusion, within an hour of moderate or heavy physical exertion, there is an elevated risk of placental abruption compared with periods of lower exertion or rest. The rate ratio of placental abruption within an hour of physical exertion was higher following heavy exertion than moderate exertion. The association between moderate or heavy physical exertion and placental abruption is lower among women who engaged in higher levels of habitual physical activity before pregnancy, and it was higher among women who had preeclampsia or eclampsia. Further research is necessary to design activity recommendations for women at high risk of placental abruption that can maximize the benefits of habitual physical activity while minimizing the risk of placental abruption associated with each episode of exertion. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Harpreet S. Chahal, Bizu Gelaye, Elizabeth Mostofsky, Murray A. Mittleman, Michelle A. Williams); Mississauga Academy of Medicine, University of Toronto Mississauga, Mississauga, Canada (Harpreet S. Chahal); Cardiovascular Epidemiology Research Unit, Beth Israel Deaconess Medical Center, Boston, Massachusetts (Elizabeth Mostofsky, Murray A. Mittleman); Universidad Peruana de Ciencias Aplicadas, Lima, Peru (Sixto E. Sanchez); Asociación Civil Proyectos en Salud , Lima, Peru (Sixto E. Sanchez); Department of Anesthesiology, Pharmacology and Therapeutics, Faculty of Medicine, University of British Columbia, Vancouver, Canada (Malcolm Maclure); Department of Obstetrics and Gynecology, San Marcos University, Hospital Madre-Nino San Bartolome, Lima, Peru (Percy Pacora, Jose A. Torres); Department of Obstetrics and Gynecology, San Marcos University, Hospital Nacional Hipólito Unanue, Lima, Peru (Jose A. Torres); Perinatology Research Branch, Division of Obstetrics and Maternal-Fetal Medicine, Division of Intramural Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland (Roberto Romero); Department of Obstetrics and Gynecology, University of Michigan Health System, University of Michigan, Ann Arbor, Michigan (Roberto Romero); Department of Epidemiology and Biostatistics, College of Human Medicine, Michigan State University, East Lansing, Michigan (Roberto Romero); Center for Molecular Medicine and Genetics, Wayne State University, Detroit, Michigan (Roberto Romero); Department of Obstetrics and Gynecology, College of Physicians and Surgeons, Columbia University, New York, New York (Cande V. Ananth); and Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York (Cande V. Ananth). H.S.C. and B.G. contributed equally to this work. This work was funded by the National Institutes of Health (grants R01 HD059827 and T37 MD001449). This research was supported, in part, by the Perinatology Research Branch, Division of Obstetrics and Maternal-Fetal Medicine, Division of Intramural Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, US Department of Health and Human Services, and, in part, with Federal funds from the National Institute of Child Health and Human Development, National Institutes of Health, US Department of Health and Human Services, under contract HHSN275201300006C. R.R. has contributed to this work as part of his official duties as an employee of the US Federal Government. We thank the participating hospitals. We also thank Elena Sanchez and the dedicated staff members of Asociacion Civil Proyectos en Salud (PROESA), Peru, for their expert technical assistance with this research. Conflict of interest: none declared. Abbreviations CI confidence interval RR rate ratio REFERENCES 1 Oyelese Y , Ananth CV. Placental abruption . Obstet Gynecol . 2006 ; 108 ( 4 ): 1005 – 1016 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Ananth CV , Wilcox AJ. Placental abruption and perinatal mortality in the United States . Am J Epidemiol . 2001 ; 153 ( 4 ): 332 – 337 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Kayani SI , Walkinshaw SA, Preston C. Pregnancy outcome in severe placental abruption . BJOG . 2003 ; 110 ( 7 ): 679 – 683 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Spinillo A , Fazzi E, Stronati M, et al. . Severity of abruptio placentae and neurodevelopmental outcome in low birth weight infants . Early Hum Dev . 1993 ; 35 ( 1 ): 45 – 54 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Harrington K , Cooper D, Lees C, et al. . Doppler ultrasound of the uterine arteries: the importance of bilateral notching in the prediction of pre-eclampsia, placental abruption or delivery of a small-for-gestational-age baby . Ultrasound Obstet Gynecol . 1996 ; 7 ( 3 ): 182 – 188 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Romero R , Kusanovic JP, Chaiworapongsa T, et al. . Placental bed disorders in preterm labor, preterm PROM, spontaneous abortion and abruptio placentae . Best Pract Res Clin Obstet Gynaecol . 2011 ; 25 ( 3 ): 313 – 327 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Ananth CV , Peltier MR, Kinzler WL, et al. . Chronic hypertension and risk of placental abruption: is the association modified by ischemic placental disease? Am J Obstet Gynecol . 2007 ; 197 ( 3 ): 273.e1 – 273.e7 . Google Scholar Crossref Search ADS WorldCat 8 Ananth CV , Vintzileos AM. Ischemic placental disease: epidemiology and risk factors . Eur J Obstet Gynecol Reprod Biol . 2011 ; 159 ( 1 ): 77 – 82 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Ananth CV , Savitz DA, Bowes WA , Jr., et al. . Influence of hypertensive disorders and cigarette smoking on placental abruption and uterine bleeding during pregnancy . Br J Obstet Gynaecol . 1997 ; 104 ( 5 ): 572 – 578 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Dahabreh IJ , Paulus JK. Association of episodic physical and sexual activity with triggering of acute cardiac events: systematic review and meta-analysis . JAMA . 2011 ; 305 ( 12 ): 1225 – 1233 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Ananth CV , Peltier MR, Chavez MR, et al. . Recurrence of ischemic placental disease . Obstet Gynecol . 2007 ; 110 ( 1 ): 128 – 133 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Ananth CV , Vintzileos AM. Maternal-fetal conditions necessitating a medical intervention resulting in preterm birth . Am J Obstet Gynecol . 2006 ; 195 ( 6 ): 1557 – 1563 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Younis JS , Samueloff A. Gestational vascular complications . Best Pract Res Clin Haematol . 2003 ; 16 ( 2 ): 135 – 151 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Dempsey JC , Ashiny Z, Qiu CF, et al. . Maternal pre-pregnancy overweight status and obesity as risk factors for cesarean delivery . J Matern Fetal Neonatal Med . 2005 ; 17 ( 3 ): 179 – 185 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Dempsey JC , Butler CL, Sorensen TK, et al. . A case-control study of maternal recreational physical activity and risk of gestational diabetes mellitus . Diabetes Res Clin Pract . 2004 ; 66 ( 2 ): 203 – 215 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Dempsey JC , Butler CL, Williams MA. No need for a pregnant pause: physical activity may reduce the occurrence of gestational diabetes mellitus and preeclampsia . Exerc Sport Sci Rev . 2005 ; 33 ( 3 ): 141 – 149 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Dempsey JC , Sorensen TK, Williams MA, et al. . Prospective study of gestational diabetes mellitus risk in relation to maternal recreational physical activity before and during pregnancy . Am J Epidemiol . 2004 ; 159 ( 7 ): 663 – 670 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Brown J , Ceysens G, Boulvain M. Exercise for pregnant women with gestational diabetes for improving maternal and fetal outcomes . Cochrane Database Syst Rev . 2017 ; 6 : CD012202 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 19 Elsasser DA , Ananth CV, Prasad V, et al. . Diagnosis of placental abruption: relationship between clinical and histopathological findings . Eur J Obstet Gynecol Reprod Biol . 2010 ; 148 ( 2 ): 125 – 130 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Maclure M . The case-crossover design: a method for studying transient effects on the risk of acute events . Am J Epidemiol . 1991 ; 133 ( 2 ): 144 – 153 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Borg GA . Psychophysical bases of perceived exertion . Med Sci Sports Exerc . 1982 ; 14 ( 5 ): 377 – 381 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Mittleman MA , Maclure M, Robins JM. Control sampling strategies for case-crossover studies: an assessment of relative efficiency . Am J Epidemiol . 1995 ; 142 ( 1 ): 91 – 98 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Rothman KJ , Greenland S, Lash TL. Modern Epidemiology . 3rd ed. Philadelphia, PA : Wolters Kluwer Health/Lippincott Williams & Wilkins ; 2008 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 24 Cohen LS , Rosenbaum JF, Heller VL. Panic attack-associated placental abruption: a case report . J Clin Psychiatry . 1989 ; 50 ( 7 ): 266 – 267 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 25 Eriksen G , Wohlert M, Ersbak V, et al. . Placental abruption. A case-control investigation . Br J Obstet Gynaecol . 1991 ; 98 ( 5 ): 448 – 452 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Michikawa T , Morokuma S, Yamazaki S, et al. . Air pollutant exposure within a few days of delivery and placental abruption in Japan . Epidemiology . 2017 ; 28 ( 2 ): 190 – 196 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Mustafic H , Jabre P, Caussin C, et al. . Main air pollutants and myocardial infarction: a systematic review and meta-analysis . JAMA . 2012 ; 307 ( 7 ): 713 – 721 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Shah AS , Lee KK, McAllister DA, et al. . Short term exposure to air pollution and stroke: systematic review and meta-analysis . BMJ . 2015 ; 350 : h1295 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Chahal HS , Mostofsky E, Mittleman MA, et al. . Aerobic fitness and risk of ventricular arrhythmia following physical exertion . Can J Cardiol . 2016 ; 32 ( 4 ): 533 – 538 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Mittleman MA , Maclure M, Tofler GH, et al. . Triggering of acute myocardial infarction by heavy physical exertion. Protection against triggering by regular exertion. Determinants of Myocardial Infarction Onset Study Investigators . N Engl J Med . 1993 ; 329 ( 23 ): 1677 – 1683 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Mostofsky E , Laier E, Levitan EB, et al. . Physical activity and onset of acute ischemic stroke: the stroke onset study . Am J Epidemiol . 2011 ; 173 ( 3 ): 330 – 336 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Borodulin KM , Evenson KR, Wen F, et al. . Physical activity patterns during pregnancy . Med Sci Sports Exerc . 2008 ; 40 ( 11 ): 1901 – 1908 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Bagaria SJ , Bagaria VB. Strategies for diagnosis and prevention of venous thromboembolism during pregnancy . J Pregnancy . 2011 ; 2011 : 206858 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Nath CA , Ananth CV, Smulian JC, et al. . Histologic evidence of inflammation and risk of placental abruption . Am J Obstet Gynecol . 2007 ; 197 ( 3 ): 319.e1 – 319.e6 . Google Scholar Crossref Search ADS WorldCat Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018.

A, Baylin,;H, Guyer,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy136pmid: 29992316

Abstract Chahal et al. (Am J Epidemiol. 2018;187(10):2073–2079) assessed the risk of placental abruption due to physical exertion using a case-crossover design. The authors found an increased risk of placental abruption following increased physical exertion in the hour prior to the abruption. The risk was greater among women who were primarily sedentary during pregnancy or prior to becoming pregnant compared with those who were more physically active. The authors used a case-crossover design to assess the association of an intermittent exposure on an acute event. Chahal et al. address the limitations of the study, including the inability to control for time-varying confounders as well as the potential for recall bias. The public health implications of the study must be carefully evaluated given that physical activity prior to and during pregnancy can lead to healthy outcomes and is likely recommended. While the current study is unable to determine the type of physical exertion associated with placental abruption, future studies are recommended to determine the type of activity that presents increased risk. Additionally, studies among larger samples and in other countries will help determine the generalizability of the results. abruptio placentae, case-crossover, exercise, physical exertion, placental abruption In this issue of the Journal, Chahal et al. (1) report an interesting association between physical exertion and placental abruption. The risk of placental abruption was higher among women engaging in moderate or heavy physical exertion in the hour before the placental abruption than among women who engaged in light or no physical exertion. Furthermore, this association was modified by habitual physical activity, with those women who habitually engaged in moderate/heavy physical activity having a lower risk (although not null) than those who did not habitually engage in physical activity in the year prior to the pregnancy. There was also potential effect modification by the presence of preeclampsia/eclampsia. This is a novel report; to the best of our knowledge, there are no prior case-crossover studies on physical exertion and placental abruption. Case-crossover studies have been used widely to study mostly intermittent exposures that have an immediate and transient effect on acute events (2). The case-crossover study design was first applied to the study of triggers of myocardial infarction (3). Interestingly, Chahal et al. made an intriguing parallel with cardiovascular disease and argued that, although unknown, similar mechanisms for the triggering of the acute event could be shared between myocardial infarction and placental abruption (1). If so, other factors that have been reported to act as triggers of myocardial infarction could also have similar effects on placental abruption, such as sexual activity, anger, or cocaine use, to name a few (4). Another interesting parallel with myocardial infarction is the effect modification by habitual physical activity. The associations between heavy physical exertion and myocardial infarction tend to be highly attenuated, or even null, among physically active people. Nonetheless, although the risk of placental abruption was much lower among women who habitually engage in physical activity compared with that of sedentary women, it was still not negligible (rate ratio = 3.0, 95% confidence interval: 1.6, 5.9). Additionally, we can consider the number of preventable cases of placental abruption due to heavy physical exertion given the elevated overall relative risk detected (rate ratio = 7.8, 95% confidence interval: 5.5, 11.0). Although the study is not exempt from limitations, the authors are to be commended for how well they address the potential limitations in their analysis and discussion. Most of the limitations are inherent to the nature of the study design. Time-varying confounders, in particular concurrent triggers, cannot be easily addressed in case-crossover studies. However, the authors conducted sensitivity analyses to rule out this possibility. More importantly, recall bias can clearly bias the results if there is differential between reporting of the hazard and control periods. In order to minimize the recall bias, women were not informed of the duration of the hypothesized hazard period and were asked sequentially about their activity during different periods of time prior to the placental abruption and prior to their pregnancy. However, it is still possible that women who report being exposed to heavy physical exertion before the outcome during the hazard period may tend to report no exposure in the control period more often, particularly among sedentary women. Pregnancy is an impactful point in a women’s life given that there are known health risks and complications for both the mother and the child. It is possible that women with placental abruption were more likely to report exercise that was out of the ordinary for them as heavy physical exertion rather than low exertion, as they search for possible reasons that the high-risk event occurred. This type of differential reporting could lead to an overestimate of the risk among usually sedentary women compared with those who engaged in physical activity prior to the pregnancy. Placental abruption is a dramatic, acute event during pregnancy and is associated with significant maternal morbidity and perinatal morbidity and mortality (5). These severe consequences may be even more drastic in developing countries where access to health care during pregnancy is usually more limited than in developed countries. The public health implications of these findings, however, must be evaluated very carefully, particularly in light of the known benefits of habitual physical activity during and prior to pregnancy. There is a high risk for these findings to be wrongly translated to the public. As Chahal et al. point out, and analogous to the case of myocardial infarction, habitual physical activity may provide a protective effect that offsets the transient hazardous association of moderate and heavy physical exertion on placental abruption risk. However, Chahal et al. recognize that this could not be assessed in the current study given the lack of a traditional control group (1). Until further studies can be done to evaluate this question, some caution is warranted to avoid sending the wrong message to all pregnant women. In the meantime, it may be important to emphasize that engaging in regular physical activity before pregnancy and maintaining a reasonable level of activity during pregnancy—while avoiding strenuous activities—may be best, particularly among women at high risk of placental abruption. However, in doing so, we should also consider the type of strenuous physical activity that places a woman at risk for placental abruption, again so as not to mislead the public of the risks of physical activity during pregnancy. Furthermore, the results from case-crossover studies should always be evaluated in the context of the absolute risk of having the outcome at any given moment and the frequency and duration of the trigger, because the reported point estimates are relative risks for a transient exposure. In the case of placental abruption and physical activity, given the relatively short window of occurrence of the outcome during pregnancy and the relatively high frequency and duration of the exposure, this may not be that critical. Finally, for some acute outcomes, avoidance of the trigger may avoid the outcome completely. For example, in the case of cellular phone use and road traffic accidents, if people do not use a cell phone while driving, they will not have the potential accident triggered by using the phone. However, for other outcomes, avoidance of the trigger may only delay the outcome. For example, if people avoid heavy physical exertion, they may delay the acute event of myocardial infarction, but at some point the vulnerable atherosclerotic plaque may be broken by any other factor that increases sympathetic activity (2). It is interesting to speculate what the case would be for placental abruption, given the relatively short window of susceptibility for this outcome. Can cases of placental abruption be avoided by avoiding the exposure, or will they just be deferred by a few days? In terms of future directions, some of the limitations addressed by the authors could be evaluated in future studies. Sample size can definitely be increased, and that could also help to explore other hypotheses that the authors were not able to explore in this particular study, such as whether risk varies according to vaginal bleeding during pregnancy, hypertensive disorders of pregnancy, and prior placental abruption, as Chahal et al. comment in the discussion (1). Also, type of physical activity remains an important question mark and one that could be explored in future studies. While a randomized controlled trial in which women are assigned to high-exertion physical activity, or various levels of physical activity, versus a sedentary pregnancy is not feasible, or ethical, observational studies of the general and at-risk populations starting at the first prenatal visit would likely be useful. The fact that the study population in Chahal et al. is from a developing country may affect some of the generalizability regarding the type of physical activity. Future studies could be conducted in countries in which rates of physical activity are higher in order to determine whether the association still exists and to determine the type, duration, frequency, and intensity of physical activity that is protective during pregnancy. It could be possible that some vigorous activities induce more risk than others, and that those vigorous activities could be more prevalent in this study population, such as heavy lifting. This remains speculative, because we are not certain whether there are differences according to type of physical activity, but investigation of these findings in other populations in future studies is definitively warranted. Finally, although other triggers, such as heat and air pollution (6, 7), have been explored with a similar epidemiologic design, many more could be addressed in future studies of placental abruption, either those triggers that have been explored for myocardial infarction, assuming that similar biological mechanisms may be in place, or others that involve other potential biological mechanisms linked specifically to this pregnancy outcome. Similarly, the literature on triggers of other pregnancy outcomes is still relatively scarce, and the case-crossover design is definitely a useful tool to explore these associations in more detail. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan (Ana Baylin, Heidi Guyer); Department of Nutritional Sciences, School of Public Health, University of Michigan, Ann Arbor, Michigan (Ana Baylin); and Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, Michigan (Heidi Guyer). Conflict of interest: none declared. REFERENCES 1 Chahal HS , Gelaye B , Mostofsky E , et al. . Physical exertion immediately prior to placental abruption: a case-crossover study . Am J Epidemiol . 2018 ; 187 ( 10 ): 2073 – 2079 . 2 Maclure M , Mittleman MA . Should we use a case-crossover design? Annu Rev Public Health . 2000 ; 21 : 193 – 221 . Google Scholar Crossref Search ADS PubMed 3 Maclure M . The case-crossover design: a method for studying transient effects on the risk of acute events . Am J Epidemiol . 1991 ; 133 ( 2 ): 144 – 153 . Google Scholar Crossref Search ADS PubMed 4 Servoss SJ , Januzzi JL , Muller JE . Triggers of acute coronary syndromes . Prog Cardiovasc Dis . 2002 ; 44 ( 5 ): 369 – 380 . Google Scholar Crossref Search ADS PubMed 5 Downes KL , Grantz KL , Shenassa ED . Maternal, labor, delivery, and perinatal outcomes associated with placental abruption: a systematic review . Am J Perinatol . 2017 ; 34 ( 10 ): 935 – 957 . Google Scholar Crossref Search ADS PubMed 6 He S , Kosatsky T , Smargiassi A , et al. . Heat and pregnancy-related emergencies: risk of placental abruption during hot weather . Environ Int . 2018 ; 111 : 295 – 300 . Google Scholar Crossref Search ADS PubMed 7 Michikawa T , Morokuma S , Yamazaki S , et al. . Air pollutant exposure within a few days of delivery and placental abruption in Japan . Epidemiology . 2017 ; 28 ( 2 ): 190 – 196 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

S, Chahal, Harpreet;Bizu, Gelaye,;A, Williams, Michelle

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy137pmid: 29992300

We appreciate the thoughtful comments by Baylin and Guyer (1) on our article, in which we evaluated the acute risk of placental abruption following physical exertion (2). Their commentary raises important issues on the etiology of placental abruption, the clinical implications of our work, and the most appropriate public health message regarding physical activity during pregnancy. Baylin and Guyer suggest a few interesting precipitants of myocardial infarction that may also increase the acute risk of placental abruption, such as cocaine use, sexual activity, and anger. We agree that these are important questions for future studies to investigate. Cocaine use over the course of pregnancy increases the baseline risk of placental abruption as well as other adverse obstetrical outcomes (3). Given the fast-acting vasoconstrictive effects of cocaine, it is possible that its use would also increase the immediate risk of placental abruption. The association of sexual activity and a transiently increased risk of cardiovascular events has been well-documented using the case-crossover approach (4), and we think this exposure would be particularly interesting for obstetrical outcomes; not only does sexual activity include physical exertion, but vaginal-penile intercourse may cause oxytocin release and uterine contractions from direct stimulation of the lower uterine segment, orgasm, and the high prostaglandin content found in semen (5). While we present an increase in the acute risk of placental abruption following physical stress, and others have found a higher risk following exposure to chemical stress (6, 7), the immediate effect following psychological stressors is, to our knowledge, unstudied. Anger is another important exposure because its physiologic effects are similar to those of heavy physical exertion (8). A case-report suggests that panic may be a potential stressor as well (9). Baylin and Guyer comment that the risk associated with episodes of physical exertion among women who were less sedentary in our study was not null but hypothesize that a null association might be found among women who are more physically active, similar to prior studies on myocardial infarction. We think this is possible as well, given that the circulating level of catecholamines correlate better with the relative intensity of exertion than with the absolute intensity and that catecholamine release is reduced after exercise training (10). Finally, we agree that the public health implications of these findings need to be addressed thoughtfully. In this case-only design, we studied 663 women who had had placental abruption and were therefore, by definition, delivering high-risk pregnancies. The majority of pregnancies do not experience this rare complication, and women with uncomplicated pregnancies ought to be encouraged to engage in safe levels of physical activity before, during, and after pregnancy. As we stated in our study, although we did not compare women who had placental abruption with those who did not, we also hypothesized that women who engaged in more physical activity would have a lower total risk of placental abruption despite the transient increase in risk associated with each episode of exertion, similar to the relationship between physical exertion and ischemic cardiovascular events. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Harpreet S. Chahal, Bizu Gelaye, Michelle A. Williams); Mississauga Academy of Medicine, University of Toronto Mississauga, Mississauga, Canada (Harpreet S. Chahal). Conflict of interest: none declared. REFERENCES 1 Baylin A , Guyer H . Invited commentary: physical exertion and placental abruption—public health implications and future directions . Am J Epidemiol . 2018 ; 187 ( 10 ): 2080 – 2082 . 2 Chahal HS , Gelaye B , Mostofsky E , et al. . Physical exertion immediately prior to placental abruption: a case-crossover study . Am J Epidemiol . 2018 ; 187 ( 10 ): 2073 – 2079 . 3 Oyelese Y , Ananth CV . Placental abruption . Obstet Gynecol . 2006 ; 108 ( 4 ): 1005 – 1016 . Google Scholar Crossref Search ADS PubMed 4 Dahabreh IJ , Paulus JK . Association of episodic physical and sexual activity with triggering of acute cardiac events: systematic review and meta-analysis . JAMA . 2011 ; 305 ( 12 ): 1225 – 1233 . Google Scholar Crossref Search ADS PubMed 5 Kavanagh J , Kelly AJ , Thomas J . Sexual intercourse for cervical ripening and induction of labour . Cochrane Database Syst Rev . 2001 ;( 2 ): CD003093 . 6 Ananth CV , Kioumourtzoglou MA , Huang Y , et al. . Exposures to air pollution and risk of acute-onset placental abruption: a case-crossover study . Epidemiology . 2018 ; 29 ( 5 ): 631 – 638 . Google Scholar Crossref Search ADS PubMed 7 Michikawa T , Morokuma S , Yamazaki S , et al. . Air pollutant exposure within a few days of delivery and placental abruption in Japan . Epidemiology . 2017 ; 28 ( 2 ): 190 – 196 . Google Scholar Crossref Search ADS PubMed 8 Smyth A , O’Donnell M , Lamelas P , et al. . Physical activity and anger or emotional upset as triggers of acute myocardial infarction: the INTERHEART study . Circulation . 2016 ; 134 ( 15 ): 1059 – 1067 . Google Scholar Crossref Search ADS PubMed 9 Cohen LS , Rosenbaum JF , Heller VL . Panic attack-associated placental abruption: a case report . J Clin Psychiatry . 1989 ; 50 ( 7 ): 266 – 267 . Google Scholar PubMed 10 Kohrt WM , Spina RJ , Ehsani AA , et al. . Effects of age, adiposity, and fitness level on plasma catecholamine responses to standing and exercise . J Appl Physiol (1985) . 1993 ; 75 ( 4 ): 1828 – 1835 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Anja, Schoeps,;van Ewijk, Reyn, ;Gisela, Kynast-Wolf,;Eric, Nebié,;Pascal, Zabré,;Ali, Sié,;Sabine, Gabrysch,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy091pmid: 29741574

Abstract Ramadan exposure in utero can be regarded as a natural experiment with which to study how nutritional conditions in utero influence susceptibility to disease later in life. We analyzed data from rural Burkina Faso on 41,025 children born between 1993 and 2012, of whom 25,093 were born to Muslim mothers. Ramadan exposure was assigned on the basis of overlap between Ramadan dates and gestation, creating 7 exclusive categories. We used proportional hazards regression with difference-in-differences analysis to estimate the association between Ramadan exposure at different gestational ages and mortality among children under 5 years of age. Under-5 mortality was 32 deaths per 1,000 child-years. Under-5 mortality among Muslims was 15% higher than that among non-Muslims (P < 0.001). In the difference-in-differences analysis, the occurrence of Ramadan during conception or the first or second trimester was associated with higher under-5 mortality rates among Muslims only. The mortality rates of children born to Muslim mothers were 33%, 29%, and 22% higher when Ramadan occurred during conception, the first trimester, and the second trimester, respectively, compared with children of non-Muslim mothers born at the same time (P = 0.01, P < 0.001, and P = 0.007). Having a Muslim mother was not associated with mortality when the child was not exposed to Ramadan, born during Ramadan, or exposed during the third trimester. Observance of Ramadan during early pregnancy can have detrimental consequences for the future health of the unborn child. child mortality, difference in differences, population-based cohort, prenatal exposure delayed effects, pregnancy, Ramadan, sub-Saharan Africa Editor’s note: Invited commentaries on this article appear on pages 2093 and 2095, and the authors’ response appears on page 2098. The first 9 months of life in utero are a crucial period for human development in which future health trajectories are set, determining physical and cognitive capabilities. Early pregnancy is a time when the developing organism is particularly vulnerable, as this is when organogenesis takes place, including development of the neural and immune systems and the shaping of endocrine and metabolic pathways (1). It is well established that some infections (including rubella virus), certain chemicals and drugs (most famously thalidomide), and deficiency in certain nutrients (e.g., folate and iodine) in early pregnancy can lead to serious fetal damage (2). There is also a growing body of evidence that nutrients, environmental chemicals, drugs, and infections can have more subtle effects on the developing organism in utero, presumably through epigenetic mechanisms. Such effects may show only later in life, a concept referred to as “developmental origins of health and disease” (3–5). The majority of studies on the topic have come from high- and middle-income countries and have focused on chronic diseases in later adulthood, such as diabetes and cardiovascular disease (6). There have been very few studies in low-income settings, although it would be of great interest to better understand developmental impacts on the immune system and infection-related mortality (7). When studying long-term effects of undernutrition in utero, researchers have employed natural experiments such as famines to avoid confounding by socioeconomic status and other factors (2, 8). The most famous is the study of the Dutch Hunger Winter of World War II, which showed that exposure to famine during gestation, especially early gestation, increases the risk of obesity, diabetes, coronary heart disease, breast cancer, mental illness, and cognitive decline later in life and increases later-life mortality rates (9–13). Other research has focused on other famines occurring during World War II, on 19th century famines, and on more recent famines such as those that took place during the Chinese Great Leap Forward (see Lumey et al. (8) for a review). One study that—like our study—focused on the association between nonfamine malnutrition and mortality exploited seasonal patterns of food availability in the Gambia, linking season of birth to smaller thymus size in children and to mortality in young adults (7). Another very particular natural experiment for studying the effects of undernutrition in utero is furnished by the Muslim tradition of daytime fasting during the holy month of Ramadan. The rotation of Ramadan through the year and the contrast with non-Muslim populations in the same location provides an opportunity to disentangle seasonal effects on fetal development (due to different availability of nutrients, as well as environmental and infectious exposures) from Ramadan effects. Studying the fetal health impacts of intermittent maternal fasting, as occurs during Ramadan, would shed light on potential long-term effects of relatively mild shocks experienced more commonly. Furthermore, effects of Ramadan fasting are relevant in their own right, given the high numbers of people affected each year. Most studies of intrauterine Ramadan exposure have focused on short-term outcomes such as low birth weight or preterm delivery, while fewer have looked beyond the immediate outcomes and studied longer-term outcomes, including cognitive abilities, body composition, symptoms of coronary heart disease and type 2 diabetes mellitus, and reduced work performance (14–20). To our knowledge, there have been no studies so far on the association between Ramadan exposure in utero and mortality outcomes in childhood or adulthood. We hypothesized that maternal observance of Ramadan during conception and early pregnancy may have negative impacts on the development of the fetus, leading to higher child mortality, especially in settings with high background levels of undernutrition among women and high infection-related child mortality. The aim of our study was to assess the influence of Ramadan exposure in utero at different gestational ages on mortality among children under 5 years of age in Burkina Faso. METHODS Study area and population This population-based cohort study was conducted using data from the Nouna Health and Demographic Surveillance System (HDSS) in northwestern Burkina Faso, which in 2012 comprised a population of nearly 100,000 inhabitants. The Nouna HDSS was established by an initial census carried out in 39 villages in 1992 (21). In 2000, 2 additional villages and the town of Nouna were included, and another 17 villages followed in 2004. The town of Nouna is a semiurban settlement with about 30,000 inhabitants, and the rural area around Nouna can be divided into 4 regions (22). The majority of inhabitants in the Nouna HDSS are Muslims (63%), while 32% are Christians, and a minority of 5% holds traditional beliefs. The study area is characterized by a sub-Sahelian climate with a dry season (November–May) and a rainy season (June–October). The majority of the population lives by subsistence farming; literacy rates are low, and there is seasonal food insecurity. More than one-quarter of children between 6 and 31 months of age are wasted (21, 23). Ramadan exposure During the month of Ramadan, which lasts about 30 days, healthy adult Muslims are obliged to abstain from eating and drinking (even water) between dawn and sunset. The Islamic lunar calendar is approximately 11 days shorter than the Gregorian calendar, so Ramadan occurs slightly earlier each year and rotates through the Gregorian year in about 33 years. For sick people and pregnant women, fasting during Ramadan is voluntary; however, they are supposed to make up for this exemption by fasting afterwards (24). While Ramadan is likely to affect fetal health through the fasting itself, other lifestyle changes taking place during Ramadan (including increased sugar intake and changes in sleeping patterns) may also be responsible for the association. Data from the Nouna HDSS do not contain information on individual Ramadan fasting, and we do not distinguish between the potential channels for associations. We assigned Ramadan exposure in utero based on a calculation of whether there was overlap between Ramadan and gestation, utilizing the date of birth in relation to the Ramadan dates and assuming a gestational period of 266 days, which is the average duration of a full-term human pregnancy. Our approach can be interpreted as an intention-to-treat analysis that helps avoid confounding due to self-selection of women into Ramadan-observant and nonobservant groups, which likely also differ in other ways. We differentiated Ramadan exposure into 7 exclusive categories: certainly not exposed, probably not exposed, conceived during Ramadan, Ramadan starting during the first, second, or third trimester, and born during Ramadan. Children in the category “probably not exposed” would not have experienced Ramadan in utero if their mothers had a normal 266-day pregnancy, as this would have started just after Ramadan. However, if they were born up to 20 days postterm, they could still have been exposed to Ramadan during conception (Figure 1). Figure 1. View largeDownload slide Assignment to prenatal Ramadan exposure by time of birth for children born in Nouna District, Burkina Faso, in 2011–2012. Assuming a gestation period of 266 days, one of 7 Ramadan exposure categories was assigned to each child. The asterisks (*) indicate birth dates, while black dots show the calculated dates of conception. The black boxes indicate the timing of Ramadan in 2010 and 2011. Mar, March; Jul, July; Nov, November. Figure 1. View largeDownload slide Assignment to prenatal Ramadan exposure by time of birth for children born in Nouna District, Burkina Faso, in 2011–2012. Assuming a gestation period of 266 days, one of 7 Ramadan exposure categories was assigned to each child. The asterisks (*) indicate birth dates, while black dots show the calculated dates of conception. The black boxes indicate the timing of Ramadan in 2010 and 2011. Mar, March; Jul, July; Nov, November. Statistical analysis Between 1993 and 2012, a total of 48,747 children were born alive in the Nouna HDSS. We excluded 7,520 children for whom the exact date of birth was unknown and 202 children for whom information on mother’s religion was missing; this resulted in a sample of 41,025 children for analysis. We calculated mortality rates in children of Muslim and non-Muslim mothers overall and stratified by the occurrence of Ramadan during pregnancy. Because our study period included 20 years of observation, Ramadan did not fully rotate through the year (from February/March in 1992 to July/August in 2012), and thus Ramadan exposure varied by month of birth (see Web Figure 1, available at https://academic.oup.com/aje). We used a difference-in-differences analysis, comparing the association of Ramadan occurrence during pregnancy with mortality between non-Muslims and Muslims, under the assumption of a common seasonal trend. The inclusion of non-Muslims in the model served to purge any remaining seasonal mortality differences correlated with Ramadan that remained after adjustment for calendar month of birth (14). For details on the model assumptions and sensitivity analyses, refer to Web Appendix 1. We used 2-level multivariate Weibull proportional hazards regression to estimate the association between Ramadan exposure in utero and child survival up to age 5 years, with random intercepts at the mother level. Death between birth and 5 years of age was the event of interest. The time variable was time from birth to death, age 5 years, loss to follow-up, or the end of the study (December 31, 2012). After confirmation of the proportional hazards assumption for the main variables of interest through the use of Kaplan-Meier graphs, we first fitted a regression model without considering Ramadan timing, to estimate the association between religion and child mortality, using a binary variable indicating whether the mother was Muslim or non-Muslim. In a second regression model, we then added a variable indicating Ramadan occurrence during pregnancy, as well as a term for the interaction of Ramadan occurrence with the Muslim indicator variable. These interaction parameters were the coefficients of interest in our study. They can be interpreted as a comparison of mortality between exposed and unexposed Muslims after correcting for seasonality (by comparing with non-Muslims). Adjustment was made for year of birth (continuous variable), calendar month of birth, and region of residence. Stata version 14 (StataCorp LLC, College Station, Texas) was used for analysis, and SAS version 9.3 (SAS Institute, Inc., Cary, North Carolina) was used for creating the figures. All P values are 2-sided. In addition, stratified analyses were conducted for the 5 main regions of the study area (22) and for 3 different study periods. Besides the main outcome of under-5 mortality, we also studied infant mortality and mortality in children aged 1–4 years. To investigate in more detail the association between being conceived during Ramadan and mortality, we created 2 separate categories: exposure for at least 14 days and exposure for less than 14 days. RESULTS Our study included 41,025 children who were born to 20,709 mothers in the Nouna HDSS between January 1993 and December 2012, for whom mother’s religion and the precise date of birth were known. The majority of mothers in this sample were Muslim (61%), while 28% were Catholic, 5% were Protestant, and 6% had an animist religion or another religion. Between 1993 and 2012, there were 4,213 deaths among children under 5 years of age during 133,203 child-years at risk, corresponding to an overall mortality rate of 31.6 deaths per 1,000 child-years. Mortality rates were higher among Muslims (33.5 deaths per 1,000 child-years) than among non-Muslims (28.7 deaths per 1,000 child-years) (Table 1). Among non-Muslims, mortality rates were somewhat lower in children who were in utero during Ramadan as compared with children who were not. This was most likely due to seasonal variations in mortality that were correlated with Ramadan occurrence in our sample and that were previously described in the study area (25). This pattern was not apparent for children of Muslim mothers. Table 1. Numbers of Live Births, Numbers of Child Deaths, and Mortality Before 5 Years of Age, by Prenatal Ramadan Exposure and Religion, Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Ramadan Exposure Category Muslims Non-Muslims Live Births No. of Child-Years No. of Child Deaths U5M Ratea Live Births No. of Child-Years No. of Child Deaths U5M Ratea No. % No. % Certainly not exposed 2,837 11.3 9,263 297 32.1 1,936 12.2 6,488 208 32.1 Probably not exposed 1,467 5.9 4,778 169 35.4 933 5.9 3,124 99 31.7 Conceived during Ramadan 2,150 8.6 6,983 252 36.1 1,338 8.4 4,485 119 26.5 Exposed during trimester 1 5,981 23.8 20,070 681 33.9 3,573 22.4 12,122 319 26.3 Exposed during trimester 2 6,058 24.1 19,183 632 32.9 3,791 23.8 11,960 319 26.7 Exposed during trimester 3 4,261 17.0 13,248 425 32.1 2,755 17.3 8,650 252 29.1 Born during Ramadan 2,339 9.3 7,586 262 34.5 1,606 10.1 5,263 179 34.0 Total 25,093 81,111 2,718 33.5 15,932 52,092 1,495 28.7 Ramadan Exposure Category Muslims Non-Muslims Live Births No. of Child-Years No. of Child Deaths U5M Ratea Live Births No. of Child-Years No. of Child Deaths U5M Ratea No. % No. % Certainly not exposed 2,837 11.3 9,263 297 32.1 1,936 12.2 6,488 208 32.1 Probably not exposed 1,467 5.9 4,778 169 35.4 933 5.9 3,124 99 31.7 Conceived during Ramadan 2,150 8.6 6,983 252 36.1 1,338 8.4 4,485 119 26.5 Exposed during trimester 1 5,981 23.8 20,070 681 33.9 3,573 22.4 12,122 319 26.3 Exposed during trimester 2 6,058 24.1 19,183 632 32.9 3,791 23.8 11,960 319 26.7 Exposed during trimester 3 4,261 17.0 13,248 425 32.1 2,755 17.3 8,650 252 29.1 Born during Ramadan 2,339 9.3 7,586 262 34.5 1,606 10.1 5,263 179 34.0 Total 25,093 81,111 2,718 33.5 15,932 52,092 1,495 28.7 Abbreviation: U5M, under-5 mortality. a Number of deaths per 1,000 child-years. Table 1. Numbers of Live Births, Numbers of Child Deaths, and Mortality Before 5 Years of Age, by Prenatal Ramadan Exposure and Religion, Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Ramadan Exposure Category Muslims Non-Muslims Live Births No. of Child-Years No. of Child Deaths U5M Ratea Live Births No. of Child-Years No. of Child Deaths U5M Ratea No. % No. % Certainly not exposed 2,837 11.3 9,263 297 32.1 1,936 12.2 6,488 208 32.1 Probably not exposed 1,467 5.9 4,778 169 35.4 933 5.9 3,124 99 31.7 Conceived during Ramadan 2,150 8.6 6,983 252 36.1 1,338 8.4 4,485 119 26.5 Exposed during trimester 1 5,981 23.8 20,070 681 33.9 3,573 22.4 12,122 319 26.3 Exposed during trimester 2 6,058 24.1 19,183 632 32.9 3,791 23.8 11,960 319 26.7 Exposed during trimester 3 4,261 17.0 13,248 425 32.1 2,755 17.3 8,650 252 29.1 Born during Ramadan 2,339 9.3 7,586 262 34.5 1,606 10.1 5,263 179 34.0 Total 25,093 81,111 2,718 33.5 15,932 52,092 1,495 28.7 Ramadan Exposure Category Muslims Non-Muslims Live Births No. of Child-Years No. of Child Deaths U5M Ratea Live Births No. of Child-Years No. of Child Deaths U5M Ratea No. % No. % Certainly not exposed 2,837 11.3 9,263 297 32.1 1,936 12.2 6,488 208 32.1 Probably not exposed 1,467 5.9 4,778 169 35.4 933 5.9 3,124 99 31.7 Conceived during Ramadan 2,150 8.6 6,983 252 36.1 1,338 8.4 4,485 119 26.5 Exposed during trimester 1 5,981 23.8 20,070 681 33.9 3,573 22.4 12,122 319 26.3 Exposed during trimester 2 6,058 24.1 19,183 632 32.9 3,791 23.8 11,960 319 26.7 Exposed during trimester 3 4,261 17.0 13,248 425 32.1 2,755 17.3 8,650 252 29.1 Born during Ramadan 2,339 9.3 7,586 262 34.5 1,606 10.1 5,263 179 34.0 Total 25,093 81,111 2,718 33.5 15,932 52,092 1,495 28.7 Abbreviation: U5M, under-5 mortality. a Number of deaths per 1,000 child-years. To estimate the association between Ramadan exposure in utero and child mortality, we performed a difference-in-differences analysis to eliminate seasonal mortality differences through comparison with non-Muslims. In adjusted multilevel regression, children for whom Ramadan occurred during conception or the first or second trimester showed higher mortality rates (37%, 33%, and 25%, respectively) than children certainly not exposed, only among Muslims, as shown by the interaction terms (P = 0.03, P = 0.01, and P = 0.05, respectively), while exposure during the third trimester and at birth was not associated with increased mortality (Table 2). Table 2. Effect of Prenatal Ramadan Exposure on Mortality Before 5 Years of Age (Difference-in-Differences Analysis) Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Parametera and Exposure Category HR 95% CI P Value Ramadan  Conceived vs. certainly not 0.86 0.69, 1.09 0.22  Trimester 1 vs. certainly not 0.88 0.73, 1.06 0.17  Trimester 2 vs. certainly not 0.91 0.76, 1.10 0.35  Trimester 3 vs. certainly not 0.92 0.76, 1.11 0.37  Born vs. certainly not 1.05 0.86, 1.29 0.63  Probably not vs. certainly not 0.99 0.77, 1.26 0.92 Mother is Muslim (yes vs. no) 0.97 0.81, 1.17 0.77 Ramadan × Muslim mother interactionb  Conceived vs. certainly not 1.37 1.03, 1.82 0.03  Trimester 1 vs. certainly not 1.33 1.06, 1.67 0.01  Trimester 2 vs. certainly not 1.25 1.00, 1.57 0.05  Trimester 3 vs. certainly not 1.10 0.87, 1.40 0.43  Born vs. certainly not 1.02 0.79, 1.34 0.86  Probably not vs. certainly not 1.14 0.83, 1.56 0.42 Year of birth (linear; per year since 1993) 0.96 0.95, 0.96 <0.001 Region of residence  Central vs. Nouna town 1.39 1.22, 1.58 <0.001  Northeast vs. Nouna town 1.94 1.74, 2.16 <0.001  Southeast vs. Nouna town 1.24 1.09, 1.40 0.001  Southwest vs. Nouna town 1.88 1.69, 2.09 <0.001 Parametera and Exposure Category HR 95% CI P Value Ramadan  Conceived vs. certainly not 0.86 0.69, 1.09 0.22  Trimester 1 vs. certainly not 0.88 0.73, 1.06 0.17  Trimester 2 vs. certainly not 0.91 0.76, 1.10 0.35  Trimester 3 vs. certainly not 0.92 0.76, 1.11 0.37  Born vs. certainly not 1.05 0.86, 1.29 0.63  Probably not vs. certainly not 0.99 0.77, 1.26 0.92 Mother is Muslim (yes vs. no) 0.97 0.81, 1.17 0.77 Ramadan × Muslim mother interactionb  Conceived vs. certainly not 1.37 1.03, 1.82 0.03  Trimester 1 vs. certainly not 1.33 1.06, 1.67 0.01  Trimester 2 vs. certainly not 1.25 1.00, 1.57 0.05  Trimester 3 vs. certainly not 1.10 0.87, 1.40 0.43  Born vs. certainly not 1.02 0.79, 1.34 0.86  Probably not vs. certainly not 1.14 0.83, 1.56 0.42 Year of birth (linear; per year since 1993) 0.96 0.95, 0.96 <0.001 Region of residence  Central vs. Nouna town 1.39 1.22, 1.58 <0.001  Northeast vs. Nouna town 1.94 1.74, 2.16 <0.001  Southeast vs. Nouna town 1.24 1.09, 1.40 0.001  Southwest vs. Nouna town 1.88 1.69, 2.09 <0.001 Abbreviations: CI, confidence interval; HR, hazard ratio. a Model: difference-in-differences analysis with a random intercept for mother; results were additionally adjusted for month of birth (estimates not displayed). b Interaction terms. Table 2. Effect of Prenatal Ramadan Exposure on Mortality Before 5 Years of Age (Difference-in-Differences Analysis) Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Parametera and Exposure Category HR 95% CI P Value Ramadan  Conceived vs. certainly not 0.86 0.69, 1.09 0.22  Trimester 1 vs. certainly not 0.88 0.73, 1.06 0.17  Trimester 2 vs. certainly not 0.91 0.76, 1.10 0.35  Trimester 3 vs. certainly not 0.92 0.76, 1.11 0.37  Born vs. certainly not 1.05 0.86, 1.29 0.63  Probably not vs. certainly not 0.99 0.77, 1.26 0.92 Mother is Muslim (yes vs. no) 0.97 0.81, 1.17 0.77 Ramadan × Muslim mother interactionb  Conceived vs. certainly not 1.37 1.03, 1.82 0.03  Trimester 1 vs. certainly not 1.33 1.06, 1.67 0.01  Trimester 2 vs. certainly not 1.25 1.00, 1.57 0.05  Trimester 3 vs. certainly not 1.10 0.87, 1.40 0.43  Born vs. certainly not 1.02 0.79, 1.34 0.86  Probably not vs. certainly not 1.14 0.83, 1.56 0.42 Year of birth (linear; per year since 1993) 0.96 0.95, 0.96 <0.001 Region of residence  Central vs. Nouna town 1.39 1.22, 1.58 <0.001  Northeast vs. Nouna town 1.94 1.74, 2.16 <0.001  Southeast vs. Nouna town 1.24 1.09, 1.40 0.001  Southwest vs. Nouna town 1.88 1.69, 2.09 <0.001 Parametera and Exposure Category HR 95% CI P Value Ramadan  Conceived vs. certainly not 0.86 0.69, 1.09 0.22  Trimester 1 vs. certainly not 0.88 0.73, 1.06 0.17  Trimester 2 vs. certainly not 0.91 0.76, 1.10 0.35  Trimester 3 vs. certainly not 0.92 0.76, 1.11 0.37  Born vs. certainly not 1.05 0.86, 1.29 0.63  Probably not vs. certainly not 0.99 0.77, 1.26 0.92 Mother is Muslim (yes vs. no) 0.97 0.81, 1.17 0.77 Ramadan × Muslim mother interactionb  Conceived vs. certainly not 1.37 1.03, 1.82 0.03  Trimester 1 vs. certainly not 1.33 1.06, 1.67 0.01  Trimester 2 vs. certainly not 1.25 1.00, 1.57 0.05  Trimester 3 vs. certainly not 1.10 0.87, 1.40 0.43  Born vs. certainly not 1.02 0.79, 1.34 0.86  Probably not vs. certainly not 1.14 0.83, 1.56 0.42 Year of birth (linear; per year since 1993) 0.96 0.95, 0.96 <0.001 Region of residence  Central vs. Nouna town 1.39 1.22, 1.58 <0.001  Northeast vs. Nouna town 1.94 1.74, 2.16 <0.001  Southeast vs. Nouna town 1.24 1.09, 1.40 0.001  Southwest vs. Nouna town 1.88 1.69, 2.09 <0.001 Abbreviations: CI, confidence interval; HR, hazard ratio. a Model: difference-in-differences analysis with a random intercept for mother; results were additionally adjusted for month of birth (estimates not displayed). b Interaction terms. In the adjusted multilevel model that did not adjust for Ramadan occurrence during pregnancy, the mortality rate among Muslims was 15% higher (95% confidence interval: 7, 24) than that among non-Muslims (Table 3). From the model with Ramadan interaction terms, we calculated the mortality rate ratio for Muslims versus non-Muslims for each Ramadan exposure category. Children born to Muslim mothers who experienced Ramadan in utero during conception had a 33% higher mortality rate than children of non-Muslim mothers in utero during the same time period (hazard ratio (HR) = 0.97 for Muslim women vs. non-Muslim women outside Ramadan, multiplied with HR = 1.37 for the interaction; P = 0.01). For trimesters 1 and 2, the increases in mortality rates were 29% and 22% (P < 0.001 and P = 0.007), respectively (Table 3, Figure 2). Among children exposed to Ramadan in the third trimester, the difference in mortality between Muslims and non-Muslims was small, and there was no increased mortality for Muslim children born during Ramadan. Having a Muslim mother was not associated with mortality when not exposed to Ramadan (HR = 0.97, P = 0.74), and Ramadan occurrence was not associated with mortality among non-Muslims in the adjusted model (Table 2). Table 3. Mortality Before 5 Years of Age in Muslims Compared With Non-Muslims, by Prenatal Ramadan Exposure, Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Prenatal Ramadan Exposure Calculationa HR 95% CI P Value Total (not considering Ramadan)b 1.15 1.07, 1.24 <0.001 Ramadan exposure categoryc  Certainly not exposed —d 0.97 0.81, 1.17 0.74  Probably not exposed 0.97 × 1.14 1.11 0.86, 1.43 0.43  Conceived during Ramadan 0.97 × 1.37 1.33 1.06, 1.66 0.01  Ramadan during trimester 1 0.97 × 1.33 1.29 1.12, 1.49 <0.001  Ramadan during trimester 2 0.97 × 1.26 1.22 1.06, 1.40 0.007  Ramadan during trimester 3 0.97 × 1.11 1.07 0.91, 1.26 0.39  Born during Ramadan 0.97 × 1.02 1.00 0.82, 1.21 0.98 Prenatal Ramadan Exposure Calculationa HR 95% CI P Value Total (not considering Ramadan)b 1.15 1.07, 1.24 <0.001 Ramadan exposure categoryc  Certainly not exposed —d 0.97 0.81, 1.17 0.74  Probably not exposed 0.97 × 1.14 1.11 0.86, 1.43 0.43  Conceived during Ramadan 0.97 × 1.37 1.33 1.06, 1.66 0.01  Ramadan during trimester 1 0.97 × 1.33 1.29 1.12, 1.49 <0.001  Ramadan during trimester 2 0.97 × 1.26 1.22 1.06, 1.40 0.007  Ramadan during trimester 3 0.97 × 1.11 1.07 0.91, 1.26 0.39  Born during Ramadan 0.97 × 1.02 1.00 0.82, 1.21 0.98 Abbreviations: CI, confidence interval; HR, hazard ratio. a Calculation of stratum HRs from the baseline HR and the interaction HR. b Survival analysis without Ramadan exposure (with a random intercept for mother; results were adjusted for year of birth, month of birth, region of residence, season at death, and sex of child). c Based on the full model shown in Table 2. d This value corresponds to the estimate for “mother is Muslim (yes vs. no)” as displayed in Table 2. Table 3. Mortality Before 5 Years of Age in Muslims Compared With Non-Muslims, by Prenatal Ramadan Exposure, Among 41,025 Children From Nouna District, Burkina Faso, 1993–2012 Prenatal Ramadan Exposure Calculationa HR 95% CI P Value Total (not considering Ramadan)b 1.15 1.07, 1.24 <0.001 Ramadan exposure categoryc  Certainly not exposed —d 0.97 0.81, 1.17 0.74  Probably not exposed 0.97 × 1.14 1.11 0.86, 1.43 0.43  Conceived during Ramadan 0.97 × 1.37 1.33 1.06, 1.66 0.01  Ramadan during trimester 1 0.97 × 1.33 1.29 1.12, 1.49 <0.001  Ramadan during trimester 2 0.97 × 1.26 1.22 1.06, 1.40 0.007  Ramadan during trimester 3 0.97 × 1.11 1.07 0.91, 1.26 0.39  Born during Ramadan 0.97 × 1.02 1.00 0.82, 1.21 0.98 Prenatal Ramadan Exposure Calculationa HR 95% CI P Value Total (not considering Ramadan)b 1.15 1.07, 1.24 <0.001 Ramadan exposure categoryc  Certainly not exposed —d 0.97 0.81, 1.17 0.74  Probably not exposed 0.97 × 1.14 1.11 0.86, 1.43 0.43  Conceived during Ramadan 0.97 × 1.37 1.33 1.06, 1.66 0.01  Ramadan during trimester 1 0.97 × 1.33 1.29 1.12, 1.49 <0.001  Ramadan during trimester 2 0.97 × 1.26 1.22 1.06, 1.40 0.007  Ramadan during trimester 3 0.97 × 1.11 1.07 0.91, 1.26 0.39  Born during Ramadan 0.97 × 1.02 1.00 0.82, 1.21 0.98 Abbreviations: CI, confidence interval; HR, hazard ratio. a Calculation of stratum HRs from the baseline HR and the interaction HR. b Survival analysis without Ramadan exposure (with a random intercept for mother; results were adjusted for year of birth, month of birth, region of residence, season at death, and sex of child). c Based on the full model shown in Table 2. d This value corresponds to the estimate for “mother is Muslim (yes vs. no)” as displayed in Table 2. Figure 2. View largeDownload slide Mortality under 5 years of age according to prenatal Ramadan exposure among children of Muslim mothers compared with children of non-Muslim mothers who were in utero during the same time period, Nouna District, Burkina Faso, 1993–2012. The hazard ratios were derived from an adjusted multilevel model (see Tables 2 and 3) that included 25,093 children born to Muslim mothers and 15,932 children born to non-Muslim mothers in Nouna District. They were calculated by dividing the under-5 mortality rates of Muslims by the under-5 mortality rates of non-Muslims. A hazard ratio of 1.3 means that under-5 mortality among Muslims was 30% higher than that among non-Muslims; a hazard ratio of 1.0 means that the mortality of Muslims and non-Muslims was the same; and hazard ratios below 1 signify lower mortality among Muslims. Vertical lines display 95% confidence intervals. Figure 2. View largeDownload slide Mortality under 5 years of age according to prenatal Ramadan exposure among children of Muslim mothers compared with children of non-Muslim mothers who were in utero during the same time period, Nouna District, Burkina Faso, 1993–2012. The hazard ratios were derived from an adjusted multilevel model (see Tables 2 and 3) that included 25,093 children born to Muslim mothers and 15,932 children born to non-Muslim mothers in Nouna District. They were calculated by dividing the under-5 mortality rates of Muslims by the under-5 mortality rates of non-Muslims. A hazard ratio of 1.3 means that under-5 mortality among Muslims was 30% higher than that among non-Muslims; a hazard ratio of 1.0 means that the mortality of Muslims and non-Muslims was the same; and hazard ratios below 1 signify lower mortality among Muslims. Vertical lines display 95% confidence intervals. In sensitivity analyses separating mortality during infancy from mortality later in childhood, Ramadan exposure during conception was associated with 68% higher infant mortality among Muslims than among non-Muslims (HR = 1.68, P = 0.004), while hazard ratios for infant mortality were 1.23 (P = 0.05) for first-trimester Ramadan occurrence and 1.17 (P = 0.13) for second-trimester Ramadan occurrence. In children aged 1–4 years, Ramadan exposure during the first (HR = 1.34, P = 0.002) and second (HR = 1.25, P = 0.02) trimesters was associated with a strong increase in mortality, but the association of Ramadan exposure during conception (HR = 1.11, P = 0.48) with mortality was weaker. When analyses were stratified by region of residence, the increased mortality rate in children exposed to Ramadan in utero was visible in all regions, with the strongest associations being seen in the Northeast and the Southwest, the regions with the highest child mortality. Because of the lower sample sizes, regional estimates were not significant at the 5% level, except for Ramadan exposure in the first or second trimester in the Southwest region, which had the largest number of children. Stratification by year of birth showed that the measures of association of Ramadan exposure with mortality during trimesters 1 and 2 were comparable between the 3 study periods (1993–1999, 2000–2006, and 2007–2012) and the association for being conceived during Ramadan was stronger during the most recent period. Among children exposed to Ramadan during conception, the association was stronger in children exposed for 14 days or more (HR = 1.45, P = 0.04). It was weaker and not significant for children exposed for less than 14 days (HR = 1.27, P = 0.21). DISCUSSION Using 20 years of surveillance data from northwestern Burkina Faso, we detected a strong association between having been exposed to Ramadan in utero and under-5 mortality. This association was present from conception to the second trimester, and particularly strong during early pregnancy, when the mortality rate of children born to Muslim mothers was around 30% higher than that among children of non-Muslim mothers born at the same time. The 15% higher mortality seen in children of Muslim mothers overall was entirely explained by Ramadan exposure, as having a Muslim mother was not associated with mortality when the pregnancy occurred outside of Ramadan. For our analysis, we assigned children of Muslim mothers as being exposed to Ramadan without having actual information on maternal fasting, as in an intention-to-treat analysis. Use of this approach avoids confounding by health issues associated with decreased fasting adherence and increased child mortality. However, if a substantial proportion of Muslim women did not adhere to Ramadan fasting during pregnancy, our results would underestimate the true association due to nondifferential misclassification. Another source of misclassification is pregnancy duration. For the Ramadan exposure assignment, we assumed a gestational age of 266 days for all children. There are, however, natural variations in gestational age at birth, and only about 50% of children are born within 7 days of the estimated due date (26). Some children born at higher gestational ages will have been classified as not exposed to Ramadan, even though they were actually exposed during conception. To ensure that our comparison category was “certainly not exposed to Ramadan,” we created a separate category of children “probably not exposed,” capturing possible exposure in very early pregnancy among children born 1–20 days postterm. A different issue is children who were born early—that is, who were in utero for less than 266 days. These children might have been assigned Ramadan exposure during conception or the beginning of trimester 1, even though they were actually not exposed to Ramadan at all. The categorization of the certainly-not-exposed children (born just before Ramadan) was not affected by this, because a shorter gestation would not lead to an overlap with Ramadan. Thus, any misclassification of pregnancy duration would only have led to an attenuation of the measures of association. If some women planned their pregnancies to avoid an overlap with Ramadan, this could have introduced bias. If this group was less vulnerable for some reason, this could have artificially increased our estimates. Such behavior should be more likely in Muslim women, because there is less reason for non-Muslims to avoid an overlap with Ramadan. Our sensitivity analyses provided no evidence for such pregnancy planning (Web Appendix 1), and previous research comparing Muslims who were pregnant during a Ramadan with those who were not found no differences in education, smoking behavior, health, or income (14–16, 19, 27), suggesting that Muslims generally do not plan their pregnancies in such a way that they avoid overlap with a Ramadan. The majority of studies on the association between Ramadan fasting and health of the offspring have focused on immediate outcomes such as birth weight and preterm delivery and have mostly found no associations (28–32). Even though low birth weight is associated with increased child mortality (33), the effect of Ramadan fasting on child mortality presumably does not act via birth weight but through an impact on the developing immune system, which is in line with the impacts of seasonality on thymus size reported from the Gambia (6). While undernutrition in utero can lead to intrauterine growth restriction that is measurable as low birth weight, as well as to epigenetic changes influencing later disease risk (9), the systems affected and the time(s) at which this shows in life will depend on the types of nutrients missing and the exact timing of undernutrition during pregnancy. For example, persons exposed to the Dutch Hunger Winter in early pregnancy had normal birth weight but suffered from a higher burden of disease as adults (9). Previous studies on long-term outcomes showed that children who were exposed to Ramadan in utero had lower academic performance (14, 15) and were more likely to suffer from learning disabilities (27), suggesting an impact on the developing nervous system. There is also evidence from Indonesia linking Ramadan exposure in utero to adult body size (17), to lower nurse-rated general adult health, and to higher prevalence of slow-healing wounds and chest pain (symptoms indicative of diabetes and coronary heart disease) (19). In line with our study, the great majority of studies on long-term outcomes of Ramadan exposure showed stronger associations if children were exposed early in pregnancy, during the time of conception and the first trimester (14, 17, 20, 34). This is biologically plausible and consistent with our knowledge on fetal programming (1, 35). An alternative explanation for the stronger associations in early pregnancy is a higher adherence to fasting during this time, especially when women are not yet aware of the pregnancy (28, 34). In pregnant women, fasting leads to increases in levels of free fatty acids and ketones and decreases in blood glucose after a relatively short period of time (12–18 hours), a process called “accelerated starvation” (36). Because of its location close to the equator, the durations of fasting periods in Burkina Faso are roughly constant over the years (around 13 hours). At the same time, preexisting undernutrition and physical labor in a hot climate may exacerbate fasting effects. Intermittent exposure to decreased glucose and increased ketone levels, the balance of calories from proteins and carbohydrates, and deficiencies in crucial micronutrients can be responsible for adverse effects on fetal health (35). From our data, however, it is impossible to say whether the association between Ramadan and mortality is due to the lack of food and water intake during the daytime or due to other lifestyle changes related to Ramadan, such as overall reduced calorie intake, increased sugar and fat intake, decreased intake of micronutrient-rich foods, changes in sleeping patterns, or other factors. The developing immune system is particularly vulnerable to maternal malnutrition in the first trimester through a range of mechanisms: micronutrient deficiencies directly affecting thymus development and hematopoiesis, impaired maternal immune transfer altering the fetal immune system’s trajectory, and maternal glucocorticoids suppressing fetal thymus and lymphocyte development as well as influencing the development of the fetal hypothalamic-pituitary-adrenal axis (37). Especially in sub-Saharan Africa, where infectious diseases (pneumonia, diarrhea, malaria) are still a major cause of under-5 mortality (38), alterations in the functioning of the immune system can be expected to influence child survival. Our finding that periconceptional Ramadan exposure was most strongly linked to mortality, and mainly to infant mortality rather than later child mortality, could mean that impacts incurred during the embryo’s formative phase are particularly severe and kill earlier or that they disrupt immune mechanisms that are important for maternal immune transfer, which is particularly relevant in infancy before vaccination takes effect. However, given our limited statistical power in subanalyses, results of the stratified analyses should not be overinterpreted. To shed more light on the mechanisms involved, it would be helpful to study cause-specific mortality and to investigate epigenetic changes where such data are available. Finally, because sex differences in immune response and in fetal programming are common (39–42), we plan to investigate effect modification by child sex as soon as newer years of surveillance data become available. In conclusion, we found that Ramadan exposure in early pregnancy leads to a 30% increase in child mortality in rural Burkina Faso, a setting where infectious causes of child mortality dominate. This suggests that changes in nutrient intake and/or lifestyle which are far less dramatic than those that occur during a famine or hunger season can have long-lasting detrimental effects on child health, which might act through impacts on immune system development (6). Besides its contribution to knowledge on the developmental origins of disease and causes of child mortality, our finding is relevant for the 1.6 billion Muslims globally, tens of millions of whom get pregnant each year. According to Islam, pregnant women are exempt from fasting during Ramadan if they are worried about their health or the health of their unborn child (24), but it has been shown in several settings that many women still fast (43–47). Conception and early pregnancy seem to be the most vulnerable periods, which complicates matters, as it would imply abstaining from fasting while trying to get pregnant and in a period when women may not yet want to reveal their pregnancy status. Furthermore, our finding may be important to women of other faiths who fast and to all women, because meal-skipping and dieting during pregnancy are common in nonreligious pregnant women as well (18, 48, 49). Even though the effect of Ramadan fasting on child mortality is probably restricted to high-mortality settings, long-term health and cognitive effects may also affect people in high-income settings. ACKNOWLEDGMENTS Author affiliations: Unit of Epidemiology and Biostatistics, Institute of Public Health, Heidelberg University, Heidelberg, Germany (Anja Schoeps, Gisela Kynast-Wolf, Sabine Gabrysch); Gutenberg School of Management and Economics, Johannes Gutenberg University, Mainz, Germany (Reyn van Ewijk); and Centre de Recherche en Santé de Nouna, Nouna, Burkina Faso (Eric Nebié, Pascal Zabré, Ali Sié). This work benefitted from support provided by the German Research Foundation to R.v.E. (grant 260639091) and to the Nouna Health and Demographic Surveillance System (research grant SFB 544 (“Control of Tropical Infectious Diseases”), 1999–2011). Conflict of interest: none declared. Abbreviations HDSS Health and Demographic Surveillance System HR hazard ratio REFERENCES 1 Ramakrishnan U , Grant F , Goldenberg T , et al. . Effect of women’s nutrition before and during early pregnancy on maternal and infant outcomes: a systematic review . Paediatr Perinat Epidemiol . 2012 ; 26 ( suppl 1 ): 285 – 301 . Google Scholar Crossref Search ADS PubMed 2 Almond D , Currie J . Killing me softly: the fetal origins hypothesis . J Econ Perspect . 2011 ; 25 ( 3 ): 153 – 172 . Google Scholar Crossref Search ADS PubMed 3 Barouki R , Gluckman PD , Grandjean P , et al. . Developmental origins of non-communicable disease: implications for research and public health . Environ Health . 2012 ; 11 : 42 . Google Scholar Crossref Search ADS PubMed 4 Barker DJ . The origins of the developmental origins theory . J Intern Med . 2007 ; 261 ( 5 ): 412 – 417 . Google Scholar Crossref Search ADS PubMed 5 Bateson P , Barker D , Clutton-Brock T , et al. . Developmental plasticity and human health . Nature . 2004 ; 430 ( 6998 ): 419 – 421 . Google Scholar Crossref Search ADS PubMed 6 Moore SE . Early life nutritional programming of health and disease in The Gambia . J Dev Orig Health Dis . 2016 ; 7 ( 2 ): 123 – 131 . Google Scholar Crossref Search ADS PubMed 7 Moore SE , Collinson AC , Tamba N’Gom P , et al. . Early immunological development and mortality from infectious disease in later life . Proc Nutr Soc . 2006 ; 65 ( 3 ): 311 – 318 . Google Scholar Crossref Search ADS PubMed 8 Lumey LH , Stein AD , Susser E . Prenatal famine and adult health . Annu Rev Public Health . 2011 ; 32 : 237 – 262 . Google Scholar Crossref Search ADS PubMed 9 Schulz LC . The Dutch Hunger Winter and the developmental origins of health and disease . Proc Natl Acad Sci USA . 2010 ; 107 ( 39 ): 16757 – 16758 . Google Scholar Crossref Search ADS PubMed 10 Roseboom T , de Rooij S , Painter R . The Dutch famine and its long-term consequences for adult health . Early Hum Dev . 2006 ; 82 ( 8 ): 485 – 491 . Google Scholar Crossref Search ADS PubMed 11 Roseboom TJ , Painter RC , van Abeelen AF , et al. . Hungry in the womb: what are the consequences? Lessons from the Dutch famine . Maturitas . 2011 ; 70 ( 2 ): 141 – 145 . Google Scholar Crossref Search ADS PubMed 12 Ekamper P , van Poppel F , Stein AD , et al. . Prenatal famine exposure and adult mortality from cancer, cardiovascular disease, and other causes through age 63 years . Am J Epidemiol . 2015 ; 181 ( 4 ): 271 – 279 . Google Scholar Crossref Search ADS PubMed 13 van Abeelen AF , Veenendaal MV , Painter RC , et al. . Survival effects of prenatal famine exposure . Am J Clin Nutr . 2012 ; 95 ( 1 ): 179 – 183 . Google Scholar Crossref Search ADS PubMed 14 Almond D , Mazumder B , van Ewijk R . In utero Ramadan exposure and children’s academic performance . Econ J . 2015 ; 125 ( 589 ): 1501 – 1533 . Google Scholar Crossref Search ADS 15 Majid MF . The persistent effects of in utero nutrition shocks over the life cycle: evidence from Ramadan fasting . J Dev Econ . 2015 ; 117 : 48 – 57 . Google Scholar Crossref Search ADS 16 Schultz-Nielsen ML , Tekin E , Greve J . Labor market effects of intrauterine exposure to nutritional deficiency: evidence from administrative data on Muslim immigrants in Denmark . Econ Hum Biol . 2016 ; 21 : 196 – 209 . Google Scholar Crossref Search ADS PubMed 17 van Ewijk RJ , Painter RC , Roseboom TJ . Associations of prenatal exposure to Ramadan with small stature and thinness in adulthood: results from a large Indonesian population-based study . Am J Epidemiol . 2013 ; 177 ( 8 ): 729 – 736 . Google Scholar Crossref Search ADS PubMed 18 Chen X . Fetus, fasting, and festival: the persistent effects of in utero social shocks . Int J Health Policy Manag . 2014 ; 3 ( 4 ): 165 – 169 . Google Scholar Crossref Search ADS PubMed 19 van Ewijk R . Long-term health effects on the next generation of Ramadan fasting during pregnancy . J Health Econ . 2011 ; 30 ( 6 ): 1246 – 1260 . Google Scholar Crossref Search ADS PubMed 20 Almond D , Mazumder B . Health Capital and the Prenatal Environment: The Effect of Maternal Fasting During Pregnancy. (NBER Working Paper 14428). Cambridge, MA : National Bureau of Economic Research ; 2008 . http://www.nber.org/papers/w14428.pdf. Accessed December 27, 2017. 21 Sié A , Louis VR , Gbangou A , et al. . The Health and Demographic Surveillance System (HDSS) in Nouna, Burkina Faso, 1993–2007 . Glob Health Action . 2010 ; 3 : 5284 . Google Scholar Crossref Search ADS 22 Schoeps A , Kynast-Wolf G , Nesbitt RC , et al. . Decreasing disparities in infant survival using surveillance data from Burkina Faso . Am J Trop Med Hyg . 2015 ; 92 ( 5 ): 1038 – 1044 . Google Scholar Crossref Search ADS PubMed 23 Beiersmann C , Bermejo Lorenzo J , Bountogo M , et al. . Malnutrition determinants in young children from Burkina Faso . J Trop Pediatr . 2013 ; 59 ( 5 ): 372 – 379 . Google Scholar Crossref Search ADS PubMed 24 Laway BA , Ashraf H . Basic rules of Ramadan: a medico-religious perspective . J Pak Med Assoc . 2015 ; 65 ( 5 suppl 1 ): S14 – S17 . Google Scholar PubMed 25 Kynast-Wolf G , Hammer GP , Müller O , et al. . Season of death and birth predict patterns of mortality in Burkina Faso . Int J Epidemiol . 2006 ; 35 ( 2 ): 427 – 435 . Google Scholar Crossref Search ADS PubMed 26 Mongelli M , Wilcox M , Gardosi J . Estimating the date of confinement: ultrasonographic biometry versus certain menstrual dates . Am J Obstet Gynecol . 1996 ; 174 ( 1 ): 278 – 281 . Google Scholar Crossref Search ADS PubMed 27 Almond D , Mazumder B . Health capital and the prenatal environment: the effect of Ramadan observance during pregnancy . Am Econ J Appl Econ . 2011 ; 3 ( 4 ): 56 – 85 . Google Scholar Crossref Search ADS 28 Arab M , Nasrollahi S . Interrelation of Ramadan fasting and birth weight . Med J Islamic World Acad Sci . 2001 ; 14 ( 3 ): 91 – 95 . 29 Cross JH , Eminson J , Wharton BA . Ramadan and birth weight at full term in Asian Moslem pregnant women in Birmingham . Arch Dis Child . 1990 ; 65 ( 10 Spec No ): 1053 – 1056 . Google Scholar Crossref Search ADS PubMed 30 Rouhani MH , Azadbakht L . Is Ramadan fasting related to health outcomes? A review on the related evidence . J Res Med Sci . 2014 ; 19 ( 10 ): 987 – 992 . Google Scholar PubMed 31 Jürges H . Ramadan fasting, sex-ratio at birth, and birth weight: no effects on Muslim infants born in Germany . Econ Lett . 2015 ; 137 : 13 – 16 . Google Scholar Crossref Search ADS 32 Daley A , Pallan M , Clifford S , et al. . Are babies conceived during Ramadan born smaller and sooner than babies conceived at other times of the year? A Born in Bradford Cohort Study . J Epidemiol Community Health . 2017 ; 71 ( 7 ): 722 – 728 . Google Scholar Crossref Search ADS PubMed 33 Watkins WJ , Kotecha SJ , Kotecha S . All-cause mortality of low birthweight infants in infancy, childhood, and adolescence: population study of England and Wales . PLoS Med . 2016 ; 13 ( 5 ): e1002018 . Google Scholar Crossref Search ADS PubMed 34 Savitri AI , Yadegari N , Bakker J , et al. . Ramadan fasting and newborn’s birth weight in pregnant Muslim women in the Netherlands . Br J Nutr . 2014 ; 112 ( 9 ): 1503 – 1509 . Google Scholar Crossref Search ADS PubMed 35 Harding JE . The nutritional basis of the fetal origins of adult disease . Int J Epidemiol . 2001 ; 30 ( 1 ): 15 – 23 . Google Scholar Crossref Search ADS PubMed 36 Metzger BE , Ravnikar V , Vileisis RA , et al. . “Accelerated starvation” and the skipped breakfast in late normal pregnancy . Lancet . 1982 ; 1 ( 8272 ): 588 – 592 . Google Scholar Crossref Search ADS PubMed 37 Palmer AC . Nutritionally mediated programming of the developing immune system . Adv Nutr . 2011 ; 2 ( 5 ): 377 – 395 . Google Scholar Crossref Search ADS PubMed 38 United Nations Children’s Fund . Levels and Trends in Child Mortality: Report 2017. New York, NY : United Nations Children’s Fund ; 2017 . https://www.unicef.org/publications/files/Child_Mortality_Report_2017.pdf. Accessed December 27, 2017. 39 Brown AS , Susser ES . Sex differences in prevalence of congenital neural defects after periconceptional famine exposure . Epidemiology . 1997 ; 8 ( 1 ): 55 – 58 . Google Scholar Crossref Search ADS PubMed 40 Tobi EW , Lumey LH , Talens RP , et al. . DNA methylation differences after exposure to prenatal famine are common and timing- and sex-specific . Hum Mol Genet . 2009 ; 18 ( 21 ): 4046 – 4053 . Google Scholar Crossref Search ADS PubMed 41 van Abeelen AF , de Rooij SR , Osmond C , et al. . The sex-specific effects of famine on the association between placental size and later hypertension . Placenta . 2011 ; 32 ( 9 ): 694 – 698 . Google Scholar Crossref Search ADS PubMed 42 Klein SL , Flanagan KL . Sex differences in immune responses . Nat Rev Immunol . 2016 ; 16 ( 10 ): 626 – 638 . Google Scholar Crossref Search ADS PubMed 43 Joosoph J , Abu J , Yu SL . A survey of fasting during pregnancy . Singapore Med J . 2004 ; 45 ( 12 ): 583 – 586 . Google Scholar PubMed 44 Mubeen SM , Mansoor S , Hussain A , et al. . Perceptions and practices of fasting in Ramadan during pregnancy in Pakistan . Iran J Nurs Midwifery Res . 2012 ; 17 ( 7 ): 467 – 471 . Google Scholar PubMed 45 van Bilsen LA , Savitri AI , Amelia D , et al. . Predictors of Ramadan fasting during pregnancy . J Epidemiol Glob Health . 2016 ; 6 ( 4 ): 267 – 275 . Google Scholar Crossref Search ADS PubMed 46 Prentice AM , Prentice A , Lamb WH , et al. . Metabolic consequences of fasting during Ramadan in pregnant and lactating women . Hum Nutr Clin Nutr . 1983 ; 37 ( 4 ): 283 – 294 . Google Scholar PubMed 47 Sultan IE , Taha IM , Hassanein MM . Ramadan fasting and maternal perspectives in healthy pregnant women: systematic review . Br J Med Med Res . 2015 ; 6 ( 6 ): 573 – 586 . Google Scholar Crossref Search ADS 48 Mazumder B , Seeskin Z . Breakfast skipping, extreme commutes, and the sex composition at birth . Biodemography Soc Biol . 2015 ; 61 ( 2 ): 187 – 208 . Google Scholar Crossref Search ADS PubMed 49 Hernández-Díaz S , Boeke CE , Romans AT , et al. . Triggers of spontaneous preterm delivery—why today? Paediatr Perinat Epidemiol . 2014 ; 28 ( 2 ): 79 – 87 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

de Rooij, Susanne R,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy088pmid: 29741567

Abstract Studies across different species have shown that moderate dietary restriction is associated with a longer life span. Surprisingly, however, when diet is restricted in prenatal life, the effect is completely the opposite. Animal studies and human epidemiologic data have shown that undernutrition in utero negatively affects health in later life and reduces life span considerably. In this issue of the Journal, Schoeps et al. (Am J Epidemiol. 2018;187(10):2085–2092) provide new evidence that variations in nutritional conditions during pregnancy relate to the future health of the unborn child. In a detailed analysis of data from Muslim and non-Muslim pregnant women in Burkina Faso, they showed that the occurrence of Ramadan in early life was strongly associated with mortality rates among children under 5 years of age. Mortality rates were highest when Ramadan had occurred in the preconception period or during the first trimester. That nutritional conditions in early life can have such profound consequences for child mortality is both astonishing and extremely relevant from a public health perspective. child mortality, famine, fasting, nutrition, prenatal exposure delayed effects, pregnancy The “developmental origins of health and disease” paradigm is currently gaining widespread momentum. The importance of a healthy early environment that sets the stage for a healthy life is increasingly being recognized. Since the beginning of this research field, with the work of Barker et al. (1) demonstrating clear associations between small size at birth and diabetes and heart disease (2), substantial attention has been given to the impact of nutrition in utero. Animal experimental work has provided a large body of evidence showing that manipulation of the prenatal and perinatal diet instigates clear effects on a wide range of health-related outcomes in the offspring, with decreased life span as a final outcome. An excellent example of such a study was performed by Ozanne and Hales (3), who demonstrated that mice that were undernourished prenatally had a 25% shorter life span than those that had not been undernourished before birth, and as such, the effects of dietary restriction in prenatal life were larger than those of postnatal diet. Studying the effects of poor nutrition in utero in humans, though, is obviously less straightforward. To overcome the limitation of being unable to experiment with nutritional conditions in humans and study its outcomes, researchers have used naturally occurring or manmade situations in which nutritional conditions varied. For example, the strong bimodal seasonality in the Gambia is associated with large differences in food availability in rural communities. Studies of the correlations between season of birth and mortality rates in young adulthood have shown that these are closely linked (4). People born during the rainy “hungry season” appeared to be up to 10 times more likely to die prematurely than those born during the dry season, which is characterized by relatively higher food abundance. Those born in the rainy season also had a 3-fold higher risk of mortality due to infectious disease (5). Over the last 100 years, famines have provided historical opportunities to study later-life health consequences of poor nutritional circumstances in early life. Different periods of famine during different periods of time across the world have been used to investigate consequences of undernutrition in early life. The most widely studied famines include the Dutch Hunger Winter, which struck the urban western part of the Netherlands at the end of World War II, and the Chinese famine, which coincided with the Great Leap Forward in 1959–1961 and was caused by the imposition of drastic changes in agriculture and economic mismanagement combined with natural disasters. Studies of the people who had been in utero during these famines have shown a range of health corollaries, especially for cardiometabolic health (6–9). Prenatal exposure to the Dutch famine has also been shown to be associated with increased mortality up to the age of 63 years (10, 11). Criticism of the famine studies has always been that the circumstances existing during these hunger periods were extremely severe and must also have caused severe stress and other physical discomfort, making it difficult to determine whether any results of prenatal famine exposure can be attributed to the state of undernutrition per se. Another point of critique that can be made is that nutritional circumstances in a war situation cannot be translated to nutritional conditions in nonwar circumstances. This is what makes the study reported by Schoeps et al. (12) in this issue of the Journal particularly interesting. The authors use the nutritional variation that accompanies the period of Ramadan, during which many Muslims participate in fasting, as a natural experiment to study the relationship between nutritional conditions in utero and child mortality. The study findings showed that, especially when Ramadan occurred during conception or the first trimester, mortality in children under age 5 years was higher by 33% and 29%, respectively, among children of Muslim mothers compared with children of non-Muslim mothers. When they looked at infant mortality only, up to the age of 1 year, exposure during the period of conception was even associated with 68% higher mortality. The results of this study largely converge with those from the Dutch famine studies in that nutritional deprivation during pregnancy, particularly the periods around conception and early pregnancy, seems to matter the most in shaping future health (6, 7, 10). Early gestation has also been demonstrated to be a critical time window for changes in the prenatal environment to affect DNA methylation levels, which are held to be responsible for the phenotypical consequences of being exposed to an adverse diet in utero (13). The finding that nutritional circumstances during the periconceptional period and early fetal life are so important in shaping future health conditions is of significant relevance from a public health point of view. The need to communicate the importance of a healthy diet, starting from the preconception period, is clearly present. The study by Schoeps et al. (12) provides important novel evidence that being in utero during the period of traditional fasting for Muslims is associated with increased child mortality. This adds to the evidence that nutrition at the very beginning of life, especially around conception and in early pregnancy, is highly important for the child’s health and may even be a matter of survival. ACKNOWLEDGMENTS Author affiliation: Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Amsterdam University Medical Center, University of Amsterdam, Amsterdam, the Netherlands (Susanne R. de Rooij). Conflict of interest: none declared. REFERENCES 1 Barker DJ , Eriksson JG , Forsén T , et al. . Fetal origins of adult disease: strength of effects and biological basis . Int J Epidemiol . 2002 ; 31 ( 6 ): 1235 – 1239 . Google Scholar Crossref Search ADS PubMed 2 Harding JE . The nutritional basis of the fetal origins of adult disease . Int J Epidemiol . 2001 ; 30 ( 1 ): 15 – 23 . Google Scholar Crossref Search ADS PubMed 3 Ozanne SE , Hales CN . Lifespan: catch-up growth and obesity in male mice . Nature . 2004 ; 427 ( 6973 ): 411 – 412 . Google Scholar Crossref Search ADS PubMed 4 Moore SE , Cole TJ , Poskitt EM , et al. . Season of birth predicts mortality in rural Gambia . Nature . 1997 ; 388 ( 6641 ): 434 . Google Scholar Crossref Search ADS PubMed 5 Moore SE , Cole TJ , Collinson AC , et al. . Prenatal or early postnatal events predict infectious deaths in young adulthood in rural Africa . Int J Epidemiol . 1999 ; 28 ( 6 ): 1088 – 1095 . Google Scholar Crossref Search ADS PubMed 6 Roseboom TJ , Painter RC , van Abeelen AF , et al. . Hungry in the womb: what are the consequences? Lessons from the Dutch famine . Maturitas . 2011 ; 70 ( 2 ): 141 – 145 . Google Scholar Crossref Search ADS PubMed 7 Lumey LH , Stein AD , Susser E . Prenatal famine and adult health . Annu Rev Public Health . 2011 ; 32 : 237 – 262 . Google Scholar Crossref Search ADS PubMed 8 Li Y , He Y , Qi L , et al. . Exposure to the Chinese famine in early life and the risk of hyperglycemia and type 2 diabetes in adulthood . Diabetes . 2010 ; 59 ( 10 ): 2400 – 2406 . Google Scholar Crossref Search ADS PubMed 9 Li Y , Jaddoe VW , Qi L , et al. . Exposure to the Chinese famine in early life and the risk of hypertension in adulthood . J Hypertens . 2011 ; 29 ( 6 ): 1085 – 1092 . Google Scholar Crossref Search ADS PubMed 10 van Abeelen AF , Veenendaal MV , Painter RC , et al. . Survival effects of prenatal famine exposure . Am J Clin Nutr . 2012 ; 95 ( 1 ): 179 – 183 . Google Scholar Crossref Search ADS PubMed 11 Ekamper P , van Poppel F , Stein AD , et al. . Prenatal famine exposure and adult mortality from cancer, cardiovascular disease, and other causes through age 63 years . Am J Epidemiol . 2015 ; 181 ( 4 ): 271 – 279 . Google Scholar Crossref Search ADS PubMed 12 Schoeps A , van Ewijk R , Kynast-Wolf G , et al. . Ramadan exposure in utero and child mortality in Burkina Faso: analysis of a population-based cohort including 41,025 children . Am J Epidemiol . 2018 ; 187 ( 10 ): 2085 – 2092 . 13 Tobi EW , Slieker RC , Stein AD , et al. . Early gestation as the critical time-window for changes in the prenatal environment to affect the adult human blood methylome . Int J Epidemiol . 2015 ; 44 ( 4 ): 1211 – 1223 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

D, Stein, Aryeh

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy089pmid: 29741561

Abstract Ramadan is observed by 1.6 billion Muslims. In an accompanying article that uses data from the Nouna Health and Demographic Surveillance System in Burkina Faso, Schoeps et al. (Am J Epidemiol. 2018;187(10):2085–2092) find that exposure to Ramadan in early pregnancy is associated with an increased risk of mortality among children under age 5 years. Ramadan exposes observant individuals to a specific pattern of nutrition and other behaviors, including changes in sleep patterns. How these behaviors might result in child mortality is not yet understood, and the findings reported in this paper should be replicated in other settings. child mortality, famine, fasting, nutrition, prenatal exposure delayed effects, pregnancy A growing body of literature has identified pregnancy, particularly periconception and early pregnancy, as a window of plasticity during which interventions can impact the developing fetus and have consequences that persist throughout postnatal life. Early work by Barker et al. (1) and others identified that birth weight, a proxy measure of cumulative exposures incurred during pregnancy, was associated with a range of later health outcomes. It is widely recognized that birth weight is but a crude measure of exposure; therefore, identifying better markers of specific exposures, or indeed the exposures themselves, is critical if epidemiologists are to be able to isolate causal factors. Given its centrality in human development, nutrition has long been identified as one such key exposure, and there is a large body of literature relating maternal prepregnancy and pregnancy nutrition to birth weight and to later child outcomes. However, the challenges of identifying women prior to conception and of measuring dietary intakes, let alone the partitioning of the ingested energy and nutrients between the mother and the developing fetus, are legion; and, of course, maternal nutritional intakes cannot be studied in isolation from the context in which the pregnant woman is living, raising questions about whether the nutritional differences observed are in fact the relevant causal factor. Sometimes, however, that context provides enough of an exposure gradient that assessment of individual intakes becomes unnecessary. Studies of famines in the Netherlands, China, and Russia have exploited dramatic reductions in food availability at a population level to infer individual-level differences in the nutritional status of pregnant women, and have examined the consequences for their offspring (2–4). In particular, the circumstances of the Dutch famine of 1944–1945, in which food rations were below 1,000 kcal/day for approximately 5 months, allow examination of differences among those exposed at specific stages of their pregnancy (2). However, there is a perverse reverse problem—while a famine is, at its core, a lack of food supply at a population level, there are usually other accompanying social stressors, including violence, forced migration, weather extremes, and other factors, and hence the specific role of nutrition cannot always be isolated. The winter of 1944–1945 was especially cold in western Europe—this made the famine more severe by limiting the ability to transfer foods by canal to the famine area, while also increasing the energy demands for survival. And the decrease in fertility associated with famine-induced amenorrhea raises questions about selectivity of the resulting birth cohort. Sometimes the variation in food availability is less severe, yet more predictable. Sub-Saharan Africa has 2 primary seasons, with food availability being very different in the dry season and the rainy season. Prior research from the Gambia has shown that season of birth is associated with later mortality (5). Ramadan is a month in which observant Muslims refrain from eating or drinking during daylight hours, resulting in a change in the usual circadian patterns of ingestion and metabolism, which is superimposed on the seasonal food context. Based as it is on the lunar calendar, Ramadan starts 11 days earlier each year. Thus, given long enough, one can differentiate the impacts of seasonal food availability and of Ramadan fasting. In an elegant paper published in the Journal this month, Schoeps et al. (6) examine whether exposure to Ramadan in selected stages of pregnancy is associated with mortality among children under age 5 years. Several aspects of the study design are of interest to epidemiologists. The study was conducted among participants in the Nouna Health and Demographic Surveillance System, a well-characterized population of approximately 100,000 individuals in northwestern Burkina Faso, established in 1992. While Burkina Faso has an equatorial climate with distinct seasons that strongly impact food availability in subsistence farming areas such as Nouna without marked variation in day length or mean temperatures, the study authors exploited the 20 years of surveillance and the variation in the timing of Ramadan to reduce the potential impact of seasonality. But the study team also had one more ace up its sleeve: The surveilled population was approximately two-thirds Muslim and one-third Christian and other religions, and the non-Muslims do not fast, while they remain subject to any seasonal or other community-wide pressures on the food system and hence nutritional intakes. Schoeps et al. used a difference-in-differences approach, common in econometric analyses and in evaluation of community-level public health programs but perhaps not as widely used in epidemiology (7), to adjust for this background experience and to isolate the potential impact of Ramadan and its associated fasting behaviors on later outcomes (6). In the whole sample, Muslims experienced a 15% increase in the under-5 mortality rate, but Muslims and non-Muslims did not experience differences in their under-5-year mortality rates when there was no exposure to Ramadan in pregnancy, and Muslim women did not experience an increase in the under-5 mortality of their children when Ramadan occurred in the later stages of pregnancy. Specifically, the authors found that exposure to Ramadan in the periconception and early pregnancy periods was associated with a 22%–30% increase in the under-5 mortality rate. Epidemiologists, when faced with a statistical association, often attempt to identify, and rule out, alternative explanations. The sample size itself was large, with 41,025 births for which date of birth was known exactly; of those 41,025 births, 4,213 children died prior to the censoring date. The authors controlled for long-term mortality trends through the inclusion of a year-of-birth variable. Indeed, there was an overall trend towards a reduction of mortality with time. Although they were not able to completely rule out a season × Ramadan interaction, as they had less than a full 33-year cycle of data available to them, any seasonal factors that impact child mortality should have been netted out through the experience of the non-Muslims, who would not have fasted. The association was observed across the geographic regions within the surveillance system. A key consideration is whether knowledge of the occurrence of Ramadan might have affected the timing of attempts to conceive, and whether the timing of pregnancy might have been differential by factors that are related to the risk of infant and child mortality. There was no suggestion of this in the data. Notably, the association was stronger for infant mortality and somewhat lower for postinfant mortality. The biological explanation for the finding is less clear. Ramadan is not, in and of itself, a psychologically stressful period. Unlike the Dutch famine, where food shortages were closely related to changes in maternal weight gain in pregnancy (8) and size at birth (9), and unlike the experience of the Gambia, where there is also an annual “hungry season” that is associated with postchildhood infectious disease mortality (5), it is not clear that Ramadan in fact results in any meaningful changes in net food intake or resting metabolic rate (10). The major behavioral change is in the timing of food intake, and therefore to some extent in the patterns of other behaviors engaged in during the month, including physical activity and sleep. Changes in drinking patterns may have resulted in mild dehydration (11). The authors hypothesize that the intermittent fasting of Ramadan affects the developing immune system (6). However, they do not provide details on specific causes of death in their population, nor do they provide any information on markers of immune function that might support this conjecture. Ramadan is observed by most of the world’s 1.6 billion Muslims. If, indeed, something about being conceived during Ramadan is raising mortality risks, this potentially affects a large proportion of the world’s population and is deserving of further study. Advising couples to refrain from attempts to conceive in the month or months before Ramadan is premature based on this single study. One hopes that further research will replicate this finding in other settings, perhaps those in which the seasonal context results in marked differences in temperature or day length over the 33-year cycle, to identify an epigenetic signal that results from this particular set of behaviors—as has been demonstrated in the context of severe, acute exposure to famine (12)—allowing exploration of any underlying mechanisms. ACKNOWLEDGMENTS Author affiliation: Hubert Department of Global Health, Rollins School of Public Heath, Emory University, Atlanta, Georgia (Aryeh D. Stein). Conflict of interest: none declared. REFERENCES 1 Barker DJ , Osmond C , Golding J , et al. . Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease . BMJ . 1989 ; 298 ( 6673 ): 564 – 567 . Google Scholar Crossref Search ADS PubMed 2 Lumey LH , Stein AD , Susser E . Prenatal famine and adult health . Annu Rev Public Health . 2011 ; 32 : 237 – 262 . Google Scholar Crossref Search ADS PubMed 3 Li C , Lumey LH . Exposure to the Chinese famine of 1959–61 in early life and long-term health conditions: a systematic review and meta-analysis . Int J Epidemiol . 2017 ; 46 ( 4 ): 1157 – 1170 . Google Scholar Crossref Search ADS PubMed 4 Stanner SA , Bulmer K , Andrès C , et al. . Does malnutrition in utero determine diabetes and coronary heart disease in adulthood? Results from the Leningrad siege study, a cross sectional study . BMJ . 1997 ; 315 ( 7119 ): 1342 – 1348 . Google Scholar Crossref Search ADS PubMed 5 Moore SE , Cole TJ , Poskitt EM , et al. . Season of birth predicts mortality in rural Gambia . Nature . 1997 ; 388 ( 6641 ): 434 . Google Scholar Crossref Search ADS PubMed 6 Schoeps A , van Ewijk R , Kynast-Wolf G , et al. . Ramadan exposure in utero and child mortality in Burkina Faso: analysis of a population-based cohort including 41,025 children . Am J Epidemiol . 2018 ; 187 ( 10 ): 2085 – 2092 . 7 Dimick JB , Ryan AM . Methods for evaluating changes in health care policy: the difference-in-differences approach . JAMA . 2014 ; 312 ( 22 ): 2401 – 2402 . Google Scholar Crossref Search ADS PubMed 8 Stein AD , Ravelli AC , Lumey LH . Famine, third-trimester pregnancy weight gain, and intrauterine growth: the Dutch Famine Birth Cohort Study . Hum Biol . 1995 ; 67 ( 1 ): 135 – 150 . Google Scholar PubMed 9 Stein AD , Zybert PA , van de Bor M , et al. . Intrauterine famine exposure and body proportions at birth: the Dutch Hunger Winter . Int J Epidemiol . 2004 ; 33 ( 4 ): 831 – 836 . Google Scholar Crossref Search ADS PubMed 10 Lessan N , Saadane I , Alkaf B , et al. . The effects of Ramadan fasting on activity and energy expenditure . Am J Clin Nutr . 2018 ; 107 ( 1 ): 54 – 61 . Google Scholar Crossref Search ADS PubMed 11 Mulyani EY , Hardinsyah , Briawan D , et al. . Hydration status of pregnant women in West Jakarta . Asia Pac J Clin Nutr . 2017 ; 26 ( suppl 1 ): S26 – S30 . Google Scholar PubMed 12 Heijmans BT , Tobi EW , Stein AD , et al. . Persistent epigenetic differences associated with prenatal exposure to famine in humans . Proc Natl Acad Sci U S A . 2008 ; 105 ( 44 ): 17046 – 17049 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Sabine, Gabrysch,;van Ewijk, Reyn,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy090pmid: 29741572

We thank Drs. de Rooij (1) and Stein (2) for their thoughtful commentaries on our article (3). While the body of evidence linking Ramadan exposure during pregnancy to adverse outcomes has grown steadily over the past few years, we totally agree that there is still comparatively little known about the mechanisms involved. What changes occur in the functioning of specific organs or body systems? Do epigenetic changes occur in relevant genes? Are there moderators of the effects? Other studies on nutritional restrictions during pregnancy, such as the Dutch famine study (4) and work from the Gambia (5), provide a fertile ground for hypotheses on mechanisms, which now need to be tested for Ramadan. Prenatal famine exposure has been shown to affect a range of adverse health outcomes (6), including diabetes, coronary heart disease, breast cancer, schizophrenia, cognitive decline, and mortality later in life, and exposure to the Dutch famine has furthermore been shown to result in persistent epigenetic changes, in the insulin-like growth factor 2 gene (IGF2) (7) and at other locations (8–10), some of which mediate the famine’s effect on body mass index and triglycerides (11). Studies in rural Gambia found that conception during the rainy season (with less food availability but more varied food and higher maternal one-carbon micronutrient levels) was associated with increased methylation at metastable epialleles, including a putative tumor suppressor and modulator of innate immunity, the vault RNA 2-1 gene (VTRNA2-1) (12–14). Prenatal Ramadan exposure has by now been linked to various outcomes, including poorer cognitive performance, a higher prevalence of symptoms indicative of coronary heart disease and type 2 diabetes, and mental disabilities, as well as altered body composition. We have now found an association with mortality in children under 5 years of age (3), in a context where infectious diseases are a major cause of child death. This led us to speculate that prenatal Ramadan exposure affects the immune system. We agree that this needs to be corroborated by future studies including cause-specific mortality and markers of immune function, such as thymic size and lymphocyte subpopulation, as studied in the Gambia (15, 16), as well as epigenetic changes in immunity-relevant genes. We must be aware that the severe and prolonged nutritional restrictions present during famines can biologically work in a very different way than the intermittent nutritional restrictions imposed during Ramadan. And as both de Rooij (1) and Stein (2) point out, during famines there are also other simultaneous exposures, such as stress, that make it hard to disentangle what exactly caused the reported effects. During Ramadan, too, there may potentially be other, concurrent exposures besides caloric restriction, such as a change in sleeping patterns, dehydration, stress, and increased sugar intake during evenings or reduced micronutrient intake. Future research needs to determine the extent to which each of these indeed plays a role for pregnant women during Ramadan and establish their biological pathways. Moreover, several exposures may interact to produce an effect, or effects may be moderated by third variables—for example, daytime fasting may be less harmful when women refrain from physically straining activities or avoid copious consumption of sugar-rich foods at night. Finally, the studies on the Dutch famine, Ramadan in pregnancy, and others highlight that we can gain valuable insights from natural experiments where true experiments are not possible and observational data suffer from confounding—which extends far beyond the field of nutrition. Epidemiologists should systematically look for natural experiments and could also benefit from applying econometric methods (difference-in-differences, regression discontinuity, instrumental variables, etc.) more frequently in these situations, as these methods are specifically suited to getting closer to causality when only observational data are available. ACKNOWLEDGMENTS Author affiliations: Unit of Epidemiology and Biostatistics, Institute of Public Health, Heidelberg University, Heidelberg, Germany (Sabine Gabrysch); and Gutenberg School of Management and Economics, Johannes Gutenberg University, Mainz, Germany (Reyn van Ewijk). Conflict of interest: none declared. REFERENCES 1 de Rooij SR . Invited commentary: a matter of survival—the detrimental consequences of adverse early-life conditions . Am J Epidemiol . 2018 ; 187 ( 10 ): 2093 – 2094 . 2 Stein AD . Invited commentary: Ramadan, pregnancy, nutrition, and epidemiology . Am J Epidemiol . 2018 ; 187 ( 10 ): 2095 – 2097 . 3 Schoeps A , van Ewijk R , Kynast-Wolf G , et al. . Ramadan exposure in utero and child mortality in Burkina Faso: analysis of a population-based cohort including 41,025 children . Am J Epidemiol . 2018 ; 187 ( 10 ): 2085 – 2092 . 4 Roseboom TJ , Painter RC , van Abeelen AF , et al. . Hungry in the womb: what are the consequences? Lessons from the Dutch famine . Maturitas . 2011 ; 70 ( 2 ): 141 – 145 . Google Scholar Crossref Search ADS PubMed 5 Moore SE . Early life nutritional programming of health and disease in The Gambia . J Dev Orig Health Dis . 2016 ; 7 ( 2 ): 123 – 131 . Google Scholar Crossref Search ADS PubMed 6 Lumey LH , Stein AD , Susser E . Prenatal famine and adult health . Annu Rev Public Health . 2011 ; 32 : 237 – 262 . Google Scholar Crossref Search ADS PubMed 7 Heijmans BT , Tobi EW , Stein AD , et al. . Persistent epigenetic differences associated with prenatal exposure to famine in humans . Proc Natl Acad Sci U S A . 2008 ; 105 ( 44 ): 17046 – 17049 . Google Scholar Crossref Search ADS PubMed 8 Tobi EW , Goeman JJ , Monajemi R , et al. . DNA methylation signatures link prenatal famine exposure to growth and metabolism . Nat Commun . 2014 ; 5 : 5592 . Google Scholar Crossref Search ADS PubMed 9 Tobi EW , Lumey LH , Talens RP , et al. . DNA methylation differences after exposure to prenatal famine are common and timing- and sex-specific . Hum Mol Genet . 2009 ; 18 ( 21 ): 4046 – 4053 . Google Scholar Crossref Search ADS PubMed 10 Tobi EW , Slieker RC , Stein AD , et al. . Early gestation as the critical time-window for changes in the prenatal environment to affect the adult human blood methylome . Int J Epidemiol . 2015 ; 44 ( 4 ): 1211 – 1223 . Google Scholar Crossref Search ADS PubMed 11 Tobi EW , Slieker RC , Luijk R , et al. . DNA methylation as a mediator of the association between prenatal adversity and risk factors for metabolic disease in adulthood . Sci Adv . 2018 ; 4 ( 1 ): eaao4364 . Google Scholar Crossref Search ADS PubMed 12 Dominguez-Salas P , Moore SE , Baker MS , et al. . Maternal nutrition at conception modulates DNA methylation of human metastable epialleles . Nat Commun . 2014 ; 5 : 3746 . Google Scholar Crossref Search ADS PubMed 13 Silver MJ , Kessler NJ , Hennig BJ , et al. . Independent genomewide screens identify the tumor suppressor VTRNA2-1 as a human epiallele responsive to periconceptional environment . Genome Biol . 2015 ; 16 : 118 . Google Scholar Crossref Search ADS PubMed 14 Waterland RA , Kellermayer R , Laritsky E , et al. . Season of conception in rural Gambia affects DNA methylation at putative human metastable epialleles . PLoS Genet . 2010 ; 6 ( 12 ): e1001252 . Google Scholar Crossref Search ADS PubMed 15 Collinson AC , Moore SE , Cole TJ , et al. . Birth season and environmental influences on patterns of thymic growth in rural Gambian infants . Acta Paediatr . 2003 ; 92 ( 9 ): 1014 – 1020 . Google Scholar Crossref Search ADS PubMed 16 Collinson AC , Ngom PT , Moore SE , et al. . Birth season and environmental influences on blood leucocyte and lymphocyte subpopulations in rural Gambian infants . BMC Immunol . 2008 ; 9 : 18 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Fabienne, Pradella,;van Ewijk, Reyn,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy132pmid: 29961865

Abstract While prenatal exposure to Ramadan has been shown to be negatively associated with general physical and mental health, studies on specific organs remain scarce. In this study, we explored whether Ramadan exposure during pregnancy affects the occurrence of wheezing, a main symptom of obstructive airway disease. Using data from the Indonesian Family Life Survey collected between 1997 and 2008 (waves 2–4), we compared wheezing occurrence among adult Muslims who had been in utero during Ramadan with that in adult Muslims who had not been in utero during Ramadan. Wheezing prevalence was higher among adult Muslims who had been in utero during Ramadan, independent of the pregnancy phase in which the exposure to Ramadan occurred. Moreover, this association tended to increase with age, being strongest among those aged about 45 years or older. This is in line with fetal programming theory, suggesting that impacts of in utero exposures often manifest only after reproductive age. Particularly strong associations were detected for smokers. The respiratory system of prenatally exposed Muslims thus seems to perform worse in mitigating later ex utero harmful influences such as smoking. This study suggests that exposure to Ramadan during pregnancy may have lasting consequences for adult lung functionality. fasting, fetal programming, nutrition, obstructive airway disease, pregnancy, prenatal exposure delayed effects, Ramadan, wheezing Respiratory diseases caused by airflow limitations rank among the top causes of death (1). Next to cigarette smoke, shocks experienced in utero and early life are main risk factors for lung dysfunction in adulthood (2, 3). As the respiratory system develops from the embryonic period onwards, shocks incurred during all pregnancy trimesters potentially impair its development. One type of prenatal shock is exposure to nutritional shortages. Lopuhaä et al. (4) found higher risks of chronic obstructive airway disease after in utero exposure to the 1944–1945 Dutch famine. However, to our knowledge, the impact of less extreme nutritional shortages during gestation on the functionability of the respiratory system in adulthood has not yet been explored. Using data from Indonesia, we investigated whether in utero exposure to Ramadan was associated with the occurrence of wheezing among adult Muslims. The intermittent fasting performed during Ramadan is less extreme than that experienced in a famine. While negative associations between in utero exposure to Ramadan and general health have been detected, studies on specific organs remain scarce. To our knowledge, this is the first study to have explored whether in utero exposure to Ramadan has an effect on lung function. Our outcome variable was wheezing—the occurrence of a whistling sound during exhalation, indicative of illnesses such as asthma, chronic obstructive airway disease, and emphysema. We particularly examined the occurrence of wheezing among smokers, as we hypothesized that prenatal exposure to Ramadan degrades the respiratory system’s capability to deal with the ex utero strain of smoking. More than 22% of the world population adheres to Islam. Moreover, improved knowledge on the early-life origins of obstructive airway disease will contribute to the development of guidelines on topics such as dieting during pregnancy. Fetal programming and the respiratory system The in utero environment codetermines an individual’s health (5). The underlying mechanism of fetal programming is that the fetus adapts to the in utero environment, with long-lasting consequences: If nutrient or oxygen transfer is limited, adaptations help the fetus survive in the short run but can have consequences for disease susceptibility in later life. Limited transfer of nutrients during critical growth phases puts fetal organs at risk of remaining underdeveloped. The development of the lung is characterized by multiple critical growth phases which start in the embryonic period. Therefore, the respiratory system is vulnerable to shocks during all phases of gestation (6–8). Adult respiratory health is impaired by in utero exposure to tobacco smoking, nutritional deficits, or placental insufficiency. Effects depend on the timing and severity of the exposure (8, 9). Epidemiologic studies have found associations between low birth weight and impaired lung function in adulthood (10–15). Epigenetic studies suggest that in utero shocks can alter DNA methylation patterns, with consequences such as increased allergy risks (16). Studies on maternal malnutrition during pregnancy and its associations with lung development in adulthood remain scarce. Studying people born around the time of the Dutch famine, Lopuhaä et al. (4) found that symptoms of chronic obstructive airway disease were experienced more frequently by persons whose time in utero overlapped with the extreme food shortage. The associations were strongest for those exposed in midgestation, while a tendency toward higher risks was also found for those exposed in early gestation (4). The associations of less extreme, or intermittent, nutritional restrictions with airway development have not yet been studied. In this study, we investigated the associations of the intermittent fasting performed during Ramadan with lung function in adulthood. Because many in utero shocks have consequences only at postreproductive ages, we expected to observe the strongest associations in later adulthood (17, 18). Moreover, we explored potential interaction effects from the combination of in utero exposure to Ramadan and subsequent ex utero exposure to harmful substances such as cigarette smoke. We hypothesized that prenatally exposed respiratory systems are weakened and more susceptible to complications stemming from other risk factors in adulthood. Ramadan during pregnancy Ramadan constitutes one of the 5 pillars of Islam. For 1 month, Muslims fast during daylight hours. Consequently, the intake of food and drinks is shifted to occur before sunrise and after sunset. Fasting hours depend on the location and time of the year during which Ramadan takes place: Because the dates of Ramadan are determined by the lunar Islamic calender, its timing shifts over the years (the lunar calender is about 11 days shorter than the Gregorian calendar). According to most interpretations of the Koran, pregnant women may refrain from fasting if they believe that fasting will harm their health or the health of the unborn child. However, they are required to make up for the missed days or compensate by way of an expiatory payment. Most Muslim women observe the fast during pregnancy, with varying fasting rates per country (19–25). For Indonesia, Majid (26) calculated an implied fasting rate of 68%–82%. Van Bilsen et al. (27) conducted a survey among 186 pregnant Muslim women in Jakarta, Indonesia. Among those women, 80% decided to fast on at least 1 day during their pregnancy, and 30% had fasted for more than 20 days. Research on prenatal Ramadan exposure and health outcomes has shown negative associations with general health and cognitive performance (26, 28–32). An important implication of these findings is that malnutrition during gestation also has effects when it occurs in intermittent forms. Worse general long-term health can largely be traced back to impaired fetal growth. During pregnancy, energy is required for the growth of the fetus and placenta, in addition to a woman’s normal energy demands. Because of this increased energy demand, during the second half of pregnancy, a woman’s blood levels of metabolic fuels and hormones quickly approach levels comparable to those of women exposed to famines (“accelerated starvation”). In the second and third trimesters of pregnancy, signs of accelerated starvation are already detected when single meals are skipped, particularly during activity-intensive daytime hours (33, 34). Adhering to the fast during Ramadan has been shown to lead to symptoms indicative of accelerated starvation (24, 25, 35). It can thus be assumed that a fetus in later gestation has to make compromises in its growth to get along with the scarce energy supply. Additionally, during early gestation, fetal growth is vulnerable in response to maternal nutrition, particularly with respect to organ development (36–38). Studies on the Dutch famine have found the strongest associations with adult health for exposure during early gestation (39). With regard to Ramadan, changes in the nutritional composition of the mother’s diet, as measured in micronutrients (40), lower total caloric intake among fasting Muslims (35, 41), or increased levels of stress hormones (42), could explain the associations. METHODS Data We used individual-level data from a longitudinal study conducted by the RAND Corporation (Santa Monica, California), the Indonesian Family Life Survey (IFLS). Indonesia is the country with the largest Muslim population worldwide. We pooled data from IFLS waves 1–4 (1993–2008). When information was available from several waves (except for the outcome measures—see next section), we used the latest available information. Information on breathing difficulties (collected in waves 2–4) was self-reported, and respondents were asked to indicate whether they had experienced wheezing or shortness of breath during the 4 weeks before the interview. From wave 1, only information on date of birth and smoking status was included. The final sample size was 28,489. Several adjustments to the sample were undertaken: First, only respondents aged 15 years or older were asked whether they had experienced breathing difficulties. Second, people who did not know their exact date of birth were excluded, because in utero Ramadan exposure is calculated on the basis of date of birth. Moreover, we detected “heaping” of reported dates of birth on several days, such as January 1 or the Indonesian day of independence on August 17. People who indicated these dates of birth were excluded. Third, we excluded all observations that did not indicate a predominantly Muslim province as the place of residence. Thus, we could ascertain that only people living in regions where Ramadan was widely practiced were included in our sample. This prevented noise in our estimation due to different traditions in non-Muslim areas of Indonesia. Outcome measures We used 2 outcome measures. In our main analysis, the outcome of interest was wheezing. An individual was considered to have had an occurrence of wheezing if she indicated having wheezed in at least 1 IFLS wave. Wheezing was therefore a dummy variable (occurred/did not occur in the last 4 weeks before the interview). However, because all data on respiratory function were self-reported, it might have happened that some persons misreported or combined symptoms. Against this background, in a second specification we used the occurrence of any breathing difficulty (wheezing/shortness of breath in the 4 four weeks before the interview; ever being diagnosed with asthma or another lung condition) as the outcome measure. Because the question on diagnosed asthma or other lung conditions was only introduced in wave 4, we only used data from wave 4 for this specification. Categories of exposure to Ramadan Anyone whose time in utero overlapped with Ramadan was classified as having been prenatally exposed. We calculated this overlap on the basis of the person’s date of birth and the historical starting and ending dates of Ramadan. Our calculations were based on the average length of human pregnancies (266 days from the day of conception). Thus, if there was an overlap between the most recent Ramadan before a person’s date of birth and the 266 days before the date of birth, he or she was considered exposed. The control group consisted of Muslims whose own time in utero did not overlap with Ramadan. These people were conceived after the end of Ramadan and more than 266 days before the start of the next Ramadan, so that no overlap between Ramadan and their time in utero occured. If one compared children born to fasting Muslims with those born to nonfasting Muslims, there would be many sources of confounding; but if one compares Muslims who (because of their birth dates) were in utero during Ramadan with Muslims who (because of their birth dates) were not in utero during Ramadan, there is much less scope for confounding. Any confounder that was to bias our results would then need to be correlated with both wheezing and moment of birth in the Islamic year. It is important to note that we studied the health impacts of exposure to Ramadan—in any form—during pregnancy, not specifically of Ramadan fasting. Ramadan includes other aspects besides fasting, such as altered sleeping patterns and changes in nutrition (high–glycemic-content foods). Moreover, information on the actual fasting behavior of the mothers of our adult respondents during pregnancy was not available. To the extent that all (or most) of the association was due to fasting (and not to other aspects of Ramadan), our estimation was an intention-to-treat estimation, in which all persons who were Muslim and whose in utero phase overlapped with Ramadan were regarded as exposed. Besides the binary classification as exposed/not exposed, we divided exposed persons into 5 subgroups in order to investigate whether associations differed with respect to the timing of Ramadan during gestation. First, we created subgroups for those who were born or conceived during Ramadan and whose own time in utero thus overlapped with only a part of Ramadan. Second, we subdivided those who were exposed to an entire Ramadan according to the pregnancy trimester during which the overlap between Ramadan and their own time in utero started (first trimester: days 1–89 of gestation; second trimester: days 90–178 of gestation; third trimester: days 179–266 of gestation) (Table 1). Each observation was classified into 1 exposure category. In order to prevent noise in the control group, we classified all persons whose conception was calculated to have occurred less than 21 days after the end of Ramadan into a separate group (“probably not exposed”); hence, they were effectively taken out of the control group. The reason for this is that if those people were born postterm and had thus been exposed to Ramadan during their first weeks in utero, they would have been erroneously classified as not exposed (see also van Ewijk (28)). Table 1. Characteristics of 26,313 Muslims and 2,176 Non-Muslims Living in Predominantly Muslim Areas (Ages ≥15 Years), Indonesian Family Life Survey, 1993–2008 Characteristic Muslims Non-Muslims % No. With Characteristica No. With Data Availableb % No. With Characteristic No. With Data Available Mean age, yearsc 34.9 (14.7) 26,313 36.5 (16.2) 2,176 Age group, years  15–29 45.7 12,021 26,313 43.6 949 2,176  30–39 23.4 6,155 26,313 20.7 451 2,176  40–49 14.5 3,808 26,313 14.1 306 2,176  50–59 9.1 2,387 26,313 11.2 244 2,176  ≥60 7.4 1,942 26,313 10.4 226 2,176 Male sex 49.5 13,019 26,313 49.4 1,074 2,176 Presence of lung condition  Any breathing difficultyd 10.9 2,489 22,893 11.7 200 1,706  Wheezing in past 4 weeks 3.4 893 26,313 3.2 69 2,176 Smoking status  Smokere 35.9 9,423 26,283 33.0 716 2,172   Male subsample 69.4 9,023 13,005 61.3 656 1,070   Female subsample 3.0 400 13,278 5.4 60 1,102 Ramadan exposure category  Certainly not in utero during Ramadan 11.3 2,974 26,313 11.2 243 2,176  Probably not in utero during Ramadan 5.7 1,490 26,313 6.7 145 2,176  In utero during Ramadan 83.0 21,849 26,313 82.2 1,788 2,176   Conceived during Ramadan 9.1 2,401 26,313 8.6 186 2,176   Ramadan started in trimester 1 25.3 6,662 26,313 24.4 532 2,176   Ramadan started in trimester 2 24.1 6,342 26,313 24.7 537 2,176   Ramadan started in trimester 3 16.3 4,284 26,313 17.0 369 2,176   Born during Ramadan 8.2 2,160 26,313 7.5 164 2,176 Characteristic Muslims Non-Muslims % No. With Characteristica No. With Data Availableb % No. With Characteristic No. With Data Available Mean age, yearsc 34.9 (14.7) 26,313 36.5 (16.2) 2,176 Age group, years  15–29 45.7 12,021 26,313 43.6 949 2,176  30–39 23.4 6,155 26,313 20.7 451 2,176  40–49 14.5 3,808 26,313 14.1 306 2,176  50–59 9.1 2,387 26,313 11.2 244 2,176  ≥60 7.4 1,942 26,313 10.4 226 2,176 Male sex 49.5 13,019 26,313 49.4 1,074 2,176 Presence of lung condition  Any breathing difficultyd 10.9 2,489 22,893 11.7 200 1,706  Wheezing in past 4 weeks 3.4 893 26,313 3.2 69 2,176 Smoking status  Smokere 35.9 9,423 26,283 33.0 716 2,172   Male subsample 69.4 9,023 13,005 61.3 656 1,070   Female subsample 3.0 400 13,278 5.4 60 1,102 Ramadan exposure category  Certainly not in utero during Ramadan 11.3 2,974 26,313 11.2 243 2,176  Probably not in utero during Ramadan 5.7 1,490 26,313 6.7 145 2,176  In utero during Ramadan 83.0 21,849 26,313 82.2 1,788 2,176   Conceived during Ramadan 9.1 2,401 26,313 8.6 186 2,176   Ramadan started in trimester 1 25.3 6,662 26,313 24.4 532 2,176   Ramadan started in trimester 2 24.1 6,342 26,313 24.7 537 2,176   Ramadan started in trimester 3 16.3 4,284 26,313 17.0 369 2,176   Born during Ramadan 8.2 2,160 26,313 7.5 164 2,176 a Number of participants with the characteristic (for binary variables). b Number of participants for whom data were available. c Values are expressed as mean (standard deviation). d Information on any breathing difficulty was available only for persons observed in the fourth wave of the survey. e Information on smoking status was unavailable for several persons. Table 1. Characteristics of 26,313 Muslims and 2,176 Non-Muslims Living in Predominantly Muslim Areas (Ages ≥15 Years), Indonesian Family Life Survey, 1993–2008 Characteristic Muslims Non-Muslims % No. With Characteristica No. With Data Availableb % No. With Characteristic No. With Data Available Mean age, yearsc 34.9 (14.7) 26,313 36.5 (16.2) 2,176 Age group, years  15–29 45.7 12,021 26,313 43.6 949 2,176  30–39 23.4 6,155 26,313 20.7 451 2,176  40–49 14.5 3,808 26,313 14.1 306 2,176  50–59 9.1 2,387 26,313 11.2 244 2,176  ≥60 7.4 1,942 26,313 10.4 226 2,176 Male sex 49.5 13,019 26,313 49.4 1,074 2,176 Presence of lung condition  Any breathing difficultyd 10.9 2,489 22,893 11.7 200 1,706  Wheezing in past 4 weeks 3.4 893 26,313 3.2 69 2,176 Smoking status  Smokere 35.9 9,423 26,283 33.0 716 2,172   Male subsample 69.4 9,023 13,005 61.3 656 1,070   Female subsample 3.0 400 13,278 5.4 60 1,102 Ramadan exposure category  Certainly not in utero during Ramadan 11.3 2,974 26,313 11.2 243 2,176  Probably not in utero during Ramadan 5.7 1,490 26,313 6.7 145 2,176  In utero during Ramadan 83.0 21,849 26,313 82.2 1,788 2,176   Conceived during Ramadan 9.1 2,401 26,313 8.6 186 2,176   Ramadan started in trimester 1 25.3 6,662 26,313 24.4 532 2,176   Ramadan started in trimester 2 24.1 6,342 26,313 24.7 537 2,176   Ramadan started in trimester 3 16.3 4,284 26,313 17.0 369 2,176   Born during Ramadan 8.2 2,160 26,313 7.5 164 2,176 Characteristic Muslims Non-Muslims % No. With Characteristica No. With Data Availableb % No. With Characteristic No. With Data Available Mean age, yearsc 34.9 (14.7) 26,313 36.5 (16.2) 2,176 Age group, years  15–29 45.7 12,021 26,313 43.6 949 2,176  30–39 23.4 6,155 26,313 20.7 451 2,176  40–49 14.5 3,808 26,313 14.1 306 2,176  50–59 9.1 2,387 26,313 11.2 244 2,176  ≥60 7.4 1,942 26,313 10.4 226 2,176 Male sex 49.5 13,019 26,313 49.4 1,074 2,176 Presence of lung condition  Any breathing difficultyd 10.9 2,489 22,893 11.7 200 1,706  Wheezing in past 4 weeks 3.4 893 26,313 3.2 69 2,176 Smoking status  Smokere 35.9 9,423 26,283 33.0 716 2,172   Male subsample 69.4 9,023 13,005 61.3 656 1,070   Female subsample 3.0 400 13,278 5.4 60 1,102 Ramadan exposure category  Certainly not in utero during Ramadan 11.3 2,974 26,313 11.2 243 2,176  Probably not in utero during Ramadan 5.7 1,490 26,313 6.7 145 2,176  In utero during Ramadan 83.0 21,849 26,313 82.2 1,788 2,176   Conceived during Ramadan 9.1 2,401 26,313 8.6 186 2,176   Ramadan started in trimester 1 25.3 6,662 26,313 24.4 532 2,176   Ramadan started in trimester 2 24.1 6,342 26,313 24.7 537 2,176   Ramadan started in trimester 3 16.3 4,284 26,313 17.0 369 2,176   Born during Ramadan 8.2 2,160 26,313 7.5 164 2,176 a Number of participants with the characteristic (for binary variables). b Number of participants for whom data were available. c Values are expressed as mean (standard deviation). d Information on any breathing difficulty was available only for persons observed in the fourth wave of the survey. e Information on smoking status was unavailable for several persons. A further reason for the separate “probably not exposed” group is the Eid al-Fitr celebration, which takes place immediately after Ramadan. The celebration, commonly referred to as “Lebaran” in Indonesia, is a festival of breaking the fast for which many Muslims travel to their home villages. The celebration lasts about 1 week. At least 7 days prior to Lebaran, all employed Muslims receive a mandatory holiday bonus of at least 1 full month’s salary. Consequently, despite higher food prices during Ramadan, people can afford more and better-quality food towards the end of Ramadan. It is possible that the health impacts of Ramadan and Lebaran are confounded during the last days of Ramadan. The diurnal fast continues, and most Muslims rather use the holiday bonus to pay for traveling to their home village, gifts for the family, or food for the celebrations. Because of the traveling, selective fertility needs to be taken into consideration. We assumed that persons who conceived around Lebaran might differ systematically from persons who did not conceive during a holiday. Even though the direction of any bias is unclear because of the complex interdependency of an extra salary, rising food prices, and celebrations, we avoided noise by excluding persons conceived during Lebaran from our control group. Statistical methods We compared data on the breathing difficulties of Muslims who were prenatally exposed to Ramadan with those of Muslims who were certainly not in utero during Ramadan. Standard errors were clustered at the household level, as within-family correlations on health outcomes were likely. The average size of Muslim households in our sample was 3.06 persons (standard deviation, 1.66). We performed logistic regression analyses and controlled for age at the time of the interview and age squared, month of birth, sex, and IFLS wave. The results of our analyses are displayed as odds ratios. Because the timing of Ramadan shifts over the years, the effects of Ramadan can be separated from seasonal effects by including month-of-birth dummy variables as covariates (28, 31). Because it lies on the equator, the times of sunrise and sunset do not vary considerably over the course of the year in Indonesia. This implies that the duration of fasting does not vary over the years and that our results were not biased due to a correlation between year of Ramadan exposure and number of hours of Ramadan exposure. We further differentiated between the various times of prenatal Ramadan exposure (conception, first trimester, second trimester, third trimester, and birth). In addition to performing the analysis for all Muslims, we conducted separate regressions by sex and by age group. Moreover, we allowed the association to vary by smoking status by including a term for interaction between exposure and smoking status. Since exposure was measured as the occurrence of an overlap between Ramadan and pregnancy, our study can be regarded as a natural experiment. Overlap between Ramadan and pregnancy occurs quasirandomly, since selective timing of pregnancy to avoid or promote Ramadan during pregnancy is rare or absent in Indonesia (28). The quasirandom prenatal exposure meant that potential confounders such as maternal characteristics, body mass index, or exposure to air pollution were canceled out; that is, we were unable to control for all factors that might influence the occurrence of breathing difficulties, but this did not bias our results, since these factors did not affect whether Ramadan occurred during pregnancy. In order to test the robustness of our results, we conducted a difference-in-differences (DID) analysis. DID entails the comparison of treatment effects between 2 groups. We included the non-Muslims and an indicator variable for exposure status, an indicator variable for religion, and an interaction of all covariates with religion. The interaction term religion × exposed compares the strength of the association of Ramadan occurrence during pregnancy with wheezing between Muslims and non-Muslims. Naturally, non-Muslims do not observe Ramadan. Hence, any “effect” of overlap between Ramadan and pregnancy among non-Muslims has to be due to residual confounding. It is this residual confounding that we therefore took out in the DID analysis. As in our main analyses, we conducted separate regressions by age group. RESULTS Baseline characteristics Our sample consisted of 28,489 observations (26,313 Muslims and 2,176 non-Muslims). Indonesia has a young population, which was reflected in our sample: The largest age group was people between 15 and 30 years of age (Table 1). In utero exposure to Ramadan and breathing difficulties Prenatal exposure to Ramadan was associated with higher risks for lung conditions in adulthood, in terms of both general breathing difficulties and wheezing (Table 2). In comparison with nonexposed Muslims, the risk of experiencing a symptom of any breathing difficulty (wheezing, shortness of breath) or being diagnosed with a lung disease (asthma, other lung condition) was 17.3% higher. For exposed men, the risk of experiencing any breathing difficulty was 20.5% higher. Table 2. Associations Between In Utero Exposure to Ramadan and Development of a Lung Condition in Adulthood (Ages ≥15 Years), Overall and by Sex, Among Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008 Ramadan Exposure Category All Muslims Female Muslims Male Muslims Any Breathing Difficultya (n = 22,893) Wheezingb (n = 26,313) Any Breathing Difficulty (n = 11,768) Wheezing (n = 13,294) Any Breathing Difficulty (n = 11,125) Wheezing (n = 13,019) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadanc 1.17 1.02, 1.35 0.022 1.26 0.97, 1.63 0.087 1.14 0.94, 1.39 0.176 1.09 0.75, 1.57 0.657 1.21 0.99, 1.47 0.061 1.45 1.00, 2.12 0.051 Exposure period  Conceived during Ramadan 1.17 0.97, 1.41 0.100 1.11 0.77, 1.58 0.582 1.14 0.88, 1.48 0.334 0.78 0.46, 1.34 0.371 1.20 0.91, 1.57 0.197 1.47 0.90, 2.42 0.124  Ramadan started in trimester 1 1.21 1.04, 1.41 0.015 1.28 1.00, 1.71 0.099 1.17 0.94, 1.46 0.151 1.18 0.78, 1.78 0.431 1.26 1.01, 1.57 0.038 1.40 0.93, 2.12 0.111  Ramadan started in trimester 2 1.15 0.99, 1.34 0.068 1.30 0.97, 1.74 0.076 1.12 0.90, 1.39 0.324 1.12 0.75, 1.69 0.578 1.19 0.96, 1.48 0.119 1.49 0.99, 2.25 0.058  Ramadan started in trimester 3 1.20 1.02, 1.41 0.032 1.25 0.92, 1.69 0.156 1.17 0.93, 1.48 0.174 1.07 0.69, 1.66 0.761 1.22 0.96, 1.54 0.105 1.46 0.94, 2.27 0.091  Born during Ramadan 1.08 0.89, 1.32 0.439 1.27 0.89, 1.81 0.197 1.09 0.83, 1.43 0.557 1.10 0.67, 1.82 0.705 1.08 0.81, 1.43 0.615 1.49 0.89, 2.49 0.127  Probably not in utero during Ramadan 1.20 0.96, 1.49 0.103 1.35 0.91, 2.01 0.133 1.22 0.90, 1.66 0.205 1.23 0.74, 2.23 0.364 1.19 0.87, 1.62 0.280 1.45 0.82, 2.57 0.200 Ramadan Exposure Category All Muslims Female Muslims Male Muslims Any Breathing Difficultya (n = 22,893) Wheezingb (n = 26,313) Any Breathing Difficulty (n = 11,768) Wheezing (n = 13,294) Any Breathing Difficulty (n = 11,125) Wheezing (n = 13,019) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadanc 1.17 1.02, 1.35 0.022 1.26 0.97, 1.63 0.087 1.14 0.94, 1.39 0.176 1.09 0.75, 1.57 0.657 1.21 0.99, 1.47 0.061 1.45 1.00, 2.12 0.051 Exposure period  Conceived during Ramadan 1.17 0.97, 1.41 0.100 1.11 0.77, 1.58 0.582 1.14 0.88, 1.48 0.334 0.78 0.46, 1.34 0.371 1.20 0.91, 1.57 0.197 1.47 0.90, 2.42 0.124  Ramadan started in trimester 1 1.21 1.04, 1.41 0.015 1.28 1.00, 1.71 0.099 1.17 0.94, 1.46 0.151 1.18 0.78, 1.78 0.431 1.26 1.01, 1.57 0.038 1.40 0.93, 2.12 0.111  Ramadan started in trimester 2 1.15 0.99, 1.34 0.068 1.30 0.97, 1.74 0.076 1.12 0.90, 1.39 0.324 1.12 0.75, 1.69 0.578 1.19 0.96, 1.48 0.119 1.49 0.99, 2.25 0.058  Ramadan started in trimester 3 1.20 1.02, 1.41 0.032 1.25 0.92, 1.69 0.156 1.17 0.93, 1.48 0.174 1.07 0.69, 1.66 0.761 1.22 0.96, 1.54 0.105 1.46 0.94, 2.27 0.091  Born during Ramadan 1.08 0.89, 1.32 0.439 1.27 0.89, 1.81 0.197 1.09 0.83, 1.43 0.557 1.10 0.67, 1.82 0.705 1.08 0.81, 1.43 0.615 1.49 0.89, 2.49 0.127  Probably not in utero during Ramadan 1.20 0.96, 1.49 0.103 1.35 0.91, 2.01 0.133 1.22 0.90, 1.66 0.205 1.23 0.74, 2.23 0.364 1.19 0.87, 1.62 0.280 1.45 0.82, 2.57 0.200 Abbreviations: CI, confidence interval; OR, odds ratio. a Ever having suffered from any breathing difficulty. The analyses of general breathing difficulties were based on data from the fourth wave of the survey only. b Having experienced wheezing in the past 4 weeks. c Results stem from 2 separate logistic regressions per column (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that adjusted for age, age2, sex, and month of birth. In the wheezing analyses, results were additionally adjusted for survey wave. Standard errors were clustered by household. Table 2. Associations Between In Utero Exposure to Ramadan and Development of a Lung Condition in Adulthood (Ages ≥15 Years), Overall and by Sex, Among Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008 Ramadan Exposure Category All Muslims Female Muslims Male Muslims Any Breathing Difficultya (n = 22,893) Wheezingb (n = 26,313) Any Breathing Difficulty (n = 11,768) Wheezing (n = 13,294) Any Breathing Difficulty (n = 11,125) Wheezing (n = 13,019) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadanc 1.17 1.02, 1.35 0.022 1.26 0.97, 1.63 0.087 1.14 0.94, 1.39 0.176 1.09 0.75, 1.57 0.657 1.21 0.99, 1.47 0.061 1.45 1.00, 2.12 0.051 Exposure period  Conceived during Ramadan 1.17 0.97, 1.41 0.100 1.11 0.77, 1.58 0.582 1.14 0.88, 1.48 0.334 0.78 0.46, 1.34 0.371 1.20 0.91, 1.57 0.197 1.47 0.90, 2.42 0.124  Ramadan started in trimester 1 1.21 1.04, 1.41 0.015 1.28 1.00, 1.71 0.099 1.17 0.94, 1.46 0.151 1.18 0.78, 1.78 0.431 1.26 1.01, 1.57 0.038 1.40 0.93, 2.12 0.111  Ramadan started in trimester 2 1.15 0.99, 1.34 0.068 1.30 0.97, 1.74 0.076 1.12 0.90, 1.39 0.324 1.12 0.75, 1.69 0.578 1.19 0.96, 1.48 0.119 1.49 0.99, 2.25 0.058  Ramadan started in trimester 3 1.20 1.02, 1.41 0.032 1.25 0.92, 1.69 0.156 1.17 0.93, 1.48 0.174 1.07 0.69, 1.66 0.761 1.22 0.96, 1.54 0.105 1.46 0.94, 2.27 0.091  Born during Ramadan 1.08 0.89, 1.32 0.439 1.27 0.89, 1.81 0.197 1.09 0.83, 1.43 0.557 1.10 0.67, 1.82 0.705 1.08 0.81, 1.43 0.615 1.49 0.89, 2.49 0.127  Probably not in utero during Ramadan 1.20 0.96, 1.49 0.103 1.35 0.91, 2.01 0.133 1.22 0.90, 1.66 0.205 1.23 0.74, 2.23 0.364 1.19 0.87, 1.62 0.280 1.45 0.82, 2.57 0.200 Ramadan Exposure Category All Muslims Female Muslims Male Muslims Any Breathing Difficultya (n = 22,893) Wheezingb (n = 26,313) Any Breathing Difficulty (n = 11,768) Wheezing (n = 13,294) Any Breathing Difficulty (n = 11,125) Wheezing (n = 13,019) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadanc 1.17 1.02, 1.35 0.022 1.26 0.97, 1.63 0.087 1.14 0.94, 1.39 0.176 1.09 0.75, 1.57 0.657 1.21 0.99, 1.47 0.061 1.45 1.00, 2.12 0.051 Exposure period  Conceived during Ramadan 1.17 0.97, 1.41 0.100 1.11 0.77, 1.58 0.582 1.14 0.88, 1.48 0.334 0.78 0.46, 1.34 0.371 1.20 0.91, 1.57 0.197 1.47 0.90, 2.42 0.124  Ramadan started in trimester 1 1.21 1.04, 1.41 0.015 1.28 1.00, 1.71 0.099 1.17 0.94, 1.46 0.151 1.18 0.78, 1.78 0.431 1.26 1.01, 1.57 0.038 1.40 0.93, 2.12 0.111  Ramadan started in trimester 2 1.15 0.99, 1.34 0.068 1.30 0.97, 1.74 0.076 1.12 0.90, 1.39 0.324 1.12 0.75, 1.69 0.578 1.19 0.96, 1.48 0.119 1.49 0.99, 2.25 0.058  Ramadan started in trimester 3 1.20 1.02, 1.41 0.032 1.25 0.92, 1.69 0.156 1.17 0.93, 1.48 0.174 1.07 0.69, 1.66 0.761 1.22 0.96, 1.54 0.105 1.46 0.94, 2.27 0.091  Born during Ramadan 1.08 0.89, 1.32 0.439 1.27 0.89, 1.81 0.197 1.09 0.83, 1.43 0.557 1.10 0.67, 1.82 0.705 1.08 0.81, 1.43 0.615 1.49 0.89, 2.49 0.127  Probably not in utero during Ramadan 1.20 0.96, 1.49 0.103 1.35 0.91, 2.01 0.133 1.22 0.90, 1.66 0.205 1.23 0.74, 2.23 0.364 1.19 0.87, 1.62 0.280 1.45 0.82, 2.57 0.200 Abbreviations: CI, confidence interval; OR, odds ratio. a Ever having suffered from any breathing difficulty. The analyses of general breathing difficulties were based on data from the fourth wave of the survey only. b Having experienced wheezing in the past 4 weeks. c Results stem from 2 separate logistic regressions per column (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that adjusted for age, age2, sex, and month of birth. In the wheezing analyses, results were additionally adjusted for survey wave. Standard errors were clustered by household. The results for wheezing confirmed those for any breathing difficulty. Associations were found for exposure during all pregnancy phases except for birth during Ramadan. Significant associations were found only among males. The lower levels of significance may be explained by the lower number of incidences of wheezing in our sample and limited statistical power in the analysis. Smokers and nonsmokers A majority of Indonesian men smoke (69.38% of Muslim men vs. 3.01% of Muslim women in our sample smoked). We allowed the estimates of exposure to vary by smoking status among the male Muslims. The negative associations were consistently stronger for exposed smokers (Table 3). Table 3. Influence of Smoking Status on Associations Between In Utero Exposure to Ramadan and the Occurence of Wheezing in Adulthold (Ages ≥15 Years) Among 13,005 Male Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008a Ramadan Exposure Category Smokers Nonsmokers OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.58 1.02, 2.44 0.042 1.24 0.67, 2.28 0.494 Exposure period  Conceived during Ramadan 1.78 1.01, 3.12 0.045 0.69 0.24, 2.00 0.496  Ramadan started in trimester 1 1.58 0.98, 2.55 0.059 1.03 0.50, 2.13 0.936  Ramadan started in trimester 2 1.64 1.02, 2.65 0.042 1.17 0.57, 2.41 0.665  Ramadan started in trimester 3 1.44 0.85, 2.41 0.174 1.63 0.79, 3.37 0.186  Born during Ramadan 1.40 0.76, 2.58 0.279 1.93 0.83, 4.50 0.129 Ramadan Exposure Category Smokers Nonsmokers OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.58 1.02, 2.44 0.042 1.24 0.67, 2.28 0.494 Exposure period  Conceived during Ramadan 1.78 1.01, 3.12 0.045 0.69 0.24, 2.00 0.496  Ramadan started in trimester 1 1.58 0.98, 2.55 0.059 1.03 0.50, 2.13 0.936  Ramadan started in trimester 2 1.64 1.02, 2.65 0.042 1.17 0.57, 2.41 0.665  Ramadan started in trimester 3 1.44 0.85, 2.41 0.174 1.63 0.79, 3.37 0.186  Born during Ramadan 1.40 0.76, 2.58 0.279 1.93 0.83, 4.50 0.129 Abbreviations: CI, confidence interval; OR, odds ratio. a Results from 2 separate logistic regressions (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that controlled for age, age2, month of birth, survey wave, and the interaction between Ramadan exposure and smoking status. Standard errors were clustered by household. Table 3. Influence of Smoking Status on Associations Between In Utero Exposure to Ramadan and the Occurence of Wheezing in Adulthold (Ages ≥15 Years) Among 13,005 Male Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008a Ramadan Exposure Category Smokers Nonsmokers OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.58 1.02, 2.44 0.042 1.24 0.67, 2.28 0.494 Exposure period  Conceived during Ramadan 1.78 1.01, 3.12 0.045 0.69 0.24, 2.00 0.496  Ramadan started in trimester 1 1.58 0.98, 2.55 0.059 1.03 0.50, 2.13 0.936  Ramadan started in trimester 2 1.64 1.02, 2.65 0.042 1.17 0.57, 2.41 0.665  Ramadan started in trimester 3 1.44 0.85, 2.41 0.174 1.63 0.79, 3.37 0.186  Born during Ramadan 1.40 0.76, 2.58 0.279 1.93 0.83, 4.50 0.129 Ramadan Exposure Category Smokers Nonsmokers OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.58 1.02, 2.44 0.042 1.24 0.67, 2.28 0.494 Exposure period  Conceived during Ramadan 1.78 1.01, 3.12 0.045 0.69 0.24, 2.00 0.496  Ramadan started in trimester 1 1.58 0.98, 2.55 0.059 1.03 0.50, 2.13 0.936  Ramadan started in trimester 2 1.64 1.02, 2.65 0.042 1.17 0.57, 2.41 0.665  Ramadan started in trimester 3 1.44 0.85, 2.41 0.174 1.63 0.79, 3.37 0.186  Born during Ramadan 1.40 0.76, 2.58 0.279 1.93 0.83, 4.50 0.129 Abbreviations: CI, confidence interval; OR, odds ratio. a Results from 2 separate logistic regressions (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that controlled for age, age2, month of birth, survey wave, and the interaction between Ramadan exposure and smoking status. Standard errors were clustered by household. Note that smoking status does not vary with Ramadan exposure (69.4% of unexposed male Muslims vs. 69.3% of exposed male Muslims in our sample smoked). A χ2 test of independence was performed to examine the relationship between Ramadan exposure and smoking status. The relationship between these variables was not significant (χ2 = 0.009, P = 0.927). Associations by age group As Table 4 shows, the risk of experiencing wheezing after in utero exposure to Ramadan increased with age. Significant associations were found when we limited the sample to respondents aged 40 years or more, and they tended to get more pronounced when we limited the sample to even older age groups. Table 4. Associations Between In Utero Exposure to Ramadan and the Occurrence of Wheezing in Adulthood (Ages ≥15 Years), by Age Group, Among Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008a Ramadan Exposure Category Age Group, years <40 (n = 18,176) ≥40 (n = 8,137) ≥45 (n = 6,061) ≥50 (n = 4,329) ≥55 (n = 2,987) ≥60 (n = 1,942) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.16 0.81, 1.66 0.423 1.35 0.90, 2.03 0.150 1.41 0.89, 2.23 0.141 1.59 0.95, 2.66 0.080 1.83 1.02, 3.30 0.045 1.98 0.98, 3.99 0.055 Exposure period  Conceived during Ramadan 1.24 0.78, 1.99 0.367 0.95 0.53, 1.69 0.856 1.10 0.59, 2.05 0.774 1.15 0.57, 2.31 0.696 0.95 0.42, 2.15 0.898 1.00 0.38, 2.61 0.996  Ramadan started in trimester 1 1.19 0.79, 1.79 0.394 1.36 0.87, 2.12 0.177 1.37 0.83, 2.26 0.223 1.62 0.92, 2.86 0.096 1.97 1.03, 3.78 0.041 2.18 1.01, 4.71 0.048  Ramadan started in trimester 2 1.29 0.87, 1.93 0.208 1.29 0.82, 2.02 0.279 1.33 0.80, 2.22 0.275 1.61 0.91, 2.85 0.102 2.15 1.11, 4.17 0.024 2.86 1.28, 6.36 0.010  Ramadan started in trimester 3 0.97 0.64, 1.49 0.903 1.55 0.97, 2.46 0.064 1.73 1.04, 2.88 0.035 1.73 0.97, 3.10 0.064 1.87 0.97, 3.62 0.064 1.71 0.75, 3.94 0.205  Born during Ramadan 1.03 0.62, 1.70 0.918 1.54 0.91, 2.62 0.108 1.44 0.79, 2.63 0.232 1.62 0.83, 3.14 0.155 1.85 0.88, 3.89 0.103 1.97 0.83, 4.64 0.123  Probably not in utero during Ramadan 1.69 1.03, 2.76 0.038 0.87 0.44, 1.75 0.703 0.96 0.45, 2.04 0.921 1.22 0.53, 2.80 0.644 1.49 0.59, 3.76 0.402 2.06 0.74, 5.79 0.168 Ramadan Exposure Category Age Group, years <40 (n = 18,176) ≥40 (n = 8,137) ≥45 (n = 6,061) ≥50 (n = 4,329) ≥55 (n = 2,987) ≥60 (n = 1,942) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.16 0.81, 1.66 0.423 1.35 0.90, 2.03 0.150 1.41 0.89, 2.23 0.141 1.59 0.95, 2.66 0.080 1.83 1.02, 3.30 0.045 1.98 0.98, 3.99 0.055 Exposure period  Conceived during Ramadan 1.24 0.78, 1.99 0.367 0.95 0.53, 1.69 0.856 1.10 0.59, 2.05 0.774 1.15 0.57, 2.31 0.696 0.95 0.42, 2.15 0.898 1.00 0.38, 2.61 0.996  Ramadan started in trimester 1 1.19 0.79, 1.79 0.394 1.36 0.87, 2.12 0.177 1.37 0.83, 2.26 0.223 1.62 0.92, 2.86 0.096 1.97 1.03, 3.78 0.041 2.18 1.01, 4.71 0.048  Ramadan started in trimester 2 1.29 0.87, 1.93 0.208 1.29 0.82, 2.02 0.279 1.33 0.80, 2.22 0.275 1.61 0.91, 2.85 0.102 2.15 1.11, 4.17 0.024 2.86 1.28, 6.36 0.010  Ramadan started in trimester 3 0.97 0.64, 1.49 0.903 1.55 0.97, 2.46 0.064 1.73 1.04, 2.88 0.035 1.73 0.97, 3.10 0.064 1.87 0.97, 3.62 0.064 1.71 0.75, 3.94 0.205  Born during Ramadan 1.03 0.62, 1.70 0.918 1.54 0.91, 2.62 0.108 1.44 0.79, 2.63 0.232 1.62 0.83, 3.14 0.155 1.85 0.88, 3.89 0.103 1.97 0.83, 4.64 0.123  Probably not in utero during Ramadan 1.69 1.03, 2.76 0.038 0.87 0.44, 1.75 0.703 0.96 0.45, 2.04 0.921 1.22 0.53, 2.80 0.644 1.49 0.59, 3.76 0.402 2.06 0.74, 5.79 0.168 Abbreviations: CI, confidence interval; OR, odds ratio. a Results stem from 2 separate logistic regressions per column (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that adjusted for age, age2, sex, month of birth, and survey wave. Standard errors were clustered by household. A separate regression was carried out for each age group. Table 4. Associations Between In Utero Exposure to Ramadan and the Occurrence of Wheezing in Adulthood (Ages ≥15 Years), by Age Group, Among Muslims Living in Predominantly Muslim Areas, Indonesian Family Life Survey, 1997–2008a Ramadan Exposure Category Age Group, years <40 (n = 18,176) ≥40 (n = 8,137) ≥45 (n = 6,061) ≥50 (n = 4,329) ≥55 (n = 2,987) ≥60 (n = 1,942) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.16 0.81, 1.66 0.423 1.35 0.90, 2.03 0.150 1.41 0.89, 2.23 0.141 1.59 0.95, 2.66 0.080 1.83 1.02, 3.30 0.045 1.98 0.98, 3.99 0.055 Exposure period  Conceived during Ramadan 1.24 0.78, 1.99 0.367 0.95 0.53, 1.69 0.856 1.10 0.59, 2.05 0.774 1.15 0.57, 2.31 0.696 0.95 0.42, 2.15 0.898 1.00 0.38, 2.61 0.996  Ramadan started in trimester 1 1.19 0.79, 1.79 0.394 1.36 0.87, 2.12 0.177 1.37 0.83, 2.26 0.223 1.62 0.92, 2.86 0.096 1.97 1.03, 3.78 0.041 2.18 1.01, 4.71 0.048  Ramadan started in trimester 2 1.29 0.87, 1.93 0.208 1.29 0.82, 2.02 0.279 1.33 0.80, 2.22 0.275 1.61 0.91, 2.85 0.102 2.15 1.11, 4.17 0.024 2.86 1.28, 6.36 0.010  Ramadan started in trimester 3 0.97 0.64, 1.49 0.903 1.55 0.97, 2.46 0.064 1.73 1.04, 2.88 0.035 1.73 0.97, 3.10 0.064 1.87 0.97, 3.62 0.064 1.71 0.75, 3.94 0.205  Born during Ramadan 1.03 0.62, 1.70 0.918 1.54 0.91, 2.62 0.108 1.44 0.79, 2.63 0.232 1.62 0.83, 3.14 0.155 1.85 0.88, 3.89 0.103 1.97 0.83, 4.64 0.123  Probably not in utero during Ramadan 1.69 1.03, 2.76 0.038 0.87 0.44, 1.75 0.703 0.96 0.45, 2.04 0.921 1.22 0.53, 2.80 0.644 1.49 0.59, 3.76 0.402 2.06 0.74, 5.79 0.168 Ramadan Exposure Category Age Group, years <40 (n = 18,176) ≥40 (n = 8,137) ≥45 (n = 6,061) ≥50 (n = 4,329) ≥55 (n = 2,987) ≥60 (n = 1,942) OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value OR 95% CI P Value In utero during Ramadan 1.16 0.81, 1.66 0.423 1.35 0.90, 2.03 0.150 1.41 0.89, 2.23 0.141 1.59 0.95, 2.66 0.080 1.83 1.02, 3.30 0.045 1.98 0.98, 3.99 0.055 Exposure period  Conceived during Ramadan 1.24 0.78, 1.99 0.367 0.95 0.53, 1.69 0.856 1.10 0.59, 2.05 0.774 1.15 0.57, 2.31 0.696 0.95 0.42, 2.15 0.898 1.00 0.38, 2.61 0.996  Ramadan started in trimester 1 1.19 0.79, 1.79 0.394 1.36 0.87, 2.12 0.177 1.37 0.83, 2.26 0.223 1.62 0.92, 2.86 0.096 1.97 1.03, 3.78 0.041 2.18 1.01, 4.71 0.048  Ramadan started in trimester 2 1.29 0.87, 1.93 0.208 1.29 0.82, 2.02 0.279 1.33 0.80, 2.22 0.275 1.61 0.91, 2.85 0.102 2.15 1.11, 4.17 0.024 2.86 1.28, 6.36 0.010  Ramadan started in trimester 3 0.97 0.64, 1.49 0.903 1.55 0.97, 2.46 0.064 1.73 1.04, 2.88 0.035 1.73 0.97, 3.10 0.064 1.87 0.97, 3.62 0.064 1.71 0.75, 3.94 0.205  Born during Ramadan 1.03 0.62, 1.70 0.918 1.54 0.91, 2.62 0.108 1.44 0.79, 2.63 0.232 1.62 0.83, 3.14 0.155 1.85 0.88, 3.89 0.103 1.97 0.83, 4.64 0.123  Probably not in utero during Ramadan 1.69 1.03, 2.76 0.038 0.87 0.44, 1.75 0.703 0.96 0.45, 2.04 0.921 1.22 0.53, 2.80 0.644 1.49 0.59, 3.76 0.402 2.06 0.74, 5.79 0.168 Abbreviations: CI, confidence interval; OR, odds ratio. a Results stem from 2 separate logistic regressions per column (first row: exposed vs. not exposed; exposure periods: classification of exposure into different pregnancy phases) that adjusted for age, age2, sex, month of birth, and survey wave. Standard errors were clustered by household. A separate regression was carried out for each age group. Robustness checks We replicated our results using a DID design (Table 5). The estimates for exposed Muslims (upper half of table) confirmed that the strength of the association increased with age. The DID analysis further allowed us to show that associations were not caused by shocks incurred during Ramadan that were independent of religion (such as rising food prices). No associations between wheezing and Ramadan exposure during pregnancy were found for non-Muslims who lived in predominantly Muslim areas (lower half of Table 5). This was confirmed by separately conducting our analyses for non-Muslims. These results again demonstrated that prenatal Ramadan exposure was not correlated with seasonality in our analyses, which also means it is unlikely that our results for Muslims were driven by residual confounding by seasonality. Table 5. Associations Between In Utero Exposure to Ramadan (Exposed vs. Certainly Not Exposed) and the Occurrence of Wheezing in Adulthood (Ages ≥15 Years) Among Muslims and Non-Muslims Living in Predominantly Muslim Areas (Difference-in-Differences Analysisa), Indonesian Family Life Survey, 1997–2008b Parameter and Age Group, years OR 95% CI P Value Muslims living in Muslim areas (exposed × Muslim)  All ages 1.71 0.82, 3.59 0.156  ≥45 1.41 0.39, 5.13 0.601  ≥50 2.09 0.55, 7.99 0.280  ≥55 3.38 0.82, 13.91 0.091 Non-Muslims living in Muslim areas (exposed)  All ages 0.65 0.32, 1.32 0.232  ≥45 0.80 0.23, 2.71 0.714  ≥50 0.63 0.18, 2.22 0.472  ≥55 0.46 0.12, 1.71 0.248 Parameter and Age Group, years OR 95% CI P Value Muslims living in Muslim areas (exposed × Muslim)  All ages 1.71 0.82, 3.59 0.156  ≥45 1.41 0.39, 5.13 0.601  ≥50 2.09 0.55, 7.99 0.280  ≥55 3.38 0.82, 13.91 0.091 Non-Muslims living in Muslim areas (exposed)  All ages 0.65 0.32, 1.32 0.232  ≥45 0.80 0.23, 2.71 0.714  ≥50 0.63 0.18, 2.22 0.472  ≥55 0.46 0.12, 1.71 0.248 Abbreviations: CI, confidence interval; OR, odds ratio. a The interaction of all covariates with religion was included in the analyses. b Results from logistic regression analyses that controlled for religion, exposure to Ramadan, religion × exposure to Ramadan, probably not being exposed, religion × probably not being exposed, age, religion × age, age2, religion × age2, month of birth, religion × month of birth, sex, religion × sex, survey wave, and religion × survey wave. A separate regresssion was carried out for each age group. Standard errors were clustered by household. Table 5. Associations Between In Utero Exposure to Ramadan (Exposed vs. Certainly Not Exposed) and the Occurrence of Wheezing in Adulthood (Ages ≥15 Years) Among Muslims and Non-Muslims Living in Predominantly Muslim Areas (Difference-in-Differences Analysisa), Indonesian Family Life Survey, 1997–2008b Parameter and Age Group, years OR 95% CI P Value Muslims living in Muslim areas (exposed × Muslim)  All ages 1.71 0.82, 3.59 0.156  ≥45 1.41 0.39, 5.13 0.601  ≥50 2.09 0.55, 7.99 0.280  ≥55 3.38 0.82, 13.91 0.091 Non-Muslims living in Muslim areas (exposed)  All ages 0.65 0.32, 1.32 0.232  ≥45 0.80 0.23, 2.71 0.714  ≥50 0.63 0.18, 2.22 0.472  ≥55 0.46 0.12, 1.71 0.248 Parameter and Age Group, years OR 95% CI P Value Muslims living in Muslim areas (exposed × Muslim)  All ages 1.71 0.82, 3.59 0.156  ≥45 1.41 0.39, 5.13 0.601  ≥50 2.09 0.55, 7.99 0.280  ≥55 3.38 0.82, 13.91 0.091 Non-Muslims living in Muslim areas (exposed)  All ages 0.65 0.32, 1.32 0.232  ≥45 0.80 0.23, 2.71 0.714  ≥50 0.63 0.18, 2.22 0.472  ≥55 0.46 0.12, 1.71 0.248 Abbreviations: CI, confidence interval; OR, odds ratio. a The interaction of all covariates with religion was included in the analyses. b Results from logistic regression analyses that controlled for religion, exposure to Ramadan, religion × exposure to Ramadan, probably not being exposed, religion × probably not being exposed, age, religion × age, age2, religion × age2, month of birth, religion × month of birth, sex, religion × sex, survey wave, and religion × survey wave. A separate regresssion was carried out for each age group. Standard errors were clustered by household. DISCUSSION In this study, in utero exposure to Ramadan led to an increased risk of breathing difficulties—specifically wheezing—in adulthood. The associations were most pronounced for smokers. The respiratory systems of prenatally exposed Muslims thus seem to perform worse in mitigating ex utero harmful influences. Our results partly confirm the findings of Lopuhaä et al. (4). Similar to their research on famines, we most consistently found associations between the occurrence of lung dysfunction and exposure to Ramadan for exposure during the first and second trimesters. However, the detected sizes of the associations were independent of the timing of exposure, and associations were also found for exposure during the third trimester. Moreover, the association seemed to increase with age and was strongest in the age group 45 years or more. This is in line with fetal programming theory, suggesting that many consequences of prenatal shocks only manifest in postreproductive age. Our results are to be regarded as intention-to-treat estimates and underestimates of the real strength of the associations. All persons with an overlap between Ramadan and pregnancy were classified as exposed, although we lacked information on actual maternal behavior during Ramadan. Consequently, children of nonfasting mothers were also classified as exposed, which biased the results towards zero. Moreover, all estimates were conditional upon survival, as all persons in the analysis were aged 15 years or older. The health outcomes of persons lost in the womb or before age 15 years and of younger children were not observed. Selectively timing pregnancies to avoid Ramadan during pregnancy is not common in Indonesia. It has been shown that parents of children who were prenatally exposed to Ramadan do not differ from parents of children without prenatal Ramadan exposure (28). Consequently, self-selection of healthy persons into the control group did not confound our results. It has furthermore been shown that persons who were exposed to Ramadan while in utero do not have a general tendency to complain more about their health (28). A main limitation of this study is that our analysis was based on self-reported health indicators that were not diagnosed by qualified personnel. However, because we found similar results for wheezing and general breathing difficulties, a potential mixup of symptoms by the interviewees was accounted for. The transferability of our results to other countries might be limited, because most Indonesian smokers consume kreteks (clove cigarettes). While the scientific evidence on whether kreteks are more or less harmful than conventional cigarettes is inconclusive, we cannot exclude the possibility that special interaction effects between kretek smoking and prenatal Ramadan exposure occur. Because obstructive airway diseases rank among the top causes of death, further research on their origins is essential. The identified effect of interaction between in utero exposure to Ramadan and ex utero exposure to smoking might also be relevant with regard to air pollution and other risk factors for airway diseases. Ramadan behavior during pregnancy is relevant for a large part of the population. Moreover, meal-skipping and dieting during pregnancy resemble intermittent fasting and might lead to similar long-term impacts on health. Sensitization of medical personnel and women of childbearing age to this issue is recommended. ACKNOWLEDGMENTS Author affiliations: Gutenberg School of Management and Economics, Faculty of Law and Economics, Johannes Gutenberg University Mainz, Mainz, Germany (Fabienne Pradella, Reyn van Ewijk). This research was funded by the German Research Foundation (DFG grant 260639091). Conflict of interest: none declared. Abbreviations DID difference-in-differences IFLS Indonesian Family Life Survey REFERENCES 1 Carraro S , Scheltema N , Bont L , et al. . Early-life origins of chronic respiratory diseases: understanding and promoting healthy ageing . Eur Respir J . 2014 ; 44 : 1682 – 1696 . Google Scholar Crossref Search ADS PubMed 2 Bush A . COPD: a pediatric disease . COPD . 2008 ; 5 ( 1 ): 53 – 67 . Google Scholar Crossref Search ADS PubMed 3 Stocks J , Sonnappa S . Early life influences on the development of chronic obstructive pulmonary disease . Ther Adv Respir Dis . 2013 ; 7 ( 3 ): 161 – 173 . Google Scholar Crossref Search ADS PubMed 4 Lopuhaä CE , Roseboom TJ , Osmond C , et al. . Atopy, lung function, and obstructive airways disease after prenatal exposure to famine . Thorax . 2000 ; 55 ( 7 ): 555 – 561 . Google Scholar Crossref Search ADS PubMed 5 Barker DJ . The developmental origins of adult disease . J Am Coll Nutr . 2004 ; 23 ( 6 suppl ): 588S – 595S . Google Scholar Crossref Search ADS PubMed 6 Kasprian G , Balassy C , Brugger PC , et al. . MRI of normal and pathological fetal lung development . Eur J Radiol . 2006 ; 57 ( 2 ): 261 – 270 . Google Scholar Crossref Search ADS PubMed 7 Joshi S , Kotecha S . Lung growth and development . Early Hum Dev . 2007 ; 83 ( 12 ): 789 – 794 . Google Scholar Crossref Search ADS PubMed 8 Harding R , Maritz G . Maternal and fetal origins of lung disease in adulthood . Semin Fetal Neonatal Med . 2012 ; 17 ( 2 ): 67 – 72 . Google Scholar Crossref Search ADS PubMed 9 Pike K , Jane Pillow J , Lucas JS . Long term respiratory consequences of intrauterine growth restriction . Semin Fetal Neonatal Med . 2012 ; 17 ( 2 ): 92 – 98 . Google Scholar Crossref Search ADS PubMed 10 Barker DJ , Godfrey KM , Fall C , et al. . Relation of birth weight and childhood respiratory infection to adult lung function and death from chronic obstructive airways disease . BMJ . 1991 ; 303 ( 6804 ): 671 – 675 . Google Scholar Crossref Search ADS PubMed 11 Stein CE , Kumaran K , Fall CH , et al. . Relation of fetal growth to adult lung function in South India . Thorax . 1997 ; 52 ( 10 ): 895 – 899 . Google Scholar Crossref Search ADS PubMed 12 Edwards CA , Osman LM , Godden DJ , et al. . Relationship between birth weight and adult lung function: controlling for maternal factors . Thorax . 2003 ; 58 ( 12 ): 1061 – 1065 . Google Scholar Crossref Search ADS PubMed 13 Lawlor DA , Ebrahim S , Smith G . Association of birth weight with adult lung function: findings from the British Women’s Heart and Health Study and a meta-analysis . Thorax . 2005 ; 60 ( 10 ): 851 – 858 . Google Scholar Crossref Search ADS PubMed 14 Hancox RJ , Poulton R , Greene JM , et al. . Associations between birth weight, early childhood weight gain, and adult lung function . Thorax . 2009 ; 64 ( 3 ): 228 – 232 . Google Scholar Crossref Search ADS PubMed 15 Pei L , Chen G , Mi J , et al. . Low birth weight and lung function in adulthood: retrospective cohort study in China, 1948–1996 . Pediatrics . 2010 ; 125 ( 4 ): e899 – e905 . Google Scholar Crossref Search ADS PubMed 16 Martino D , Prescott S . Epigenetics and prenatal influences on asthma and allergic airways disease . Chest . 2011 ; 139 ( 3 ): 640 – 647 . Google Scholar Crossref Search ADS PubMed 17 Godfrey KM , Barker DJ . Fetal nutrition and adult disease . Am J Clin Nutr . 2000 ; 71 : 1344S – 1352S . Google Scholar Crossref Search ADS PubMed 18 Metcalfe NB , Monaghan P . Compensation for a bad start: grow now, pay later? Trends Ecol Evol . 2001 ; 16 ( 5 ): 254 – 260 . Google Scholar Crossref Search ADS PubMed 19 Savitri AI , Yadegari N , Bakker J , et al. . Ramadan fasting and newborn’s birth weight in pregnant Muslim women in the Netherlands . Br J Nutr . 2014 ; 112 ( 9 ): 1503 – 1509 . Google Scholar Crossref Search ADS PubMed 20 Robinson T , Raisler J . “Each one is a doctor for herself”: Ramadan fasting among pregnant Muslim women in the United States . Ethn Dis . 2005 ; 15 ( 1 suppl 1 ): S1-99 – S1-103 . 21 Ziaee V , Kihanidoost Z , Younesian M , et al. . The effect of Ramadan fasting on outcome of pregnancy . Iran J Pediatr . 2010 ; 20 ( 2 ): 181 – 186 . Google Scholar PubMed 22 Arab M , Nasrollahi S . Interrelation of Ramadan fasting and birth weight . Med J Islamic World Acad Sci . 2001 ; 14 ( 3 ): 91 – 95 . 23 Joosoph J , Abu J , Yu SL . A survey of fasting during pregnancy . Singapore Med J . 2004 ; 45 ( 12 ): 583 – 586 . Google Scholar PubMed 24 Prentice AM , Prentice A , Lamb WH , et al. . Metabolic consequences of fasting during Ramadan in pregnant and lactating women . Hum Nutr Clin Nutr . 1983 ; 37 ( 4 ): 283 – 284 . Google Scholar PubMed 25 Malhotra A , Scott PH , Scott J , et al. . Metabolic changes in Asian Muslim pregnant mothers observing the Ramadan fast in Britain . Br J Nutr . 1989 ; 61 ( 3 ): 663 – 672 . Google Scholar Crossref Search ADS PubMed 26 Majid MF . The persistent effects of in utero nutrition shocks over the life cycle: evidence from Ramadan fasting . J Dev Econ . 2015 ; 117 : 48 – 57 . Google Scholar Crossref Search ADS 27 van Bilsen LA , Savitri AI , Amelia D , et al. . Predictors of Ramadan fasting during pregnancy . J Epidemiol Glob Health . 2016 ; 6 ( 4 ): 267 – 275 . Google Scholar Crossref Search ADS PubMed 28 van Ewijk R . Long-term health effects on the next generation of Ramadan fasting during pregnancy . J Health Econ . 2011 ; 30 ( 6 ): 1246 – 1260 . Google Scholar Crossref Search ADS PubMed 29 van Ewijk RJ , Painter RC , Roseboom TJ . Associations of prenatal exposure to Ramadan with small stature and thinness in adulthood: results from a large Indonesian population-based study . Am J Epidemiol . 2013 ; 177 ( 8 ): 729 – 736 . Google Scholar Crossref Search ADS PubMed 30 Almond D , Mazumder B , van Ewijk R . In utero Ramadan exposure and children’s academic performance . Econ J . 2015 ; 125 ( 589 ): 1501 – 1533 . Google Scholar Crossref Search ADS 31 Almond D , Mazumder B . Health capital and the prenatal environment: the effect of Ramadan observance during pregnancy . Am Econ J Appl Econ . 2011 ; 3 ( 4 ): 56 – 85 . Google Scholar Crossref Search ADS 32 Chen X . Fetus, fasting, and festival: the persistent effects of in utero social shocks . Int J Health Policy Manag . 2014 ; 3 ( 4 ): 165 – 169 . Google Scholar Crossref Search ADS PubMed 33 Metzger BE , Ravnikar V , Vileisis RA , et al. . “Accelerated starvation” and the skipped breakfast in late normal pregnancy . Lancet . 1982 ; 1 ( 8272 ): 588 – 592 . Google Scholar Crossref Search ADS PubMed 34 Burbos N , Shiner AM , Morris E . Severe metabolic acidosis as a consequence of acute starvation in pregnancy . Arch Gynecol Obstet . 2009 ; 279 ( 3 ): 399 – 400 . Google Scholar Crossref Search ADS PubMed 35 Arab M . Ketonuria and serum glucose of fasting pregnant women at the end of a day in Ramadan . Acta Med Iran . 2004 ; 42 ( 4 ): 209 – 212 . 36 Heppe DH , Medina-Gomez C , Hofman A , et al. . Maternal first-trimester diet and childhood bone mass: the Generation R Study . Am J Clin Nutr . 2013 ; 98 ( 1 ): 224 – 232 . Google Scholar Crossref Search ADS PubMed 37 Rifas-Shiman SL , Rich-Edwards JW , Willett WC , et al. . Changes in dietary intake from the first to the second trimester of pregnancy . Paediatr Perinat Epidemiol . 2006 ; 20 ( 1 ): 35 – 42 . Google Scholar Crossref Search ADS PubMed 38 Wu G , Bazer FW , Cudd TA , et al. . Maternal nutrition and fetal development . J Nutr . 2004 ; 134 ( 9 ): 2169 – 2172 . Google Scholar Crossref Search ADS PubMed 39 Painter RC , Roseboom TJ , Bleker OP . Prenatal exposure to the Dutch famine and disease in later life: an overview . Reprod Toxicol . 2005 ; 20 ( 3 ): 345 – 352 . Google Scholar Crossref Search ADS PubMed 40 Savitri AI , Amelia D , Painter RC , et al. . Ramadan during pregnancy and birth weight of newborns . J Nutr Sci . 2018 ; 7 : e5 . Google Scholar Crossref Search ADS PubMed 41 Larijani B , Zahedi F , Sanjari M , et al. . The effect of Ramadan fasting on fasting serum glucose in healthy adults . Med J Malaysia . 2003 ; 58 ( 5 ): 678 – 680 . Google Scholar PubMed 42 Seckl JR , Holmes MC . Mechanisms of disease: glucocorticoids, their placental metabolism and fetal ‘programming’ of adult pathophysiology . Nat Clin Pract Endocrinol Metab . 2007 ; 3 ( 6 ): 479 – 488 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

L, Kinney, Gregory;A, Santorico, Stephanie;A, Young, Kendra;H, Cho, Michael;J, Castaldi, Peter;Raul, San José Estépar,;C, Ross, James;G, Dy, Jennifer;J, Make, Barry;A, Regan, Elizabeth;A, Lynch, David;C, Everett, Douglas;M, Lutz, Sharon;K, Silverman, Edwin;R, Washko, George;D, Crapo, James;E, Hokanson, John

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy087pmid: 29771274

Abstract Chronic obstructive pulmonary disease (COPD) is a syndrome caused by damage to the lungs that results in decreased pulmonary function and reduced structural integrity. Pulmonary function testing (PFT) is used to diagnose and stratify COPD into severity groups, and computed tomography (CT) imaging of the chest is often used to assess structural changes in the lungs. We hypothesized that the combination of PFT and CT phenotypes would provide a more powerful tool for assessing underlying morphologic differences associated with pulmonary function in COPD than does PFT alone. We used factor analysis of 26 variables to classify 8,157 participants recruited into the COPDGene cohort between January 2008 and June 2011 from 21 clinical centers across the United States. These factors were used as predictors of all-cause mortality using Cox proportional hazards modeling. Five factors explained 80% of the covariance and represented the following domains: factor 1, increased emphysema and decreased pulmonary function; factor 2, airway disease and decreased pulmonary function; factor 3, gas trapping; factor 4, CT variability; and factor 5, hyperinflation. After more than 46,079 person-years of follow-up, factors 1 through 4 were associated with mortality and there was a significant synergistic interaction between factors 1 and 2 on death. Considering CT measures along with PFT in the assessment of COPD can identify patients at particularly high risk for death. chronic obstructive pulmonary disease, Cox proportional hazards, factor analysis, mortality Chronic obstructive pulmonary disease (COPD) is defined by reduced pulmonary function and is associated with reduced quality of life, more hospitalizations, and higher risk for death (1, 2). Cigarette smoking is the major environmental risk factor for development of COPD (3). Although COPD is defined by a ratio of less than 0.70 of forced expiratory volume at 1 second (FEV1) to forced vital capacity (FVC), there is substantial heterogeneity in the clinical and pathological manifestations of the disease (4, 5). Identifying the pathophysiologic processes reflecting underlying disease heterogeneity could lead to more targeted therapies for prevention and treatment. Chest computed tomography (CT) can aid visualization and quantification of anatomic features of the lung; however, some key features of interest are correlated. This presents challenges in the use of these features as covariates in multivariable statistical models but also presents an opportunity to better define multidimensional pathologic processes in COPD. These CT features may represent primary structural abnormalities in the lung that lead to reduced pulmonary function and, ultimately, COPD diagnosis. By taking advantage of the correlation structure of chest CT and pulmonary function data, factor analysis can reduce the number of variables to a small and manageable set of uncorrelated factors (6). We hypothesized that this smaller set of continuous vectors may represent disease axes that can be used to identify the underlying pathophysiologic heterogeneity within COPD. We hypothesized that this method, applied to a large heterogeneous dataset of COPD, would result in novel insights into disease phenotypes and prediction of mortality. METHODS The Genetic Epidemiology of COPD Study The Genetic Epidemiology of COPD (COPDGene) Study is a multicenter (21 clinical sites in the United States), observational study designed so genetic factors associated with COPD can be identified and COPD-related phenotypes characterized (7). Into this study were recruited 10,192 adult current and former smokers who were non-Hispanic whites (two-thirds of cohort) or who were black (one-third of cohort), ages 44–81 years, with at least a 10 pack-year history of smoking. Participants with known COPD were recruited from outpatient pulmonary clinics and other smokers were recruited through personal contact with friends and relatives of clinic patients, advertisements, and outreach to community groups and other organizations. Study centers were instructed to target recruitment of participants without COPD from community sources rather than pulmonary clinics serving other lung diseases. Nonclinic-based recruitment identified participants with and without COPD and identified undetected COPD in many participants. Exclusion criteria were as follows: women who were pregnant, a history of lung disease other than asthma, surgical removal of at least 1 lung lobe, active cancer treatment or suspected lung cancer, chest radiation therapy, metal objects in the chest, recent COPD exacerbation treated with antibiotics or steroids (these patients were invited to participate at a later date), recent eye surgery, past myocardial infarction or other cardiac hospitalization, recent chest or abdominal surgery, inability to use albuterol, a first- or second-degree relative participating in the study, or multiple racial categories. All participants provided written informed consent, and the overall study was approved by the institutional review boards at all of the participating centers. All participants were assessed for pulmonary function using spirometry and lung morphology using high-dose inspiration and low-dose expiration chest CT imaging. Although COPDGene is enriched for COPD cases, participants included people with and without COPD and represented a full range of pulmonary function (Web Figure 1, available at https://academic.oup.com/aje). Thus, COPDGene provides a broad spectrum of pulmonary phenotypes, based on findings of CT imaging, after specific exclusion criteria were applied (7), allowing unique insights into this population of current and former smokers. Table 1 details characteristics of the COPDGene participants included in the present analysis compared with the full cohort. We excluded subjects without complete data on the phenotypes of interest or death follow-up. There were a total of 5,612 non-Hispanic white and 2,545 black participants with complete data. Table 1. Characteristics of Individuals Included in This Analysis With Complete Factor Data Compared With the Entire Cohort, Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011 Characteristic Analysis Cohort (n = 8,157) COPDGene Cohort (n = 10,192) P Value Mean (SD) No. % Mean (SD) No. % Age, years 59.7 (9.0) 59.6 (9.0) 0.45 Female sex 3,775 46.3 4,742 46.5 0.74 Black race 2,545 31.2 3,408 33.4 0.001 Current smoker 4,276 52.4 5,414 53.1 0.35 Pack-years of smoking 44.4 (24.9) 44.2 (25.0) 0.59 COPD (yes) 3,604 44.2 4,484 44.0 0.07 FEV1% predicteda 76.8 (25.3) 76.4 (25.6) 0.29 FEV1/FVC 0.67 (0.16) 0.67 (0.16) 1.00 COPD classificationb 0.88  PRISM 992 12.2 1,257 12.4  Gold 0 3,561 43.7 4,388 43.3  Gold 1 651 8.0 794 7.8  Gold 2 1,574 19.3 1,922 19.0  Gold 3 918 11.2 1,162 11.5  Gold 4 461 5.6 606 6.0 Characteristic Analysis Cohort (n = 8,157) COPDGene Cohort (n = 10,192) P Value Mean (SD) No. % Mean (SD) No. % Age, years 59.7 (9.0) 59.6 (9.0) 0.45 Female sex 3,775 46.3 4,742 46.5 0.74 Black race 2,545 31.2 3,408 33.4 0.001 Current smoker 4,276 52.4 5,414 53.1 0.35 Pack-years of smoking 44.4 (24.9) 44.2 (25.0) 0.59 COPD (yes) 3,604 44.2 4,484 44.0 0.07 FEV1% predicteda 76.8 (25.3) 76.4 (25.6) 0.29 FEV1/FVC 0.67 (0.16) 0.67 (0.16) 1.00 COPD classificationb 0.88  PRISM 992 12.2 1,257 12.4  Gold 0 3,561 43.7 4,388 43.3  Gold 1 651 8.0 794 7.8  Gold 2 1,574 19.3 1,922 19.0  Gold 3 918 11.2 1,162 11.5  Gold 4 461 5.6 606 6.0 Abbreviations: FEV1%, percentage of forced expiratory volume at 1 second; COPDGene, the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study; GOLD, Global Initiative for Chronic Obstructive Lung Disease; PRISM, preserved ratio impaired spirometry; SD, standard deviation. a Based on Third National Health and Nutrition Examination Survey reference values (9). b GOLD stage is missing for 63 patients in the COPDGene cohort because spirometry could not be completed or spirometry data were missing. Percentages are calculated on the basis of a total of 10,129 patients. Table 1. Characteristics of Individuals Included in This Analysis With Complete Factor Data Compared With the Entire Cohort, Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011 Characteristic Analysis Cohort (n = 8,157) COPDGene Cohort (n = 10,192) P Value Mean (SD) No. % Mean (SD) No. % Age, years 59.7 (9.0) 59.6 (9.0) 0.45 Female sex 3,775 46.3 4,742 46.5 0.74 Black race 2,545 31.2 3,408 33.4 0.001 Current smoker 4,276 52.4 5,414 53.1 0.35 Pack-years of smoking 44.4 (24.9) 44.2 (25.0) 0.59 COPD (yes) 3,604 44.2 4,484 44.0 0.07 FEV1% predicteda 76.8 (25.3) 76.4 (25.6) 0.29 FEV1/FVC 0.67 (0.16) 0.67 (0.16) 1.00 COPD classificationb 0.88  PRISM 992 12.2 1,257 12.4  Gold 0 3,561 43.7 4,388 43.3  Gold 1 651 8.0 794 7.8  Gold 2 1,574 19.3 1,922 19.0  Gold 3 918 11.2 1,162 11.5  Gold 4 461 5.6 606 6.0 Characteristic Analysis Cohort (n = 8,157) COPDGene Cohort (n = 10,192) P Value Mean (SD) No. % Mean (SD) No. % Age, years 59.7 (9.0) 59.6 (9.0) 0.45 Female sex 3,775 46.3 4,742 46.5 0.74 Black race 2,545 31.2 3,408 33.4 0.001 Current smoker 4,276 52.4 5,414 53.1 0.35 Pack-years of smoking 44.4 (24.9) 44.2 (25.0) 0.59 COPD (yes) 3,604 44.2 4,484 44.0 0.07 FEV1% predicteda 76.8 (25.3) 76.4 (25.6) 0.29 FEV1/FVC 0.67 (0.16) 0.67 (0.16) 1.00 COPD classificationb 0.88  PRISM 992 12.2 1,257 12.4  Gold 0 3,561 43.7 4,388 43.3  Gold 1 651 8.0 794 7.8  Gold 2 1,574 19.3 1,922 19.0  Gold 3 918 11.2 1,162 11.5  Gold 4 461 5.6 606 6.0 Abbreviations: FEV1%, percentage of forced expiratory volume at 1 second; COPDGene, the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study; GOLD, Global Initiative for Chronic Obstructive Lung Disease; PRISM, preserved ratio impaired spirometry; SD, standard deviation. a Based on Third National Health and Nutrition Examination Survey reference values (9). b GOLD stage is missing for 63 patients in the COPDGene cohort because spirometry could not be completed or spirometry data were missing. Percentages are calculated on the basis of a total of 10,129 patients. Pulmonary function variables Pulmonary function testing was performed following American Thoracic Society guidelines (8) using the Easy-One spirometer (ndd Medical Technologies, Andover, Massachusetts). Spirometry was performed at baseline and repeated after the administration of 180 μg of inhaled albuterol. FEV1, FVC, peak expiratory flow, and forced expiratory flow at 25%–75% of the FVC were obtained for all participants, FEV1 percent (FEV1%) predicted and FVC percent (FCV%) predicted values were calculated using the Third National Health and Nutrition Examination Survey reference values (9). The FEV1/FVC ratio was calculated using the absolute measures in liters. FEV1% predicted, FVC% predicted, peak expiratory flow, forced expiratory flow at 25%–75% of the FVC, and FEV1/FVC measure unique aspects of pulmonary function and each was included in the factor analysis. CT variables CT images of the chest were acquired at full inspiration and relaxed exhalation, as described previously (10, 11). Multiple CT-based metrics of the lung were obtained, including densitometric assessments of the lung parenchyma to provide objective assessments of emphysema-like tissue (CT threshold of −950 Hounsfield units (HU) and −910 HU expressed as a percentage of total lung parenchyma) and gas trapping (i.e., expiration-to-inspiration attenuation ratio, which is thought to reflect small airways disease). The percentage of lung tissue below a threshold of −856 HU was used to represent a quantitative metric of gas trapping on the expiratory CT scan. Additional measures of central airway morphology were also used to provide objective assessments of airway wall thickening. One such measure is the wall area percentage. This is calculated as the 100 multiplied by the ratio of the airway wall area divided by the total bronchial cross-sectional area (wall plus lumen). Multiple investigations have demonstrated that increases in these measures (e.g., increased wall area percentage) reflect airway wall thickening and spirometric and clinical impairment. These measures are commonly obtained at select sites in the tracheobronchial tree, such as the third-generation (i.e., segmental) airways. Measures of airway wall thickness and the square root of the wall area of derived airways with lumen circumference of 10 mm and 15 mm were calculated as described previously (12, 13). Total lung capacity was measured in liters using volumetric CT imaging obtained during a breath hold at full inspiration with the subject supine. Functional residual capacity was measured in liters using volumetric CT scans obtained at the end of relaxed exhalation while supine. CT variables included in the factor analysis were designed to represent the broad range of CT phenotypes related to COPD. Other covariates Height was measured in centimeters using a stadiometer, weight was measured in kilograms,, and body mass index (BMI) was calculated by dividing the weight by height squared (using weight in kilograms and height in meters) (14). Current smoking status and pack-years of smoking were determined by questionnaire. Self-reported physician diagnoses of comorbid conditions were also determined by questionnaire. Statistical methods for factor analysis Before conducting the factor analysis, the distributions of variables were assessed for normality and Box-Cox transformations were considered for each nonnormally distributed variable. See Web Table 1 for transformations that were performed. Note that for all transformations, the scale direction was preserved to facilitate interpretation of the loading scores in clinically relevant terms. Before performing factor analyses in the full cohort, we stratified the cohort by sex and race (non-Hispanic white or black) and assessed the dimensionality of the variables, each centered at 0 and scaled to have variance of 1, using principal components analysis based on the number of eigenvalues that were greater than 1. In addition, factor analysis was performed in the 4 strata we compared for factor similarities and differences. Horn’s parallel analysis was also conducted based on factor analysis fit to minimize the sum of squares of off-diagonal residuals of the resulting correlation matrix (15). Factor scores were computed using the Varimax rotation. Analyses were all conducted in R, version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria) using the psych package. All-cause mortality Assessment of death in COPDGene was conducted using multiple approaches. A longitudinal follow-up data collection effort was conducted using automated telephony and web-based survey instruments every 6 months for all available participants (16). Participant contact through this system resulted in identification of deceased participants and subsequent follow-up request for confirmation of death. Searches based on the Social Security Death Index (SSDI) are also conducted at regular intervals in COPDGene. Individual study-center institutional review board restrictions allowed an SSDI search to be conducted for 8,675 subjects in October 2016 by a central study search and by 9 sites performing their own searches. Results were aggregated centrally. Assessment of vital status (i.e., alive vs. dead) was backdated 3 months to account for the expected lag time between death and its appearance in the SSDI dataset. We included 333 participants who were unable to be searched through SSDI but were active participants in the longitudinal follow-up (participant returned a longitudinal follow-up survey within 7 months of the search). Participant follow-up time was the time between their baseline study visit and SSDI identified death, report of death from institutional review board–restricted study centers, or most recent, active, longitudinal follow-up participation. A total of 1,454 participants have been lost to follow-up (i.e., they have no SSDI identifier and no study contact has been made after the baseline study visit). The Cox proportional hazard model, based on time to death, was used to model prediction of death in the sample. Continuous factor scores were tested for interactions as well as nonlinear associations with death. RESULTS The study population for the current analysis with complete data is similar to the overall cohort with respect to age (P = 0.5), sex (P = 0.7), smoking status (P = 0.4), and history of pack-years of smoking (P = 0.6), but differed with respect to race, with fewer black participants included (P = 0.001) (Table 1). There were no differences in COPD case status (P = 0.07), pulmonary function (P for FEV1 = 0.3 and P for FEV1/FVC = 1.0), or Global Initiative for Chronic Obstructive Lung Disease stage (a measure of COPD severity) (P = 0.9) between the study population with complete data and the overall cohort. Principal component analysis was performed separately in the subgroups (male non-Hispanic whites, female non-Hispanic whites, male blacks, female blacks) to assess the dimensionality of the underlying factor model and all models yielded 5 to 6 eigenvalues greater than 1. The first 6 principal components explained between 82% and 85% of the variability for all groups. Horn’s parallel analysis indicated no more than 7 factors existed, and that no more than 6 principal components explained variability beyond background noise. Beginning with the white male group (n = 2,973), factor analysis was conducted starting with 7 factors with factors subsequently removed until all factors had absolute factor loadings greater than 0.7. This yielded a 5-factor model. Likewise, in each of the other subgroups, a 5-factor model was supported by the data. Correlating these factor loadings among all subgroups revealed 5 consistent factors. Using the factor model from the white male subgroup, factor scores were derived for each of the other subgroups. These were then correlated with the scores derived from their respective subgroup analyses. The correlations were all quite high: 0.84 for factor 2 in female black participants and greater than 0.96 for all other subgroups and all 5 factors (Web Table 2). These correlations support the same underlying factor model for all subgroups. Given evidence for the same underlying factors explaining correlation among the variables, a single-factor model was fit on the basis of the combined set of data, using the same approach. From the factor analysis with varimax rotation, 5 factors were identified (Table 2). These 5 factors explained 80% of the total variance of the 26 variables included in the final analysis, with the first factor accounting for 37% of the variance of these measures and the remaining factors accounting for progressively less of the total variance—17%, 10%, 9%, and 7% of the total variance, respectively, for factors 2 through 5. These factors accounted for a majority of the individual-measure variances of pulmonary function (72%–98%), inspiration CT density measures (52%–99%), expiration CT measures (74%–99%), but substantially less of the specific airway disease measurements (33%–36%) (Table 2). Table 2. Factor Loadings for a 5-Factor Model Based on the Combined Data Set (n = 8,157), Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011 Variable Emphysema Disease Axis Airway Disease Axis Gas Trapping CT Intensity Variability TLC and FRC Proportion of Variance Explaineda Pulmonary function  FEV1/FVC −0.63b −0.61b −0.23 0.12 0.05 0.83  FEV1% predictedc −0.41 −0.89b −0.1 0.03 −0.08 0.98  FVC% predictedc −0.06 −0.82b 0.06 −0.05 −0.19 0.72  Peak expiratory flow −0.4 −0.7b −0.19 0.08 0.16 0.72  Forced expiratory flow 25%–75% −0.51b −0.7b −0.22 0.07 0.11 0.81  TLC (predicted-race adjusted) −0.04 −0.08 −0.03 0.01 0.95b 0.9  FRC (predicted-race adjusted) 0.05 −0.02 0.04 0.03 0.94b 0.88 Inspiratory CT  Less than −856 HU 0.86b −0.18 0.03 −0.44 0.01 0.96  Less than −910 HU 0.96b −0.05 0.11 −0.2 0 0.98  Less than −950 HU 0.96b 0.08 0.18 0.1 0 0.96  Inspiration histogram, 15th percentile −0.94b −0.01 −0.12 0.19 0.01 0.93  Emphysema, lower one-third, % 0.92b 0.04 0.12 0.02 0.05 0.87  Emphysema, upper one-third, % 0.91b 0.09 0.19 0.15 −0.04 0.89  Inspiration intensity, mean −0.88b 0.06 −0.06 0.41 −0.01 0.94  Inspiration intensity, SD −0.03 0.26 0.12 0.66b −0.01 0.52  Exp/insp attenuation ratio 0.31 0.41 0.81b −0.24 −0.01 0.99 Expiratory CT  Less than −910 HU 0.75b 0.33 0.54b 0.15 0.03 0.98   Less than −950 HU 0.73b 0.33 0.48 0.3 0.04 0.97  Gas trapping, % 0.68b 0.23 0.65b −0.05 0.03 0.95  Expiration histogram, 15th percentile −0.66b −0.25 −0.58b 0.1 −0.01 0.84  Expiration intensity, mean −0.62b −0.27 −0.64b 0.36 0 0.99  Expiration intensity, SD 0.1 −0.1 −0.19 0.82b 0.05 0.74 Airway measurements  Wall area, % segmental −0.07 0.57b 0.13 0.07 −0.08 0.36  Pi 10 −0.16 0.51b 0.13 0.16 −0.1 0.34  Pi 15 −0.21 0.51b 0.1 0.14 −0.02 0.33  BMI −0.26 0.07 −0.13 0.50b 0.01 0.33 Proportion of variance explained 0.37 0.17 0.1 0.09 0.07 Cumulative variance explained 0.37 0.53 0.64 0.72 0.8 Variable Emphysema Disease Axis Airway Disease Axis Gas Trapping CT Intensity Variability TLC and FRC Proportion of Variance Explaineda Pulmonary function  FEV1/FVC −0.63b −0.61b −0.23 0.12 0.05 0.83  FEV1% predictedc −0.41 −0.89b −0.1 0.03 −0.08 0.98  FVC% predictedc −0.06 −0.82b 0.06 −0.05 −0.19 0.72  Peak expiratory flow −0.4 −0.7b −0.19 0.08 0.16 0.72  Forced expiratory flow 25%–75% −0.51b −0.7b −0.22 0.07 0.11 0.81  TLC (predicted-race adjusted) −0.04 −0.08 −0.03 0.01 0.95b 0.9  FRC (predicted-race adjusted) 0.05 −0.02 0.04 0.03 0.94b 0.88 Inspiratory CT  Less than −856 HU 0.86b −0.18 0.03 −0.44 0.01 0.96  Less than −910 HU 0.96b −0.05 0.11 −0.2 0 0.98  Less than −950 HU 0.96b 0.08 0.18 0.1 0 0.96  Inspiration histogram, 15th percentile −0.94b −0.01 −0.12 0.19 0.01 0.93  Emphysema, lower one-third, % 0.92b 0.04 0.12 0.02 0.05 0.87  Emphysema, upper one-third, % 0.91b 0.09 0.19 0.15 −0.04 0.89  Inspiration intensity, mean −0.88b 0.06 −0.06 0.41 −0.01 0.94  Inspiration intensity, SD −0.03 0.26 0.12 0.66b −0.01 0.52  Exp/insp attenuation ratio 0.31 0.41 0.81b −0.24 −0.01 0.99 Expiratory CT  Less than −910 HU 0.75b 0.33 0.54b 0.15 0.03 0.98   Less than −950 HU 0.73b 0.33 0.48 0.3 0.04 0.97  Gas trapping, % 0.68b 0.23 0.65b −0.05 0.03 0.95  Expiration histogram, 15th percentile −0.66b −0.25 −0.58b 0.1 −0.01 0.84  Expiration intensity, mean −0.62b −0.27 −0.64b 0.36 0 0.99  Expiration intensity, SD 0.1 −0.1 −0.19 0.82b 0.05 0.74 Airway measurements  Wall area, % segmental −0.07 0.57b 0.13 0.07 −0.08 0.36  Pi 10 −0.16 0.51b 0.13 0.16 −0.1 0.34  Pi 15 −0.21 0.51b 0.1 0.14 −0.02 0.33  BMI −0.26 0.07 −0.13 0.50b 0.01 0.33 Proportion of variance explained 0.37 0.17 0.1 0.09 0.07 Cumulative variance explained 0.37 0.53 0.64 0.72 0.8 Abbreviations; BMI, body mass index; CT, computed tomography; Exp, expiration; FEV1/FVC, ratio of forced expiratory volume at 1 second to forced vital capacity; FEV1%, percentage of forced expiratory volume at 1 second; FRC, functional residual capacity; FVC%, percentage of forced vital capacity; HU, Hounsfield unit; Insp, inspiration; Pi 10, airway wall thickness at an internal perimeter of 10 mm; Pi 15, airway wall thickness at an internal perimeter of 15 mm ; SD, standard deviation; TLC, total lung capacity. a Proportion of variance in the row variable explained by the 5 factors. b Factors loading ≥│0.5│. c Based on Third National Health and Nutrition Examination Survey reference values (9). Table 2. Factor Loadings for a 5-Factor Model Based on the Combined Data Set (n = 8,157), Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011 Variable Emphysema Disease Axis Airway Disease Axis Gas Trapping CT Intensity Variability TLC and FRC Proportion of Variance Explaineda Pulmonary function  FEV1/FVC −0.63b −0.61b −0.23 0.12 0.05 0.83  FEV1% predictedc −0.41 −0.89b −0.1 0.03 −0.08 0.98  FVC% predictedc −0.06 −0.82b 0.06 −0.05 −0.19 0.72  Peak expiratory flow −0.4 −0.7b −0.19 0.08 0.16 0.72  Forced expiratory flow 25%–75% −0.51b −0.7b −0.22 0.07 0.11 0.81  TLC (predicted-race adjusted) −0.04 −0.08 −0.03 0.01 0.95b 0.9  FRC (predicted-race adjusted) 0.05 −0.02 0.04 0.03 0.94b 0.88 Inspiratory CT  Less than −856 HU 0.86b −0.18 0.03 −0.44 0.01 0.96  Less than −910 HU 0.96b −0.05 0.11 −0.2 0 0.98  Less than −950 HU 0.96b 0.08 0.18 0.1 0 0.96  Inspiration histogram, 15th percentile −0.94b −0.01 −0.12 0.19 0.01 0.93  Emphysema, lower one-third, % 0.92b 0.04 0.12 0.02 0.05 0.87  Emphysema, upper one-third, % 0.91b 0.09 0.19 0.15 −0.04 0.89  Inspiration intensity, mean −0.88b 0.06 −0.06 0.41 −0.01 0.94  Inspiration intensity, SD −0.03 0.26 0.12 0.66b −0.01 0.52  Exp/insp attenuation ratio 0.31 0.41 0.81b −0.24 −0.01 0.99 Expiratory CT  Less than −910 HU 0.75b 0.33 0.54b 0.15 0.03 0.98   Less than −950 HU 0.73b 0.33 0.48 0.3 0.04 0.97  Gas trapping, % 0.68b 0.23 0.65b −0.05 0.03 0.95  Expiration histogram, 15th percentile −0.66b −0.25 −0.58b 0.1 −0.01 0.84  Expiration intensity, mean −0.62b −0.27 −0.64b 0.36 0 0.99  Expiration intensity, SD 0.1 −0.1 −0.19 0.82b 0.05 0.74 Airway measurements  Wall area, % segmental −0.07 0.57b 0.13 0.07 −0.08 0.36  Pi 10 −0.16 0.51b 0.13 0.16 −0.1 0.34  Pi 15 −0.21 0.51b 0.1 0.14 −0.02 0.33  BMI −0.26 0.07 −0.13 0.50b 0.01 0.33 Proportion of variance explained 0.37 0.17 0.1 0.09 0.07 Cumulative variance explained 0.37 0.53 0.64 0.72 0.8 Variable Emphysema Disease Axis Airway Disease Axis Gas Trapping CT Intensity Variability TLC and FRC Proportion of Variance Explaineda Pulmonary function  FEV1/FVC −0.63b −0.61b −0.23 0.12 0.05 0.83  FEV1% predictedc −0.41 −0.89b −0.1 0.03 −0.08 0.98  FVC% predictedc −0.06 −0.82b 0.06 −0.05 −0.19 0.72  Peak expiratory flow −0.4 −0.7b −0.19 0.08 0.16 0.72  Forced expiratory flow 25%–75% −0.51b −0.7b −0.22 0.07 0.11 0.81  TLC (predicted-race adjusted) −0.04 −0.08 −0.03 0.01 0.95b 0.9  FRC (predicted-race adjusted) 0.05 −0.02 0.04 0.03 0.94b 0.88 Inspiratory CT  Less than −856 HU 0.86b −0.18 0.03 −0.44 0.01 0.96  Less than −910 HU 0.96b −0.05 0.11 −0.2 0 0.98  Less than −950 HU 0.96b 0.08 0.18 0.1 0 0.96  Inspiration histogram, 15th percentile −0.94b −0.01 −0.12 0.19 0.01 0.93  Emphysema, lower one-third, % 0.92b 0.04 0.12 0.02 0.05 0.87  Emphysema, upper one-third, % 0.91b 0.09 0.19 0.15 −0.04 0.89  Inspiration intensity, mean −0.88b 0.06 −0.06 0.41 −0.01 0.94  Inspiration intensity, SD −0.03 0.26 0.12 0.66b −0.01 0.52  Exp/insp attenuation ratio 0.31 0.41 0.81b −0.24 −0.01 0.99 Expiratory CT  Less than −910 HU 0.75b 0.33 0.54b 0.15 0.03 0.98   Less than −950 HU 0.73b 0.33 0.48 0.3 0.04 0.97  Gas trapping, % 0.68b 0.23 0.65b −0.05 0.03 0.95  Expiration histogram, 15th percentile −0.66b −0.25 −0.58b 0.1 −0.01 0.84  Expiration intensity, mean −0.62b −0.27 −0.64b 0.36 0 0.99  Expiration intensity, SD 0.1 −0.1 −0.19 0.82b 0.05 0.74 Airway measurements  Wall area, % segmental −0.07 0.57b 0.13 0.07 −0.08 0.36  Pi 10 −0.16 0.51b 0.13 0.16 −0.1 0.34  Pi 15 −0.21 0.51b 0.1 0.14 −0.02 0.33  BMI −0.26 0.07 −0.13 0.50b 0.01 0.33 Proportion of variance explained 0.37 0.17 0.1 0.09 0.07 Cumulative variance explained 0.37 0.53 0.64 0.72 0.8 Abbreviations; BMI, body mass index; CT, computed tomography; Exp, expiration; FEV1/FVC, ratio of forced expiratory volume at 1 second to forced vital capacity; FEV1%, percentage of forced expiratory volume at 1 second; FRC, functional residual capacity; FVC%, percentage of forced vital capacity; HU, Hounsfield unit; Insp, inspiration; Pi 10, airway wall thickness at an internal perimeter of 10 mm; Pi 15, airway wall thickness at an internal perimeter of 15 mm ; SD, standard deviation; TLC, total lung capacity. a Proportion of variance in the row variable explained by the 5 factors. b Factors loading ≥│0.5│. c Based on Third National Health and Nutrition Examination Survey reference values (9). CT measures of quantitative measures of emphysema on CT scan loaded strongly on factor 1, with the highest factor-loading scores being inspiratory CT volume less than −910 HU and less than −950 HU (factor loading = 0.96 for both) (Table 2). Measures of emphysema distribution also loaded highest on factor 1, as did the analogous measures of density on the expiration CT scans. Pulmonary function measures FEV1/FVC, FEV1% predicted, and forced expiratory flow 25%–75% loaded negatively on factor 1 (−0.63, −0.41, and −0.51, respectively). The airway measurements did not load strongly on factor 1. Based on the factor-loading scores, high CT measures of lung density with concomitant low pulmonary function, we interpret factor 1 to represent a multidimensional (i.e., low attenuation areas and lower pulmonary function) emphysema disease axis. Factor 2 was represented by strong factor loadings for the physiologic pulmonary function measures, with factor-loading scores range from −0.61 to −0.89, except for total lung capacity and functional residual capacity (Table 2). Airway measurements of CT morphology loaded highest on factor 2: 0.57 for segmental wall area percentage and 0.51 for square root of the wall area of derived airways with lumen circumference of 10 mm and 15 mm. The measures of CT density, particularly from the inspiratory CT scans, did not load strongly on factor 2. The morphologic measures of the airways, combined with low pulmonary function loading on factor 2, indicated to us that factor 2 represents a multidimensional airway disease axis. Physiologic measures of pulmonary function did not load strongly on factors 3 or 4. CT measures of low attenuation on expiratory CT scans, which are not present on inspiration CT images, indicate gas trapping and represent a gas-trapping disease axis (factor-loading score for expiration-to-inspiration attenuation ratio, 0.81). Measures of CT density variability measured by the standard deviation of the CT histogram and BMI loaded on factor 4 and therefore represent a complex axis capturing risk associated with BMI (low BMI possibly suggesting cachexia, high BMI suggesting obesity), as well as CT “noise” potentially capturing risk associated with both high and low attenuation present in individuals (e.g., low attenuation attributable to emphysema combined with high attenuation attributable to fibrotic lung diseases). Total lung capacity and functional residual capacity are the only variables that loaded strongly on factor 5 (Table 2). The coefficients for the derivation of factor scores of this model are provided in Web Table 3. Relationships between disease axes and all-cause mortality A total of 950 deaths occurred over a mean follow-up time of 6.3 years, representing 46,079 person-years of follow-up. Older age (P < 0.0001), male sex (P = 0.018), being a current versus former smoker (P = 0.048), and pack-years of smoking (P = 0.0003) were all positively associated with death (Table 3). Lower BMI was associated with higher risk for death (P < 0.0001) in this population of current and former smokers enriched for more severe COPD. The emphysema disease axis was associated with higher risk for death (P = 0.045). The airway disease axis was associated with the greatest risk for death (P < 0.0001) and, in addition to the linear term, a squared term for the airway disease axis was also significantly associated with greater risk for death (P = 0.027). Table 3. Cox Proportional Hazard Model of Death, Survival Follow-up December 2016, the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study Cohort, United States, 2008–2011 Variable β Estimate Standard Error P Value Hazard Ratio 95% CI Age 0.03058 0.00556 <0.0001 1.031 1.02, 1.04 Male sex 0.00334 0.00141 0.0178 1.003 1.00, 1.01 Current smoker 0.19166 0.09702 0.0482 1.211 1.002, 1.465 Pack-years of smoking 0.30179 0.08248 0.0003 1.352 1.15, 1.59 BMI (1 unit) −0.0569 0.00867 <0.0001 0.945 0.929, 0.961 High blood pressure 0.17596 0.08129 0.0304 1.192 1.017, 1.398 Emphysema disease axisa 0.11584 0.05784 0.0452 1.123 1.002, 1.258 Airway disease axisb 0.64179 0.05853 <0.0001 1.900 1.694, 2.131 Interaction between the emphysema and airway disease axes 0.16892 0.0474 0.0004 Airway disease axis squared 0.07635 0.0345 0.0269 1.079 1.009, 1.155 Gas trapping 0.03526 0.05231 0.5004 1.036 0.935, 1.148 CT intensity variability/noise 0.30924 0.04911 <0.0001 1.362 1.237, 1.500 TLC and FRC −0.0398 0.04004 0.3205 0.961 0.888, 1.039 Variable β Estimate Standard Error P Value Hazard Ratio 95% CI Age 0.03058 0.00556 <0.0001 1.031 1.02, 1.04 Male sex 0.00334 0.00141 0.0178 1.003 1.00, 1.01 Current smoker 0.19166 0.09702 0.0482 1.211 1.002, 1.465 Pack-years of smoking 0.30179 0.08248 0.0003 1.352 1.15, 1.59 BMI (1 unit) −0.0569 0.00867 <0.0001 0.945 0.929, 0.961 High blood pressure 0.17596 0.08129 0.0304 1.192 1.017, 1.398 Emphysema disease axisa 0.11584 0.05784 0.0452 1.123 1.002, 1.258 Airway disease axisb 0.64179 0.05853 <0.0001 1.900 1.694, 2.131 Interaction between the emphysema and airway disease axes 0.16892 0.0474 0.0004 Airway disease axis squared 0.07635 0.0345 0.0269 1.079 1.009, 1.155 Gas trapping 0.03526 0.05231 0.5004 1.036 0.935, 1.148 CT intensity variability/noise 0.30924 0.04911 <0.0001 1.362 1.237, 1.500 TLC and FRC −0.0398 0.04004 0.3205 0.961 0.888, 1.039 Abbreviations: BMI, body mass index; CI, confidence interval; CT, computed tomography; FRC, functional residual capacity; TLC, total lung capacity. a Hazard ratio for factor 1 is presented for factor 2 = 0. b Hazard ratio for factor 2 is presented for factor 1 = 0. Table 3. Cox Proportional Hazard Model of Death, Survival Follow-up December 2016, the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study Cohort, United States, 2008–2011 Variable β Estimate Standard Error P Value Hazard Ratio 95% CI Age 0.03058 0.00556 <0.0001 1.031 1.02, 1.04 Male sex 0.00334 0.00141 0.0178 1.003 1.00, 1.01 Current smoker 0.19166 0.09702 0.0482 1.211 1.002, 1.465 Pack-years of smoking 0.30179 0.08248 0.0003 1.352 1.15, 1.59 BMI (1 unit) −0.0569 0.00867 <0.0001 0.945 0.929, 0.961 High blood pressure 0.17596 0.08129 0.0304 1.192 1.017, 1.398 Emphysema disease axisa 0.11584 0.05784 0.0452 1.123 1.002, 1.258 Airway disease axisb 0.64179 0.05853 <0.0001 1.900 1.694, 2.131 Interaction between the emphysema and airway disease axes 0.16892 0.0474 0.0004 Airway disease axis squared 0.07635 0.0345 0.0269 1.079 1.009, 1.155 Gas trapping 0.03526 0.05231 0.5004 1.036 0.935, 1.148 CT intensity variability/noise 0.30924 0.04911 <0.0001 1.362 1.237, 1.500 TLC and FRC −0.0398 0.04004 0.3205 0.961 0.888, 1.039 Variable β Estimate Standard Error P Value Hazard Ratio 95% CI Age 0.03058 0.00556 <0.0001 1.031 1.02, 1.04 Male sex 0.00334 0.00141 0.0178 1.003 1.00, 1.01 Current smoker 0.19166 0.09702 0.0482 1.211 1.002, 1.465 Pack-years of smoking 0.30179 0.08248 0.0003 1.352 1.15, 1.59 BMI (1 unit) −0.0569 0.00867 <0.0001 0.945 0.929, 0.961 High blood pressure 0.17596 0.08129 0.0304 1.192 1.017, 1.398 Emphysema disease axisa 0.11584 0.05784 0.0452 1.123 1.002, 1.258 Airway disease axisb 0.64179 0.05853 <0.0001 1.900 1.694, 2.131 Interaction between the emphysema and airway disease axes 0.16892 0.0474 0.0004 Airway disease axis squared 0.07635 0.0345 0.0269 1.079 1.009, 1.155 Gas trapping 0.03526 0.05231 0.5004 1.036 0.935, 1.148 CT intensity variability/noise 0.30924 0.04911 <0.0001 1.362 1.237, 1.500 TLC and FRC −0.0398 0.04004 0.3205 0.961 0.888, 1.039 Abbreviations: BMI, body mass index; CI, confidence interval; CT, computed tomography; FRC, functional residual capacity; TLC, total lung capacity. a Hazard ratio for factor 1 is presented for factor 2 = 0. b Hazard ratio for factor 2 is presented for factor 1 = 0. We explored both the airway and emphysema axes by assessing their relationship with death, using deciles of each axis. The nonlinear risk observed across deciles of both axes prompted us to test for a statistical relationship (Web Figure 2). Furthermore, there was a statistically significant synergistic interaction between the emphysema disease axis and the airway disease axis (P = 0.001) on death risk. Neither the gas-trapping factor nor the total lung capacity and functional residual capacity factor were related to all-cause mortality (P = 0.5 and P = 0.3, respectively). The CT intensity variability factor that included BMI was positively associated with a higher risk of death (P < 0.0001). The complex relationship among the emphysema and airway disease axes with death is summarized in Figure 1. The z-axis represents the probability of all-cause mortality ranging from less than 5% to 40% for each decile of loading score for factors 1 and 2 in a Cox proportional hazards model adjusted for age, sex, current smoking, pack-years of smoking, BMI, high blood pressure, each of the 5 factors, the interaction between factors 1 and 2, and a quadratic term for factor 2. As can be seen in Figure 1 in dark blue (the decile of emphysema axis ranging from 1, small loading score, to 10, large loading score), there was not a significant increase in death for low levels of the emphysema disease axis, and even at high levels of the emphysema disease axis, the death rate was not elevated at the lower end of the distribution of the airway disease axis. The airway disease axis was strongly associated with death at all levels of the emphysema axis (decile of airway axis in Figure 1 ranging from 1, small loading score, to 10, large loading score) and with the increase in the death rate being more than a simple linear function. The synergistic interaction between the emphysema and airway disease axes can be seen as the greater mortality rate associated with the higher levels of both (i.e., the progression dark blue to the upper, rear quadrant of the surface plot shown in red in Figure 1. Figure 1. View largeDownload slide The relationship among the emphysema and airway disease axes with death, Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011. The z-axis represents the probability of all-cause mortality, ranging from 4% (dark blue), 5%–10% (purple), 10%–15% (blue), 15%–20% (green), 20%–25% (orange), 25%–30% (yellow), 30%–35% (red), to greater than 35% (dark red) for each decile of loading score for factors 1 (emphysema axis) and 2 (airway axis) in a Cox proportional hazards model adjusted for age, sex, being a current smoker, pack-years of smoking, body mass index (calculated by dividing the weight in kilograms by height in meters squared), high blood pressure, each of the 5 factors, the interaction between factors 1 and 2, and a quadratic term for factor 2. The x- and y-axes represent deciles of each axis, ranging from 1, representing a small loading score, to 10, representing a large loading score. Figure 1. View largeDownload slide The relationship among the emphysema and airway disease axes with death, Genetic Epidemiology of Chronic Obstructive Pulmonary Disease Study, United States, 2008–2011. The z-axis represents the probability of all-cause mortality, ranging from 4% (dark blue), 5%–10% (purple), 10%–15% (blue), 15%–20% (green), 20%–25% (orange), 25%–30% (yellow), 30%–35% (red), to greater than 35% (dark red) for each decile of loading score for factors 1 (emphysema axis) and 2 (airway axis) in a Cox proportional hazards model adjusted for age, sex, being a current smoker, pack-years of smoking, body mass index (calculated by dividing the weight in kilograms by height in meters squared), high blood pressure, each of the 5 factors, the interaction between factors 1 and 2, and a quadratic term for factor 2. The x- and y-axes represent deciles of each axis, ranging from 1, representing a small loading score, to 10, representing a large loading score. DISCUSSION COPD has long been recognized as a heterogeneous disease (4). In recent uses of multidimensional analyses, researchers have characterized this heterogeneity by clustering individuals into discrete phenotypic categories (5). Our approach attempts to identify multidimensional vectors on the basis of combined spirometric and CT data, with each person contributing to each vector depending upon their values for all variables within this multivariable analysis. This provides a continuous distribution for each vector representing an underlying physiologic process, which can be interpreted based on the factor loadings of each individual variable. Conducting factor analysis of subjects in the COPDGene Study revealed 5 unique, multidimensional factors from the correlation structure of pulmonary function and morphologic measures obtained from chest CT imaging. Spirometric measures contributed to 2 of the factors, which were defined on the basis of morphologic measures from the CT images: the emphysema disease axis, characterized by low attenuation areas from inspiration CT, and the airway disease axis, characterized by measures of airway wall thickness. These 2 disease axes were associated with death. Furthermore, a synergistic interaction became apparent such that high levels of both factors were associated with the greatest risk for death. Morphologic measures from chest CT imaging and measures of pulmonary function were included in this analysis. Vector labels from chest CT variables indicate the observed morphologic differences. For example, the emphysema axis represents strong factor loadings of low attenuation area on inspiration CT scans (i.e., less than −950 HU and less than −910HU), and the airway axis represents strong factor loadings of airway thickness (i.e., segmental wall area percentage, square root of the wall area of a derived airway with lumen circumferences of 10 mm and 15 mm). These were labeled as disease axes on the basis of their strong inverse factor loadings of pulmonary function (i.e., FEV1/FVC for the emphysema and airway disease axes and FEV1% predicted for the airway disease axis). Low pulmonary function is a well-established risk factor for death (2), and the results of our study are consistent with that observation. In addition, these analyses partition pulmonary function variables into the proportion associated with an emphysema disease axis and the proportion associated with an airway disease axis. To illustrate, reading Table 2 from left to right, the loading scores reported for the row labeled FEV1 show a negative relationship with the emphysema disease axis in column 1 (loading score = −0.41), a stronger negative relationship with the emphysema disease in column 2 (loading score = −0.89), and a weaker negative relationship with the gas-trapping axis in column 3. To assess the impact of these disease axes on death independent of pulmonary function, FEV1% predicted and the FEV1/FVC ratio were removed from the disease axes and were included directly as independent variables in the Cox proportional hazard model along with the disease axes. The airway disease axis remained independently associated with death rate, whereas the emphysema disease axis was only marginally associated with overall death rate after adjustment for pulmonary function (data not shown). These analyses indicate an important role for pulmonary function on death associated with emphysema as well as an independent role associated with the airway disease axis. Limitations of this study include the use of several measures of CT intensity on inspiration and expiration that have not been directly shown to have clinical relevance. These measures fell into predictable factors, however, which suggests the values chosen (e.g., inspiration percentage of the lung less than −856, less than −910, and less than −950) were correlated measures of a similar disease process with clinical relevance. COPDGene is a large study of current and former smokers with a history of smoking cigarettes for more than 10 pack-years. This likely makes the generalizability of the results questionable when applied to a population with shorter history of cigarette smoking. The COPDGene Study has experienced loss to follow-up over this time, and this can induce bias. Table 1 indicates that all characteristics except race were not statistically different between the main cohort and the cohort used for the factor analysis. Disease axes, compared with disease clusters, may be difficult to interpret in a clinical setting, making the direct applicability of this approach complex for physicians. This approach does avoid the potential misinterpretation of assessing individual variables in the presence of highly correlated data. However, future work should include assessment of inflection points in the risk for all-cause mortality in each of the factors that define high-risk subgroups on the basis of the continuous disease axes. Although clustering approaches have not achieved strong separation for COPD subtypes, inflection points in risk may suggest reasonable clinical cutpoints for individuals. Disease axes may provide important insights into the pathophysiologic processes leading to COPD and death associated with COPD and provide potential targets for intervention for prevention or treatment of COPD. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Colorado School of Public Health, University of Colorado Anschutz Medical Campus, Aurora, Colorado (Gregory L. Kinney, Kendra A. Young, John E. Hokanson); Department of Mathematical and Statistical Sciences, University of Colorado Denver, Denver, Colorado (Stephanie A. Santorico); Human Medical Genetics and Genomics Program, University of Colorado School of Medicine, Aurora, Colorado (Stephanie A. Santorico); Division of Biostatistics and Bioinformatics, Office of Academic Affairs, National Jewish Health, Denver, Colorado (Stephanie A. Santorico, Douglas C. Everett); Channing Division of Network Medicine, Brigham and Women’s Hospital, Boston, Massachusetts (Michael H. Cho, Peter J. Castaldi, Raul San José Estépar, James C. Ross, Edwin K. Silverman, George R. Washko); Department of Electrical & Computer Engineering, Northeastern University, Boston, Massachusetts (Jennifer G. Dy); Department of Medicine, National Jewish Health, Denver, Colorado (Barry J. Make, Elizabeth A. Regan, James D. Crapo); Department of Radiology, National Jewish Health, Denver, Colorado (David A. Lynch); and Department of Biostatistics and Informatics, University of Colorado Anschutz Medical Campus, Aurora, Colorado (Sharon M. Lutz). This project was supported by grants R01 HL089897, R01 HL089856, and K08 HL097029 (to S.M.L.) from the National Heart, Lung, and Blood Institute. The Genetic Epidemiology of COPD Study is also supported by the COPD Foundation through contributions made to an industry advisory board composed of AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Novartis, Pfizer, Siemens, and Sunovion. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute or the National Institutes of Health. Conflict of interest: E.K.S. has received honoraria and consulting fees from Merck, grant support and consulting fees from GlaxoSmithKline, and honoraria from Novartis. These funding sources played no role in the design of the study or the decision to submit the manuscript for publication. The other authors report no conflicts. Abbreviations BMI body mass index COPD chronic obstructive pulmonary disease COPDGene Genetic Epidemiology of Chronic Obstructive Pulmonary Disease CT computed tomography FEV1 forced expiratory volume at 1 second FEV1% percentage of forced expiratory volume at 1 second FVC forced vital capacity FVC% percentage of forced vital capacity HU Hounsfield unit SSDI Social Security Death Index REFERENCES 1 Friedlander AL , Lynch D , Dyar LA , et al. . Phenotypes of chronic obstructive pulmonary disease . COPD . 2007 ; 4 ( 4 ): 355 – 384 . Google Scholar Crossref Search ADS PubMed 2 Silverman EK . Exacerbations in chronic obstructive pulmonary disease: do they contribute to disease progression? Proc Am Thorac Soc . 2007 ; 4 ( 8 ): 586 – 590 . Google Scholar Crossref Search ADS PubMed 3 Løkke A , Lange P , Scharling H , et al. . Developing COPD: a 25 year follow up study of the general population . Thorax . 2006 ; 61 ( 11 ): 935 – 939 . Google Scholar Crossref Search ADS PubMed 4 Sanders KJC , Ash SY , Washko GR , et al. . Imaging approaches to understand disease complexity: chronic obstructive pulmonary disease as a clinical model . J Appl Physiol (1985) . 2018 ; 124 ( 2 ): 512 – 520 . Google Scholar Crossref Search ADS PubMed 5 Castaldi PJ , Benet M , Petersen H , et al. . Do COPD subtypes really exist? COPD heterogeneity and clustering in 10 independent cohorts . Thorax . 2017 ; 72 ( 11 ): 998 – 1006 . Google Scholar Crossref Search ADS PubMed 6 Abdi H , Williams LJ . Principal component analysis . Wiley Interdiscip Rev Comput Stat . 2010 ; 2 ( 4 ): 433 – 459 . Google Scholar Crossref Search ADS 7 Regan EA , Hokanson JE , Murphy JR , et al. . Genetic epidemiology of COPD (COPDGene) study design . COPD . 2010 ; 7 ( 1 ): 32 – 43 . Google Scholar Crossref Search ADS PubMed 8 Miller MR , Hankinson J , Brusasco V , et al. . Standardisation of spirometry . Eur Respir J . 2005 ; 26 ( 2 ): 319 – 338 . Google Scholar Crossref Search ADS PubMed 9 Hankinson JL , Odencrantz JR , Fedan KB , et al. . Spirometric reference values from a sample of the general US population . Am J Respir Crit Care Med . 1999 ; 159 ( 1 ): 179 – 187 . Google Scholar Crossref Search ADS PubMed 10 Han MK , Kazerooni EA , Lynch DA , et al. . Chronic obstructive pulmonary disease exacerbations in the COPDGene study: associated radiologic phenotypes . Radiology . 2011 ; 261 ( 1 ): 274 – 282 . Google Scholar Crossref Search ADS PubMed 11 Diaz AA , Come CE , Ross JC , et al. . Association between airway caliber changes with lung inflation and emphysema assessed by volumetric CT scan in subjects with COPD . Chest . 2012 ; 141 ( 3 ): 736 – 744 . Google Scholar Crossref Search ADS PubMed 12 Kim WJ , Silverman EK , Hoffman E , et al. . CT metrics of airway disease and emphysema in severe COPD . Chest . 2009 ; 136 ( 2 ): 396 – 404 . Google Scholar Crossref Search ADS PubMed 13 Nakano Y , Wong JC , de Jong PA , et al. . The prediction of small airway dimensions using computed tomography . Am J Respir Crit Care Med . 2005 ; 171 ( 2 ): 142 – 146 . Google Scholar Crossref Search ADS PubMed 14 Xu J , Murphy SL , Kochanek KD , et al. . Deaths: final data for 2013 . Natl Vital Stat Rep . 2016 ; 64 ( 2 ): 1 – 119 . Google Scholar PubMed 15 Horn JL . A rationale and test for the number of factors in factor analysis . Psychometrika . 1965 ; 30 ( 2 ): 179 – 185 . Google Scholar Crossref Search ADS PubMed 16 Stewart JI , Moyle S , Criner GJ , et al. . Automated telecommunication to obtain longitudinal follow-up in a multicenter cross-sectional COPD study . COPD . 2012 ; 9 ( 5 ): 466 – 472 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Xiao, Xiling; Zhang, Zhe; Chang, Ellen T; Liu, Zhiwei; Liu, Qing; Cai, Yonglin; Chen, Guomin; Huang, Qi-Hong; Xie, Shang-Hang; Cao, Su-Mei; Shao, Jian-Yong; Jia, Wei-Hua; Zheng, Yuming; Liao, Jian; Chen, Yufeng; Lin, Longde; Ernberg, Ingemar; Huang, Guangwu; Zeng, Yi; Zeng, Yi-Xin; Adami, Hans-Olov; Ye, Weimin

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy095pmid: 29701753

Abstract Because persistent inflammation may render the nasopharyngeal mucosa susceptible to carcinogenesis, chronic ear-nose-throat (ENT) disease and its treatment might influence the risk of nasopharyngeal carcinoma (NPC). Existing evidence is, however, inconclusive and often based on methodologically suboptimal epidemiologic studies. In a population-based case-control study in southern China, we enrolled 2,532 persons with NPC and 2,597 controls, aged 20–74 years, from 2010 to 2014. Odds ratios were estimated for associations between NPC risk and history of ENT and related medications. Any history of chronic ENT disease was associated with a 34% increased risk of NPC. Similarly, use of nasal drops or aspirin was associated with approximately doubled risk of NPC. However, in secondary analyses restricted to chronic ENT diseases and related medication use at least 5 years prior to diagnosis/interview, most results were statistically nonsignificant, except a history of uncured ENT diseases, untreated nasal polyps, and earlier age at first diagnosis of ENT disease and first or most recent aspirin use. Overall, these findings suggest that ENT disease and related medication use are most likely early indications rather than causes of NPC, although the possibility of a modestly increased NPC risk associated with these diseases and related medications cannot be excluded. case-control study, medical history, medication use, nasopharyngeal carcinoma Nasopharyngeal carcinoma (NPC) has a remarkable geographical and racial/ethnic distribution, and it is endemic in parts of southern China. Numerous studies have suggested that NPC develops through interactions among genetic and environmental factors and Epstein-Barr virus (EBV) infection. The etiologic role of EBV in NPC is supported by molecular analyses, as well as by serological studies (1, 2). Assumed environmental risk factors for NPC include salted fish consumption, smoking, and poor oral health (3, 4), while genetic studies have consistently shown that certain human leukocyte antigens influence NPC risk (5). Inflammation has been proposed as one of the hallmarks of cancer (6), and persistent inflammation and infection of the respiratory tract in particular may render the nasopharyngeal mucosa more susceptible to carcinogenesis. Individual history of ear-nose-throat (ENT) disease, including sinusitis and otitis media, has been shown to increase NPC risk (7–11). Modern medicines, such as aspirin and nasal drops, as well as traditional herbal medicines, such as balms and essential oils, are commonly used in southern China to relieve symptoms of headache and nasal obstruction. Nasal balms or oils may increase NPC risk (12), whereas an inverse association between aspirin use and NPC risk has been reported (13). However, published studies of the use of these medications were either small in size or hospital-based, and observed associations may be due to reverse causality, recall bias, or confounding by underlying ENT disease. To fill the existing knowledge gap regarding the possible relationship of medical history and medication use with NPC risk, we conducted a large, population-based case-control study in southern China, where NPC is endemic. This study enabled a more rigorous investigation of the potential independent etiologic roles of ENT-related medical history and medication use in NPC development. METHODS Study population The NPC Genes, Environment, and EBV study (NPCGEE) is a collaborative population-based case-control study of NPC based in the Zhaoqing area of Guangdong Province and in the Wuzhou and Guiping/Pingnan areas of Guangxi Autonomous Region. These areas encompass 13 cities/counties (Deqing, Fengkai, Gaoyao, Huaiji, Sihui, Zhaoqing, Guangning, Wuzhou, Cenxi, Cangwu, Tengxian, Pingnan, and Guiping) in southern China, with a total population of approximately 8 million. Recruitment of cases and controls was described previously (14). In brief, cases were aged 20–74 years at diagnosis between March 2010 and December 2013, living in the described geographic area, and without a prior history of malignant disease or congenital or acquired immunodeficiency. All cases were histopathologically confirmed by pathology reports. We established a rapid case-ascertainment system including 10 hospitals and 2 cancer research institutions that directly notified study investigators of newly diagnosed NPC cases. In the Zhaoqing area, 1,528 eligible cases were identified between March 2010 and August 2013; in the Wuzhou area, 792 eligible cases were identified between April 2010 and September 2013; in the Guiping/Pingnan area, 727 eligible cases were identified between July 2010 and December 2013. The number of cases identified in each region was close to the expected number of incident NPC cases based on historical incidence rates. Of eligible patients who were contacted by study staff, 1,306 (85% of 1,528 cases) in the Zhaoqing area, 689 (87% of 792 cases) in the Wuzhou area, and 559 (77% of 727 cases) in the Guiping/Pingnan area were enrolled in the study. Controls were randomly selected every 6–12 months between November 2010 and November 2014 from continuously updated total population registries covering the Zhaoqing, Wuzhou, and Guiping/Pingnan populations, with frequency matching to the 5-year-age and sex distribution of the cases according to geographic region. Eligible controls were required to be residents of the study area without a prior history of malignant disease or congenital or acquired immune deficiency. Controls who had worked outside of the study area for 10 years or more, as identified with the help of the local government in each town or community, were not considered part of the study base, because they were highly unlikely to return to the study area if they were diagnosed with NPC; therefore, they were replaced. Of 3,932 potential controls selected from the total population registries, 730 (19%) could not be linked to an identifiable person because of invalid contact information. Of the 3,202 who were identified, 138 (4%) had emigrated out of the study area, 90 (3%) were deceased or incapacitated, and 326 (10%) refused to participate. Of the 2,648 (83% of 3,202) enrolled controls, 2,133 (81%) were initial selections from the population registry, and the other 515 (19%) were replacements. This study was approved by the institutional review boards of Sun Yat-sen University Cancer Center, the Institute for Viral Disease Control and Prevention of the Chinese Center for Disease Control and Prevention, Guangxi Medical University, and Harvard T.H. Chan School of Public Health, as well as the Regional Research Ethics Vetting Board in Stockholm, Sweden. All subjects granted written or oral informed consent to participate. Data collection Trained interviewers used a structured electronic questionnaire to conduct audiotaped face-to-face or telephone interviews with study participants. Although blinding to case-control status was not feasible, in an effort to reduce information bias, we required all interviewers, who were unaware of the study hypotheses, to interview an approximately equal number of cases and controls. The questionnaire covered demographic characteristics, residential history, occupational history, medical history of chronic ENT disease, medication use, family medical history, dietary habits, cigarette smoking, alcohol and tea drinking, and use of Chinese herbal medicine, among other topics. Questionnaire data were automatically flagged for logic errors and missing values, and errors were corrected by making comparisons against audio recordings or by contacting participants again. Five chronic ENT diseases were investigated: chronic sinusitis, chronic pharyngitis, chronic otitis media, nasal polyps, and septal abnormalities. For each disease, questions addressed whether the subject had ever been diagnosed (yes or no), age at diagnosis, number of times of diagnosed, age at most recent occurrence, any use of medications or surgical treatment, and whether the disease had been cured (i.e., was no longer symptomatic). Persistent, chronic ENT disease was classified based on having an interval of 3 months or less between multiple episodes (15). Medications assessed were nasal drops (use for at least 3 months in 1 year), aspirin (use for at least 3 months in 1 year), balm or peppermint (any use), flower oil (any use), and qufeng oil (any use). Qufeng oil, a traditional Chinese medicine, consists mainly of methyl salicylate, peppermint oil, and camphor oil, and is used topically on the skin to treat arthralgia. For nasal drops and aspirin, additional questions included reasons for use (free text), ages at initiation and cessation of use, frequency of use (times per day, week, or month), and duration of use (years). After data cleaning, 1 case and 17 controls had misplaced data, and 6 controls outside the eligible range at interview were excluded. We further excluded 1 case with missing data on medical history and medication use and 20 cases and 28 controls with poor-quality questionnaire data as determined by the interviewers. After excluding these 73 subjects, 2,532 cases and 2,597 controls were included in the present analysis. Statistical analysis We used multivariate unconditional logistic regression models to estimate odds ratios and corresponding 95% confidence intervals for risk of NPC associated with history of chronic ENT diseases and medication use. In secondary analyses, to reduce the potential for reverse causality, chronic ENT diseases and use of related medications within the 5 years prior to diagnosis (for cases) or interview (for controls) were excluded. Based on prior knowledge and analyses in this study population (3, 4), potential confounders included in the multivariate models were age (in 5-year groups), sex, residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (concrete building, clay brick cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current, former, or never), first-degree family history of NPC (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use (never, yearly, monthly, or weekly or more). Crude differences between NPC cases and controls were compared using the χ2 test for categorical variables and Student’s t test for continuous variables. Where continuous variables were classified into categories, the median value was used as the cutpoint. Data analyses were performed with SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina). All statistical tests were 2-sided, and a P value of <0.05 was considered statistically significant. RESULTS Descriptive characteristics Table 1 shows the distribution of demographic and other characteristics among the 2,532 NPC cases and 2,597 population-based controls. Because we began interviewing controls about 1 year later than cases, cases were slightly younger than controls. Cases were less educated and more likely to live in cottages, have blue-collar jobs, have a first-degree family history of NPC, have consumed more salt-preserved fish in adulthood, and have used herbal medicines at least weekly, compared with controls. Table 1. Characteristics of Nasopharyngeal Carcinoma Cases and Controls Enrolled in a Population-Based Case-Control Study in Southern China, 2010–2014 Characteristic . Case Group (n = 2,532) . Control Group (n = 2,597) . P Value . No. . % . No. . % . Residential area 0.3  Zhaoqing 1,286 50.8 1,321 50.9  Wuzhou 688 27.2 665 25.6  Guiping/Pingnan 558 22.0 611 23.5 Sex 0.97  Male 1,860 73.5 1,909 73.5  Female 672 26.5 688 26.5 Age at diagnosis/interview, years  Overalla 48.54 (10.70) 49.75 (10.90) <0.001b  20–34 232 9.2 207 8.1 0.07  35–39 277 11.0 246 9.5  40–44 418 16.5 399 15.4  45–49 492 19.4 491 18.9  50–54 333 13.2 347 13.4  55–59 353 13.9 382 14.7  60–74 427 16.9 525 20.2 Educational level, years 0.004  ≤6 1,005 39.7 932 35.9  7–9 1,013 40.0 1,040 40.1  10–12 407 16.1 484 18.6  >12 107 4.2 141 5.4 Current housing type <0.001  Building (concrete structure) 1,820 71.9 2,019 77.7  Cottage (clay brick structure) 702 27.7 575 22.1  Boat 10 0.4 2 0.1  Missing 0 0 1 0.04 Current occupation <0.001  Unemployed 78 3.1 96 3.7  Farmer 855 33.8 984 37.9  Blue-collar 1,023 40.4 900 34.7  White-collar 350 13.8 416 16.0  Other/unknown 226 8.9 201 7.7 Cigarette smoking 0.12  Current smoker 1,121 44.3 1,213 46.7  Former smoker 182 7.2 152 5.9  Never smoker 1,228 48.5 1,230 47.4  Missing 1 0.04 2 0.1 First-degree family history of NPC <0.001  No 2,208 87.2 2,483 95.6  Yes 272 10.7 70 2.7  Unknown 46 1.8 43 1.7  Missing 6 0.2 1 0.04 Alcohol drinking 0.11  Never 1,735 68.5 1,815 69.9  Ever 791 31.2 766 29.5  Missing 6 0.2 16 0.6 Tea drinking <0.001  Never 1,618 63.8 1,513 58.3  Ever 911 36.0 1,081 41.6  Missing 3 0.1 3 0.1 Salt-preserved fish consumption in adult diet, categoryc <0.001  1 846 33.4 797 30.7  2 489 19.3 583 22.5  3 488 19.3 582 22.4  4 678 26.8 596 23.0  Missing 31 1.2 39 1.5 Herbal medicine use 0.03  Nonusers 355 14.0 419 16.1  Yearly 965 38.1 974 37.5  Monthly 905 35.7 942 36.3  Weekly or more often 269 10.6 220 8.5  Missing 38 1.5 42 1.6 Characteristic . Case Group (n = 2,532) . Control Group (n = 2,597) . P Value . No. . % . No. . % . Residential area 0.3  Zhaoqing 1,286 50.8 1,321 50.9  Wuzhou 688 27.2 665 25.6  Guiping/Pingnan 558 22.0 611 23.5 Sex 0.97  Male 1,860 73.5 1,909 73.5  Female 672 26.5 688 26.5 Age at diagnosis/interview, years  Overalla 48.54 (10.70) 49.75 (10.90) <0.001b  20–34 232 9.2 207 8.1 0.07  35–39 277 11.0 246 9.5  40–44 418 16.5 399 15.4  45–49 492 19.4 491 18.9  50–54 333 13.2 347 13.4  55–59 353 13.9 382 14.7  60–74 427 16.9 525 20.2 Educational level, years 0.004  ≤6 1,005 39.7 932 35.9  7–9 1,013 40.0 1,040 40.1  10–12 407 16.1 484 18.6  >12 107 4.2 141 5.4 Current housing type <0.001  Building (concrete structure) 1,820 71.9 2,019 77.7  Cottage (clay brick structure) 702 27.7 575 22.1  Boat 10 0.4 2 0.1  Missing 0 0 1 0.04 Current occupation <0.001  Unemployed 78 3.1 96 3.7  Farmer 855 33.8 984 37.9  Blue-collar 1,023 40.4 900 34.7  White-collar 350 13.8 416 16.0  Other/unknown 226 8.9 201 7.7 Cigarette smoking 0.12  Current smoker 1,121 44.3 1,213 46.7  Former smoker 182 7.2 152 5.9  Never smoker 1,228 48.5 1,230 47.4  Missing 1 0.04 2 0.1 First-degree family history of NPC <0.001  No 2,208 87.2 2,483 95.6  Yes 272 10.7 70 2.7  Unknown 46 1.8 43 1.7  Missing 6 0.2 1 0.04 Alcohol drinking 0.11  Never 1,735 68.5 1,815 69.9  Ever 791 31.2 766 29.5  Missing 6 0.2 16 0.6 Tea drinking <0.001  Never 1,618 63.8 1,513 58.3  Ever 911 36.0 1,081 41.6  Missing 3 0.1 3 0.1 Salt-preserved fish consumption in adult diet, categoryc <0.001  1 846 33.4 797 30.7  2 489 19.3 583 22.5  3 488 19.3 582 22.4  4 678 26.8 596 23.0  Missing 31 1.2 39 1.5 Herbal medicine use 0.03  Nonusers 355 14.0 419 16.1  Yearly 965 38.1 974 37.5  Monthly 905 35.7 942 36.3  Weekly or more often 269 10.6 220 8.5  Missing 38 1.5 42 1.6 Abbreviation: NPC, nasopharyngeal carcinoma. a Values are expressed as mean (standard deviation). bP value was determined by a 2-sided t test. Other P values were determined by a χ2 test. c Energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake. Open in new tab Table 1. Characteristics of Nasopharyngeal Carcinoma Cases and Controls Enrolled in a Population-Based Case-Control Study in Southern China, 2010–2014 Characteristic . Case Group (n = 2,532) . Control Group (n = 2,597) . P Value . No. . % . No. . % . Residential area 0.3  Zhaoqing 1,286 50.8 1,321 50.9  Wuzhou 688 27.2 665 25.6  Guiping/Pingnan 558 22.0 611 23.5 Sex 0.97  Male 1,860 73.5 1,909 73.5  Female 672 26.5 688 26.5 Age at diagnosis/interview, years  Overalla 48.54 (10.70) 49.75 (10.90) <0.001b  20–34 232 9.2 207 8.1 0.07  35–39 277 11.0 246 9.5  40–44 418 16.5 399 15.4  45–49 492 19.4 491 18.9  50–54 333 13.2 347 13.4  55–59 353 13.9 382 14.7  60–74 427 16.9 525 20.2 Educational level, years 0.004  ≤6 1,005 39.7 932 35.9  7–9 1,013 40.0 1,040 40.1  10–12 407 16.1 484 18.6  >12 107 4.2 141 5.4 Current housing type <0.001  Building (concrete structure) 1,820 71.9 2,019 77.7  Cottage (clay brick structure) 702 27.7 575 22.1  Boat 10 0.4 2 0.1  Missing 0 0 1 0.04 Current occupation <0.001  Unemployed 78 3.1 96 3.7  Farmer 855 33.8 984 37.9  Blue-collar 1,023 40.4 900 34.7  White-collar 350 13.8 416 16.0  Other/unknown 226 8.9 201 7.7 Cigarette smoking 0.12  Current smoker 1,121 44.3 1,213 46.7  Former smoker 182 7.2 152 5.9  Never smoker 1,228 48.5 1,230 47.4  Missing 1 0.04 2 0.1 First-degree family history of NPC <0.001  No 2,208 87.2 2,483 95.6  Yes 272 10.7 70 2.7  Unknown 46 1.8 43 1.7  Missing 6 0.2 1 0.04 Alcohol drinking 0.11  Never 1,735 68.5 1,815 69.9  Ever 791 31.2 766 29.5  Missing 6 0.2 16 0.6 Tea drinking <0.001  Never 1,618 63.8 1,513 58.3  Ever 911 36.0 1,081 41.6  Missing 3 0.1 3 0.1 Salt-preserved fish consumption in adult diet, categoryc <0.001  1 846 33.4 797 30.7  2 489 19.3 583 22.5  3 488 19.3 582 22.4  4 678 26.8 596 23.0  Missing 31 1.2 39 1.5 Herbal medicine use 0.03  Nonusers 355 14.0 419 16.1  Yearly 965 38.1 974 37.5  Monthly 905 35.7 942 36.3  Weekly or more often 269 10.6 220 8.5  Missing 38 1.5 42 1.6 Characteristic . Case Group (n = 2,532) . Control Group (n = 2,597) . P Value . No. . % . No. . % . Residential area 0.3  Zhaoqing 1,286 50.8 1,321 50.9  Wuzhou 688 27.2 665 25.6  Guiping/Pingnan 558 22.0 611 23.5 Sex 0.97  Male 1,860 73.5 1,909 73.5  Female 672 26.5 688 26.5 Age at diagnosis/interview, years  Overalla 48.54 (10.70) 49.75 (10.90) <0.001b  20–34 232 9.2 207 8.1 0.07  35–39 277 11.0 246 9.5  40–44 418 16.5 399 15.4  45–49 492 19.4 491 18.9  50–54 333 13.2 347 13.4  55–59 353 13.9 382 14.7  60–74 427 16.9 525 20.2 Educational level, years 0.004  ≤6 1,005 39.7 932 35.9  7–9 1,013 40.0 1,040 40.1  10–12 407 16.1 484 18.6  >12 107 4.2 141 5.4 Current housing type <0.001  Building (concrete structure) 1,820 71.9 2,019 77.7  Cottage (clay brick structure) 702 27.7 575 22.1  Boat 10 0.4 2 0.1  Missing 0 0 1 0.04 Current occupation <0.001  Unemployed 78 3.1 96 3.7  Farmer 855 33.8 984 37.9  Blue-collar 1,023 40.4 900 34.7  White-collar 350 13.8 416 16.0  Other/unknown 226 8.9 201 7.7 Cigarette smoking 0.12  Current smoker 1,121 44.3 1,213 46.7  Former smoker 182 7.2 152 5.9  Never smoker 1,228 48.5 1,230 47.4  Missing 1 0.04 2 0.1 First-degree family history of NPC <0.001  No 2,208 87.2 2,483 95.6  Yes 272 10.7 70 2.7  Unknown 46 1.8 43 1.7  Missing 6 0.2 1 0.04 Alcohol drinking 0.11  Never 1,735 68.5 1,815 69.9  Ever 791 31.2 766 29.5  Missing 6 0.2 16 0.6 Tea drinking <0.001  Never 1,618 63.8 1,513 58.3  Ever 911 36.0 1,081 41.6  Missing 3 0.1 3 0.1 Salt-preserved fish consumption in adult diet, categoryc <0.001  1 846 33.4 797 30.7  2 489 19.3 583 22.5  3 488 19.3 582 22.4  4 678 26.8 596 23.0  Missing 31 1.2 39 1.5 Herbal medicine use 0.03  Nonusers 355 14.0 419 16.1  Yearly 965 38.1 974 37.5  Monthly 905 35.7 942 36.3  Weekly or more often 269 10.6 220 8.5  Missing 38 1.5 42 1.6 Abbreviation: NPC, nasopharyngeal carcinoma. a Values are expressed as mean (standard deviation). bP value was determined by a 2-sided t test. Other P values were determined by a χ2 test. c Energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake. Open in new tab ENT diseases Table 2 presents adjusted odds ratios for the association between chronic ENT disease and risk of NPC. Any history of chronic ENT disease was associated with a 34% (95% confidence interval (CI): 12, 59) higher risk of NPC. Having been treated for chronic ENT disease was associated with a similar magnitude of increase in NPC risk. Having been cured of chronic ENT disease was, however, not significantly associated with NPC risk, whereas uncured ENT disease conferred an increased risk (odds ratio (OR) = 1.53, 95% CI: 1.25, 1.89). After exclusion of individuals first diagnosed with ENT disease within the past 5 years, however, odds ratios were attenuated toward (or even past) the null, and none remained statistically significant, except a history of uncured ENT disease (OR = 1.28, 95% CI: 1.00, 1.64) and initial diagnosis of ENT disease at a young age (OR = 1.35, 95% CI: 1.04, 1.76). Table 2. Odds Ratios for Nasopharyngeal Carcinoma Associated With Chronic Ear-Nose-Throat Diseases in Southern China, 2010–2014 Variable . Any History of Chronic ENT Disease . ENT Disease Diagnosed More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,413) . No. of Controls (n = 2,533) . ORa . 95% CIa . Chronic ENT diseases  No 2,161 2,304 1.00 Referent 2,161 2,304 1.00 Referent  Yes 371 293 1.34 1.12, 1.59 240 223 1.16 0.94, 1.42   Untreated 61 52 1.29 0.87, 1.92 36 42 0.96 0.60, 1.54   Some diseases treated 310 241 1.35 1.12, 1.63 216 187 1.23 0.99, 1.53   Uncured 260 183 1.53 1.25, 1.89 167 141 1.28 1.00, 1.64   Some diseases cured 111 110 1.01 0.76, 1.35 85 88 1.02 0.73, 1.40  Missing 0 0 12 6  Age at first diagnosis, years   1–30 181 140 1.39 1.09, 1.77 151 118 1.35 1.04, 1.76   31–74 178 147 1.26 1.00, 1.60 89 105 0.94 0.69, 1.27   Unknown 12 6 12 6 Variable . Any History of Chronic ENT Disease . ENT Disease Diagnosed More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,413) . No. of Controls (n = 2,533) . ORa . 95% CIa . Chronic ENT diseases  No 2,161 2,304 1.00 Referent 2,161 2,304 1.00 Referent  Yes 371 293 1.34 1.12, 1.59 240 223 1.16 0.94, 1.42   Untreated 61 52 1.29 0.87, 1.92 36 42 0.96 0.60, 1.54   Some diseases treated 310 241 1.35 1.12, 1.63 216 187 1.23 0.99, 1.53   Uncured 260 183 1.53 1.25, 1.89 167 141 1.28 1.00, 1.64   Some diseases cured 111 110 1.01 0.76, 1.35 85 88 1.02 0.73, 1.40  Missing 0 0 12 6  Age at first diagnosis, years   1–30 181 140 1.39 1.09, 1.77 151 118 1.35 1.04, 1.76   31–74 178 147 1.26 1.00, 1.60 89 105 0.94 0.69, 1.27   Unknown 12 6 12 6 Abbreviations: CI, confidence interval; ENT, ear-nose-throat; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab Table 2. Odds Ratios for Nasopharyngeal Carcinoma Associated With Chronic Ear-Nose-Throat Diseases in Southern China, 2010–2014 Variable . Any History of Chronic ENT Disease . ENT Disease Diagnosed More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,413) . No. of Controls (n = 2,533) . ORa . 95% CIa . Chronic ENT diseases  No 2,161 2,304 1.00 Referent 2,161 2,304 1.00 Referent  Yes 371 293 1.34 1.12, 1.59 240 223 1.16 0.94, 1.42   Untreated 61 52 1.29 0.87, 1.92 36 42 0.96 0.60, 1.54   Some diseases treated 310 241 1.35 1.12, 1.63 216 187 1.23 0.99, 1.53   Uncured 260 183 1.53 1.25, 1.89 167 141 1.28 1.00, 1.64   Some diseases cured 111 110 1.01 0.76, 1.35 85 88 1.02 0.73, 1.40  Missing 0 0 12 6  Age at first diagnosis, years   1–30 181 140 1.39 1.09, 1.77 151 118 1.35 1.04, 1.76   31–74 178 147 1.26 1.00, 1.60 89 105 0.94 0.69, 1.27   Unknown 12 6 12 6 Variable . Any History of Chronic ENT Disease . ENT Disease Diagnosed More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,413) . No. of Controls (n = 2,533) . ORa . 95% CIa . Chronic ENT diseases  No 2,161 2,304 1.00 Referent 2,161 2,304 1.00 Referent  Yes 371 293 1.34 1.12, 1.59 240 223 1.16 0.94, 1.42   Untreated 61 52 1.29 0.87, 1.92 36 42 0.96 0.60, 1.54   Some diseases treated 310 241 1.35 1.12, 1.63 216 187 1.23 0.99, 1.53   Uncured 260 183 1.53 1.25, 1.89 167 141 1.28 1.00, 1.64   Some diseases cured 111 110 1.01 0.76, 1.35 85 88 1.02 0.73, 1.40  Missing 0 0 12 6  Age at first diagnosis, years   1–30 181 140 1.39 1.09, 1.77 151 118 1.35 1.04, 1.76   31–74 178 147 1.26 1.00, 1.60 89 105 0.94 0.69, 1.27   Unknown 12 6 12 6 Abbreviations: CI, confidence interval; ENT, ear-nose-throat; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab Focusing on each chronic ENT disease separately, we found that a history of chronic sinusitis was associated with a 1.3-fold increased risk of NPC, as was treated chronic sinusitis. While cured chronic sinusitis was not significantly associated with NPC risk, uncured chronic sinusitis was positively associated with a close to 50% increased risk of NPC. After restriction of the analysis to those first diagnosed at least 5 years earlier, the associations were generally attenuated and were statistically nonsignificant (Web Table 1, available at https://academic.oup.com/aje). Chronic otitis media was associated with an almost doubled risk of NPC, and the results were similar for treated and uncured chronic otitis media (Web Table 2). Older age at first diagnosis, older age at the most recent occurrence, and persistent chronic otitis media were associated with increased risk of NPC. When we excluded individuals who were first diagnosed within the past 5 years, nearly all estimates were attenuated, and only older age at the most recent diagnosis of chronic otitis media (i.e., having had a more recent occurrence) remained significant (OR = 2.05, 95% CI: 1.06, 3.97). A history of nasal polyps was not significantly associated with NPC risk, but untreated and uncured nasal polyps were associated with significantly greater risk (Web Table 3). When we excluded individuals who were first diagnosed with nasal polyps within the past 5 years, all associations were attenuated and most were statistically nonsignificant, except for untreated nasal polyps (OR = 3.26, 95% CI: 1.14, 9.34). No associations were found between NPC risk and chronic pharyngitis and septal abnormalities, in the primary or secondary analyses (Web Tables 4 and 5). Medications Table 3 shows the adjusted odds ratios for associations between use of nasal drops and risk of NPC. Any history of use of nasal drops was associated with an almost doubled risk of NPC. Use of nasal drops for nasal obstruction demonstrated a significantly increased risk for NPC (OR = 2.82), whereas use of nasal drops for allergic rhinitis did not (OR = 1.23). Older age at initiation and more recent use of nasal drops were more strongly associated with NPC risk than earlier and more distant past use. Shorter duration of nasal drops use (1–6 years), was more strongly associated with NPC risk than longer duration of use (7–40 years), but more frequent use (at least daily) was more strongly associated with risk than less frequent use. However, when individuals who started to use nasal drops within the past 5 years were excluded from the analysis, associations with nasal drops use were weakened, and all were statistically nonsignificant. Table 3. Odds Ratios for Nasopharyngeal Carcinoma Associated With Nasal Drops Use in Southern China, 2010–2014 Variable . Any History of Nasal Drops . First Use of Nasal Drops More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,497) . No. of Controls (n = 2,590) . ORa . 95% CIa . History of nasal drops use (at least 3 months in 1 year)  No 2,458 2,558 1.00 Referent 2,458 2,558 1.00 Referent  Yes 74 39 1.98 1.31, 3.01 39 32 1.25 0.75, 2.07 Reason for use of nasal drops  Nasal obstruction 51 20 2.82 1.63, 4.88 25 15 1.80 0.90, 3.59  Allergic rhinitis 13 10 1.23 0.50, 3.00 12 9 1.39 0.54, 3.55  Other 10 9 1.06 0.41, 2.73 2 8 0.24 0.05, 1.15 Age at first use of nasal drops, years  1–34 32 20 1.69 0.92, 3.09 20 17 1.30 0.64, 2.65  35–64 42 19 2.28 1.29, 4.02 19 15 1.20 0.58, 2.44 Age at most recent use of nasal drops, years  1–42 32 19 1.68 0.92, 3.07 21 16 1.34 0.67, 2.68  43–74 42 20 2.29 1.30, 4.05 18 16 1.16 0.56, 2.41 Duration of nasal drops use, years  1–6 52 21 2.55 1.51, 4.31 21 16 1.31 0.67, 2.57  7–40 22 18 1.23 0.61, 2.48 18 16 1.17 0.55, 2.51 No. of uses of nasal drops  ≥1/day 39 15 2.51 1.35, 4.65 21 13 1.48 0.72, 3.06  ≥1/week 15 12 1.38 0.60, 3.17 7 11 0.70 0.25, 2.02  ≥1/month 19 12 1.84 0.83, 3.92 11 8 1.54 0.57, 4.16  Missing 1 0 0 0 Variable . Any History of Nasal Drops . First Use of Nasal Drops More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,497) . No. of Controls (n = 2,590) . ORa . 95% CIa . History of nasal drops use (at least 3 months in 1 year)  No 2,458 2,558 1.00 Referent 2,458 2,558 1.00 Referent  Yes 74 39 1.98 1.31, 3.01 39 32 1.25 0.75, 2.07 Reason for use of nasal drops  Nasal obstruction 51 20 2.82 1.63, 4.88 25 15 1.80 0.90, 3.59  Allergic rhinitis 13 10 1.23 0.50, 3.00 12 9 1.39 0.54, 3.55  Other 10 9 1.06 0.41, 2.73 2 8 0.24 0.05, 1.15 Age at first use of nasal drops, years  1–34 32 20 1.69 0.92, 3.09 20 17 1.30 0.64, 2.65  35–64 42 19 2.28 1.29, 4.02 19 15 1.20 0.58, 2.44 Age at most recent use of nasal drops, years  1–42 32 19 1.68 0.92, 3.07 21 16 1.34 0.67, 2.68  43–74 42 20 2.29 1.30, 4.05 18 16 1.16 0.56, 2.41 Duration of nasal drops use, years  1–6 52 21 2.55 1.51, 4.31 21 16 1.31 0.67, 2.57  7–40 22 18 1.23 0.61, 2.48 18 16 1.17 0.55, 2.51 No. of uses of nasal drops  ≥1/day 39 15 2.51 1.35, 4.65 21 13 1.48 0.72, 3.06  ≥1/week 15 12 1.38 0.60, 3.17 7 11 0.70 0.25, 2.02  ≥1/month 19 12 1.84 0.83, 3.92 11 8 1.54 0.57, 4.16  Missing 1 0 0 0 Abbreviations: CI, confidence interval; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab Table 3. Odds Ratios for Nasopharyngeal Carcinoma Associated With Nasal Drops Use in Southern China, 2010–2014 Variable . Any History of Nasal Drops . First Use of Nasal Drops More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,497) . No. of Controls (n = 2,590) . ORa . 95% CIa . History of nasal drops use (at least 3 months in 1 year)  No 2,458 2,558 1.00 Referent 2,458 2,558 1.00 Referent  Yes 74 39 1.98 1.31, 3.01 39 32 1.25 0.75, 2.07 Reason for use of nasal drops  Nasal obstruction 51 20 2.82 1.63, 4.88 25 15 1.80 0.90, 3.59  Allergic rhinitis 13 10 1.23 0.50, 3.00 12 9 1.39 0.54, 3.55  Other 10 9 1.06 0.41, 2.73 2 8 0.24 0.05, 1.15 Age at first use of nasal drops, years  1–34 32 20 1.69 0.92, 3.09 20 17 1.30 0.64, 2.65  35–64 42 19 2.28 1.29, 4.02 19 15 1.20 0.58, 2.44 Age at most recent use of nasal drops, years  1–42 32 19 1.68 0.92, 3.07 21 16 1.34 0.67, 2.68  43–74 42 20 2.29 1.30, 4.05 18 16 1.16 0.56, 2.41 Duration of nasal drops use, years  1–6 52 21 2.55 1.51, 4.31 21 16 1.31 0.67, 2.57  7–40 22 18 1.23 0.61, 2.48 18 16 1.17 0.55, 2.51 No. of uses of nasal drops  ≥1/day 39 15 2.51 1.35, 4.65 21 13 1.48 0.72, 3.06  ≥1/week 15 12 1.38 0.60, 3.17 7 11 0.70 0.25, 2.02  ≥1/month 19 12 1.84 0.83, 3.92 11 8 1.54 0.57, 4.16  Missing 1 0 0 0 Variable . Any History of Nasal Drops . First Use of Nasal Drops More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,497) . No. of Controls (n = 2,590) . ORa . 95% CIa . History of nasal drops use (at least 3 months in 1 year)  No 2,458 2,558 1.00 Referent 2,458 2,558 1.00 Referent  Yes 74 39 1.98 1.31, 3.01 39 32 1.25 0.75, 2.07 Reason for use of nasal drops  Nasal obstruction 51 20 2.82 1.63, 4.88 25 15 1.80 0.90, 3.59  Allergic rhinitis 13 10 1.23 0.50, 3.00 12 9 1.39 0.54, 3.55  Other 10 9 1.06 0.41, 2.73 2 8 0.24 0.05, 1.15 Age at first use of nasal drops, years  1–34 32 20 1.69 0.92, 3.09 20 17 1.30 0.64, 2.65  35–64 42 19 2.28 1.29, 4.02 19 15 1.20 0.58, 2.44 Age at most recent use of nasal drops, years  1–42 32 19 1.68 0.92, 3.07 21 16 1.34 0.67, 2.68  43–74 42 20 2.29 1.30, 4.05 18 16 1.16 0.56, 2.41 Duration of nasal drops use, years  1–6 52 21 2.55 1.51, 4.31 21 16 1.31 0.67, 2.57  7–40 22 18 1.23 0.61, 2.48 18 16 1.17 0.55, 2.51 No. of uses of nasal drops  ≥1/day 39 15 2.51 1.35, 4.65 21 13 1.48 0.72, 3.06  ≥1/week 15 12 1.38 0.60, 3.17 7 11 0.70 0.25, 2.02  ≥1/month 19 12 1.84 0.83, 3.92 11 8 1.54 0.57, 4.16  Missing 1 0 0 0 Abbreviations: CI, confidence interval; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab As shown in Table 4, aspirin use was also associated with an almost doubled risk of NPC. The major reasons for use of aspirin were arthritis and myocardial infarction, neither of which was significantly associated with NPC risk, whereas unspecified “other” reasons for use were associated with significantly greater risk (OR = 2.01). Relative risks did not vary appreciably by age at initiating aspirin, but more distant past use was more strongly associated with NPC risk than more recent use (for last use before age 50 years, OR = 2.58, 95% CI: 1.86, 3.58). Shorter duration of aspirin use (1–5 years) was more strongly associated with NPC risk than longer duration (6–51 years), but more frequent use (at least daily) was associated with greater risk than less frequent use. After restriction of the analysis to those who started using aspirin at least 5 years earlier, aspirin use overall was not significantly associated with NPC risk, but earlier age at initiation and most recent use remained significantly associated with greater NPC risk, although odds ratios were attenuated toward the null (for both, OR = 1.68). Table 4. Odds Ratios for Nasopharyngeal Carcinoma Associated With Aspirin Use in Southern China, 2010–2014 Variable . Aspirin Use . First Use of Aspirin More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,390) . No. of Controls (n = 2,551) . ORa . 95% CIa . History of aspirin use (at least 3 months in 1 year)  No 2,291 2,468 1.00 Referent 2,291 2,468 1.00 Referent  Yes 241 129 1.91 1.52, 2.41 99 83 1.21 0.89, 1.66 Reason for use of aspirin  Arthritis 16 15 1.37 0.66, 2.87 11 9 1.39 0.56, 3.46  Myocardial infarction 0 2 0 2  Other 225 112 2.01 1.57, 2.57 88 72 1.22 0.88, 1.71 Age at first use of aspirin, years  1–40 130 63 2.05 1.49, 2.83 72 43 1.68 1.13, 2.52  41–74 111 66 1.77 1.28, 2.46 27 40 0.69 0.41, 1.17 Age at most recent use of aspirin, years  16–49 147 55 2.58 1.86, 3.58 63 35 1.68 1.08, 2.60  50–74 94 74 1.38 0.99, 1.92 36 48 0.85 0.53, 1.34 Duration of aspirin use, years  1–5 158 64 2.43 1.79, 3.31 33 35 0.87 0.53, 1.45  6–51 83 65 1.38 0.98, 1.95 66 48 1.48 0.99, 2.20 No. of uses of aspirin  ≥1/day 89 34 2.54 1.67, 3.85 23 18 1.28 0.66, 2.48  ≥1/week 35 20 1.67 0.94, 2.96 12 9 1.09 0.43, 2.74  ≥1/month 116 73 1.72 1.26, 2.35 63 54 1.24 0.84, 1.82  Missing 1 2 1 2 Variable . Aspirin Use . First Use of Aspirin More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,390) . No. of Controls (n = 2,551) . ORa . 95% CIa . History of aspirin use (at least 3 months in 1 year)  No 2,291 2,468 1.00 Referent 2,291 2,468 1.00 Referent  Yes 241 129 1.91 1.52, 2.41 99 83 1.21 0.89, 1.66 Reason for use of aspirin  Arthritis 16 15 1.37 0.66, 2.87 11 9 1.39 0.56, 3.46  Myocardial infarction 0 2 0 2  Other 225 112 2.01 1.57, 2.57 88 72 1.22 0.88, 1.71 Age at first use of aspirin, years  1–40 130 63 2.05 1.49, 2.83 72 43 1.68 1.13, 2.52  41–74 111 66 1.77 1.28, 2.46 27 40 0.69 0.41, 1.17 Age at most recent use of aspirin, years  16–49 147 55 2.58 1.86, 3.58 63 35 1.68 1.08, 2.60  50–74 94 74 1.38 0.99, 1.92 36 48 0.85 0.53, 1.34 Duration of aspirin use, years  1–5 158 64 2.43 1.79, 3.31 33 35 0.87 0.53, 1.45  6–51 83 65 1.38 0.98, 1.95 66 48 1.48 0.99, 2.20 No. of uses of aspirin  ≥1/day 89 34 2.54 1.67, 3.85 23 18 1.28 0.66, 2.48  ≥1/week 35 20 1.67 0.94, 2.96 12 9 1.09 0.43, 2.74  ≥1/month 116 73 1.72 1.26, 2.35 63 54 1.24 0.84, 1.82  Missing 1 2 1 2 Abbreviations: CI, confidence interval; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab Table 4. Odds Ratios for Nasopharyngeal Carcinoma Associated With Aspirin Use in Southern China, 2010–2014 Variable . Aspirin Use . First Use of Aspirin More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,390) . No. of Controls (n = 2,551) . ORa . 95% CIa . History of aspirin use (at least 3 months in 1 year)  No 2,291 2,468 1.00 Referent 2,291 2,468 1.00 Referent  Yes 241 129 1.91 1.52, 2.41 99 83 1.21 0.89, 1.66 Reason for use of aspirin  Arthritis 16 15 1.37 0.66, 2.87 11 9 1.39 0.56, 3.46  Myocardial infarction 0 2 0 2  Other 225 112 2.01 1.57, 2.57 88 72 1.22 0.88, 1.71 Age at first use of aspirin, years  1–40 130 63 2.05 1.49, 2.83 72 43 1.68 1.13, 2.52  41–74 111 66 1.77 1.28, 2.46 27 40 0.69 0.41, 1.17 Age at most recent use of aspirin, years  16–49 147 55 2.58 1.86, 3.58 63 35 1.68 1.08, 2.60  50–74 94 74 1.38 0.99, 1.92 36 48 0.85 0.53, 1.34 Duration of aspirin use, years  1–5 158 64 2.43 1.79, 3.31 33 35 0.87 0.53, 1.45  6–51 83 65 1.38 0.98, 1.95 66 48 1.48 0.99, 2.20 No. of uses of aspirin  ≥1/day 89 34 2.54 1.67, 3.85 23 18 1.28 0.66, 2.48  ≥1/week 35 20 1.67 0.94, 2.96 12 9 1.09 0.43, 2.74  ≥1/month 116 73 1.72 1.26, 2.35 63 54 1.24 0.84, 1.82  Missing 1 2 1 2 Variable . Aspirin Use . First Use of Aspirin More Than 5 Years Before Interview . No. of Cases (n = 2,532) . No. of Controls (n = 2,597) . ORa . 95% CIa . No. of Cases (n = 2,390) . No. of Controls (n = 2,551) . ORa . 95% CIa . History of aspirin use (at least 3 months in 1 year)  No 2,291 2,468 1.00 Referent 2,291 2,468 1.00 Referent  Yes 241 129 1.91 1.52, 2.41 99 83 1.21 0.89, 1.66 Reason for use of aspirin  Arthritis 16 15 1.37 0.66, 2.87 11 9 1.39 0.56, 3.46  Myocardial infarction 0 2 0 2  Other 225 112 2.01 1.57, 2.57 88 72 1.22 0.88, 1.71 Age at first use of aspirin, years  1–40 130 63 2.05 1.49, 2.83 72 43 1.68 1.13, 2.52  41–74 111 66 1.77 1.28, 2.46 27 40 0.69 0.41, 1.17 Age at most recent use of aspirin, years  16–49 147 55 2.58 1.86, 3.58 63 35 1.68 1.08, 2.60  50–74 94 74 1.38 0.99, 1.92 36 48 0.85 0.53, 1.34 Duration of aspirin use, years  1–5 158 64 2.43 1.79, 3.31 33 35 0.87 0.53, 1.45  6–51 83 65 1.38 0.98, 1.95 66 48 1.48 0.99, 2.20 No. of uses of aspirin  ≥1/day 89 34 2.54 1.67, 3.85 23 18 1.28 0.66, 2.48  ≥1/week 35 20 1.67 0.94, 2.96 12 9 1.09 0.43, 2.74  ≥1/month 116 73 1.72 1.26, 2.35 63 54 1.24 0.84, 1.82  Missing 1 2 1 2 Abbreviations: CI, confidence interval; OR, odds ratio. a Adjusted for sex, age (5-year categories), residential area (Zhaoqing, Wuzhou, or Guiping/Pingnan), educational level (in years: ≤6, 7–9, 10–12, or >12), current housing type (building, cottage, or boat), current occupation (unemployed, farmer, blue-collar, white-collar, or other/unknown), cigarette smoking (current smoker, former smoker, or never smoker), first-degree family history of nasopharyngeal carcinoma (yes, no, or unknown), alcohol drinking (never, ever), tea drinking (never, ever), salt-preserved fish consumption in adulthood (energy-adjusted intake categories: 1 = lowest intake (none) to 4 = highest intake), and herbal medicine use. Open in new tab Web Table 6 shows estimated associations with use of balm or peppermint and use of flower or qufeng oil. Both were associated with a significantly elevated risk of NPC (OR = 2.63, 95% CI: 2.00, 3.46, and OR = 2.50, 95% CI: 1.97, 3.18, respectively). Use of both types of herbal medicines was associated with a further increased risk of NPC (OR = 2.95). Because age at first use of these medications was not assessed, individuals who initiated use within the past 5 years could not be excluded from the analysis. DISCUSSION This population-based case-control study, conducted in the NPC-endemic region of southern China, is one of the largest to investigate ENT-related medical history and medication use in the etiology of NPC. A recent history of chronic ENT disease, especially chronic sinusitis and otitis media, was associated with an increased risk of NPC, whereas chronic pharyngitis, nasal polyps, and septal abnormalities had no association with NPC. We also found that having been treated for (but not cured of) chronic ENT disease, chronic sinusitis, or chronic otitis media, or having uncured ENT disease, chronic sinusitis, chronic otitis media, or nasal polyps, was associated with an increased NPC risk. Our results are generally consistent with those from previous studies, including case-control studies in Thailand, Taiwan, the United States, Guangzhou, and Shanghai, China, which reported relative risks from 1.8 to 3.8 in association with a history of overall and specific chronic ENT diseases (7–11, 16, 17). However, after exclusion of individuals first diagnosed with ENT disease within the past 5 years, we found that these associations were substantially weakened and statistically nonsignificant, and only the associations with early age at first diagnosis of ENT disease and untreated nasal polyps remained significant, suggesting an influence of reverse causality or recall bias. The positive associations with treated and uncured ENT disease could be due to greater severity of preclinical disease within 5 years before NPC diagnosis. Nonspecific symptoms of NPC include epistaxis, unilateral nasal obstruction, and auditory complaints due to cranial nerve palsies (18). These symptoms are shared by benign ENT diseases, which should be included in the differential diagnosis for NPC. Although our results are perhaps most likely explained by recall bias or especially reverse causality, where ENT disease is a preclinical symptom of NPC, we could not preclude the possibility of a modest excess risk of NPC associated with ENT disease. A biologically plausible explanation for our findings and those from previous studies, including some that excluded ENT disease within 5 or 10 years of NPC diagnosis (9, 11, 16, 17), is a carcinogenic effect of chronic inflammation. This theory is well supported by epidemiologic evidence on other cancers, as exemplified by the likely association between inflammatory bowel disease and colorectal cancer (19), between Helicobacter pylori infection and gastric cancer (20), and many others (21, 22). However, research on potential inflammatory mechanisms of NPC development remains sparse. With reference to medications for ENT disease, most nasal drops act primarily via vasoconstriction to relieve symptoms of nasal obstruction but not to reduce inflammation. Our study found that a history of nasal drops use within the past 5 years was associated with increased risk of NPC. When individuals who started to use nasal drops within the past 5 years were excluded from the analysis, the association became weaker and statistically nonsignificant. This finding, which may be due to confounding by indication, probably reflects the inflamed condition of the upper respiratory tract before NPC diagnosis. After restriction of the analysis to those who started using aspirin at least 5 years earlier, aspirin use overall was not significantly associated with NPC risk, but earlier age at initiation and more distant past use were significantly associated with greater NPC risk. Aspirin use has been shown to be associated with reduced risk of several common malignancies (23–25). Results of epidemiologic studies analyzing aspirin use and head and neck cancers have been inconsistent (26–28). Thus far, only one small, hospital-based case-control study found a reduced risk of NPC in association with regular aspirin use (13). Although reverse causality must again be considered as a potential explanation for our finding of a positive association between aspirin use and NPC, this result might alternatively be due to the unique involvement of EBV infection, especially EBV reactivation, in the etiology of NPC (29). Given that aspirin can induce EBV activation and lytic cycle replication in EBV-positive cells (30, 31), aspirin might increase NPC risk by reactivating EBV. However, another EBV-associated malignancy, Hodgkin lymphoma, has been shown in some studies to be inversely associated with aspirin use (32, 33). This may be due to the distinctive role of EBV in the pathogenesis of malignancies from epithelial or lymphatic origin. Thus, the potential effect, if any, of aspirin use on NPC development, and whether it might interact with EBV infection, remains uncertain. Given that aspirin use for unspecified reasons other than arthritis and myocardial infarction was most strongly associated with NPC risk, the observed association could have been due to aspirin use for relieving pain, which is a preclinical symptom of NPC. Consistent with the results of a study conducted 4 decades ago in Taiwan, we found that use of balm or peppermint and flower or qufeng oil was associated with increased NPC risk (12), perhaps due to alleviation of potential premonitory symptoms of NPC. However, because we did not collect information about age at first use of these products, we could not conduct a secondary analysis to examine whether the observed positive associations were due to confounding by indication. Our study has some weaknesses. First, all information was retrospective and self-reported. Because NPC cases may overestimate or be more likely than controls to recall their history of chronic ENT diseases and related medication use, especially close to the time of NPC diagnosis, we conducted secondary analyses restricted to chronic ENT diseases and related medication use at least 5 years prior to diagnosis/interview. In general, the magnitude of the associations was attenuated, suggesting an influence of reverse causality and/or recall bias. Second, the lack of blinding of interviewers to case-control status could also have contributed to some amount of information bias. Third, invalid contact information for potential controls might have resulted in some underrepresentation of younger, urban, residentially mobile controls. However, we adjusted for age and socioeconomic factors in the multivariate analysis, and the overall control participation rate was relatively high. Although health-care utilization among enrolled controls might have been lower than in the source population (34), resulting in underascertainment of ENT diseases among controls and overestimation of associations with NPC risk, the attenuation of associations after restriction to diagnoses at least 5 years ago indicates that any such bias was modest. In conclusion, our results indicate that positive association between ENT diseases and NPC risk detected in our study and others is probably due to bias, especially reverse causation resulting from preclinical ENT disease within 5 years of NPC diagnosis. In light of the lack of an association between more distant past ENT disease and NPC risk, our findings call into question any potential role of chronic ENT inflammation in the etiology of NPC. Similarly, the observed associations between the medication use related to ENT diseases and NPC risk might well be due to confounding by indication. However, we could not exclude the possibility of a modest excess NPC risk in association with ENT diseases and their related medication use. Future studies, with particular attention to these pitfalls, are needed to clarify whether there exists a weak association between chronic ENT diseases and/or related medication use, and NPC risk. ACKNOWLEDGMENTS Author affiliations: Department of Otolaryngology–Head and Neck Surgery, First Affiliated Hospital of Guangxi Medical University, Nanning, China (Xiling Xiao, Zhe Zhang, Guangwu Huang); Exponent, Inc., Center for Health Sciences, Menlo Park, California (Ellen T. Chang); Division of Epidemiology, Department of Health Research and Policy, Stanford University School of Medicine, Stanford, California (Ellen T. Chang); Division of Cancer Epidemiology and Genetics, National Cancer Institute, Rockville, Maryland (Zhiwei Liu); Department of Cancer Prevention Center, Sun Yat-sen University Cancer Center, Guangzhou, China (Qing Liu, Shang-Hang Xie, Su-Mei Cao); State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Sun Yat-sen University Cancer Center, Guangzhou, China (Qing Liu, Shang-Hang Xie, Su-Mei Cao, Jian-Yong Shao, Wei-Hua Jia, Yi-Xin Zeng); Department of Clinical Laboratory, Wuzhou Red Cross Hospital, Wuzhou, China (Yonglin Cai, Yuming Zheng); Wuzhou Health System Key Laboratory for Nasopharyngeal Carcinoma Etiology and Molecular Mechanism, Wuzhou, China (Yonglin Cai, Yuming Zheng); State Key Laboratory for Infectious Diseases Prevention and Control, Institute for Viral Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, China (Guomin Chen, Yi Zeng); Sihui Cancer Institute, Sihui, China (Qi-Hong Huang); Cangwu Institute for Nasopharyngeal Carcinoma Control and Prevention, Wuzhou, China (Jian Liao); Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden (Yufeng Chen, Hans-Olov Adami, Weimin Ye); Key Laboratory of High-Incidence Tumor Prevention and Treatment, Guangxi Medical University, Ministry of Education, Nanning, China (Longde Lin); Department of Microbiology, Tumor and Cell Biology, Karolinska Institutet, Stockholm, Sweden (Ingemar Ernberg); Beijing Hospital, Beijing, China (Yi-Xin Zeng); and Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Hans-Olov Adami). G.H., Y.Z., Y.-X.Z., H.-O.A., and W.Y. contributed equally to this work. This work was supported by the US National Cancer Institute (grant R01 CA115873), Swedish Research Council (grants 2015-02625, 2015-06268, and 2017-05814), and Karolinska Institutet (Distinguished Professor Award to H.-O.A. (Dnr: 2368/10-221)). The work from the Guiping/Pingnan area was supported by grants from the New Century Excellent Talents in University (grant NCET-12-0654), National Basic Research Program of China (grant 2011CB504300), and Guangxi Natural Science Foundation (grant 2013GXNSFGA 019002). We thank the members of the External Advisory Board, including Drs. Curtis Harris and Allan Hildesheim (National Cancer Institute), Dr. Mary-Claire King (University of Washington), Dr. Xihong Lin (Harvard School of Public Health), Dr. Youlin Qiao (Chinese Academy of Medical Sciences), and Dr. Weicheng You (Peking University Health Science Center). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. Conflict of interest: none declared. Abbreviations CI confidence interval EBV Epstein-Barr virus ENT ear-nose-throat NPC nasopharyngeal carcinoma OR odds ratio REFERENCES 1 Chang ET , Adami HO. The enigmatic epidemiology of nasopharyngeal carcinoma . Cancer Epidemiol Biomarkers Prev . 2006 ; 15 ( 10 ): 1765 – 1777 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Wu HC , Lin YJ, Lee JJ, et al. . Functional analysis of EBV in nasopharyngeal carcinoma cells . Lab Invest . 2003 ; 83 ( 6 ): 797 – 812 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Chang ET , Liu Z, Hildesheim A, et al. . Active and passive smoking and risk of nasopharyngeal carcinoma: a population-based case-control study in southern China . Am J Epidemiol . 2017 ; 185 ( 12 ): 1272 – 1280 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Liu Z , Chang ET, Liu Q, et al. . Oral hygiene and risk of nasopharyngeal carcinoma-a population-based case-control study in China . Cancer Epidemiol Biomarkers Prev . 2016 ; 25 ( 8 ): 1201 – 1207 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Bei JX , Zuo XY, Liu WS, et al. . Genetic susceptibility to the endemic form of NPC . Chin Clin Oncol . 2016 ; 5 ( 2 ): 15 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Mantovani A . Cancer: inflaming metastasis . Nature . 2009 ; 457 ( 7225 ): 36 – 37 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Ekburanawat W , Ekpanyaskul C, Brennan P, et al. . Evaluation of non-viral risk factors for nasopharyngeal carcinoma in Thailand: results from a case-control study . Asian Pac J Cancer Prev . 2010 ; 11 ( 4 ): 929 – 932 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 8 Henderson BE , Louie E, SooHoo Jing J, et al. . Risk factors associated with nasopharyngeal carcinoma . N Engl J Med . 1976 ; 295 ( 20 ): 1101 – 1106 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Huang WY , Lin CC, Jen YM, et al. . Association between adult otitis media and nasopharyngeal cancer: a nationwide population-based cohort study . Radiother Oncol . 2012 ; 104 ( 3 ): 338 – 342 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Hung SH , Chen PY, Lin HC, et al. . Association of rhinosinusitis with nasopharyngeal carcinoma: a population-based study . Laryngoscope . 2014 ; 124 ( 7 ): 1515 – 1520 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Wu EL , Riley CA, Hsieh MC, et al. . Chronic sinonasal tract inflammation as a precursor to nasopharyngeal carcinoma and sinonasal malignancy in the United States . Int Forum Allergy Rhinol . 2017 ; 7 ( 8 ): 786 – 793 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Lin TM , Chen KP, Lin CC, et al. . Retrospective study on nasopharyngeal carcinoma . J Natl Cancer Inst . 1973 ; 51 ( 5 ): 1403 – 1408 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Di Maso M , Bosetti C, La Vecchia C, et al. . Regular aspirin use and nasopharyngeal cancer risk: a case-control study in Italy . Cancer Epidemiol . 2015 ; 39 ( 4 ): 545 – 547 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Ye W , Chang ET, Liu Z, et al. . Development of a population-based cancer case-control study in southern china . Oncotarget . 2017 ; 8 ( 50 ): 87073 – 87085 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 15 Sham JS , Wei WI, Lau SK, et al. . Serous otitis media: an opportunity for early recognition of nasopharyngeal carcinoma . Arch Otolaryngol Head Neck Surg . 1992 ; 118 ( 8 ): 794 – 797 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Yu MC , Garabrant DH, Huang TB, et al. . Occupational and other non‐dietary risk factors for nasopharyngeal carcinoma in Guangzhou, China . Int J Cancer . 1990 ; 45 ( 6 ): 1033 – 1039 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Yuan JM , Wang XL, Xiang YB, et al. . Non‐dietary risk factors for nasopharyngeal carcinoma in Shanghai, China . Int J Cancer . 2000 ; 85 ( 3 ): 364 – 369 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Chua MLK , Wee JTS, Hui EP, et al. . Nasopharyngeal carcinoma . Lancet . 2016 ; 387 ( 10022 ): 1012 – 1024 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Adami HO , Bretthauer M, Emilsson L, et al. . The continuing uncertainty about cancer risk in inflammatory bowel disease . Gut . 2016 ; 65 ( 6 ): 889 – 893 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Venerito M , Vasapolli R, Rokkas T, et al. . Helicobacter pylori, gastric cancer and other gastrointestinal malignancies . Helicobacter . 2017 ; 22 ( suppl 1 ): e12413 . Google Scholar Crossref Search ADS WorldCat 21 Wu Y , Antony S, Meitzler JL, et al. . Molecular mechanisms underlying chronic inflammation-associated cancers . Cancer Lett . 2014 ; 345 ( 2 ): 164 – 173 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Chai EZ , Siveen KS, Shanmugam MK, et al. . Analysis of the intricate relationship between chronic inflammation and cancer . Biochem J . 2015 ; 468 ( 1 ): 1 – 15 . Google Scholar Crossref Search ADS PubMed WorldCat 23 García Rodríguez LA , Soriano-Gabarró M, Bromley S, et al. . New use of low-dose aspirin and risk of colorectal cancer by stage at diagnosis: a nested case-control study in UK general practice . BMC Cancer . 2017 ; 17 : 637 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Risch HA , Lu L, Streicher SA, et al. . Aspirin use and reduced risk of pancreatic cancer . Cancer Epidemiol Biomarkers Prev . 2017 ; 26 ( 1 ): 68 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Yang YS , Kornelius E, Chiou JY, et al. . Low-dose aspirin reduces breast cancer risk in women with diabetes: a nationwide retrospective cohort study in Taiwan . J Womens Health (Larchmt) . 2017 ; 26 ( 12 ): 1278 – 1284 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Jayaprakash V , Rigual NR, Moysich KB, et al. . Chemoprevention of head and neck cancer with aspirin: a case-control study . Arch Otolaryngol Head Neck Surg . 2006 ; 132 ( 11 ): 1231 – 1236 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Macfarlane TV , Lefevre K, Watson MC. Aspirin and non-steroidal anti-inflammatory drug use and the risk of upper aerodigestive tract cancer . Br J Cancer . 2014 ; 111 ( 9 ): 1852 – 1859 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Rosenquist K , Wennerberg J, Schildt EB, et al. . Oral status, oral infections and some lifestyle factors as risk factors for oral and oropharyngeal squamous cell carcinoma. A population-based case-control study in southern Sweden . Acta Otolaryngol . 2005 ; 125 ( 12 ): 1327 – 1336 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Chen Y , Zhao W, Lin L, et al. . Nasopharyngeal Epstein-Barr virus load: an efficient supplementary method for population-based nasopharyngeal carcinoma screening . PLoS One . 2015 ; 10 ( 7 ): e0132669 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Liu SF , Wang H, Li ZJ, et al. . Aspirin induces lytic cytotoxicity in Epstein-Barr virus-positive cells . Eur J Pharmacol . 2008 ; 589 ( 1–3 ): 8 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Ordonez P , Nandakumar A, Koriyama C, et al. . Cytotoxic effects of NF-κB inhibitors in combination with anti-herpes agents on Epstein-Barr virus–positive gastric carcinoma in vitro . Mol Med Rep . 2016 ; 14 ( 3 ): 2359 – 2367 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Chang ET , Cronin-Fenton DP, Friis S, et al. . Aspirin and other nonsteroidal anti-inflammatory drugs in relation to Hodgkin lymphoma risk in northern Denmark . Cancer Epidemiol Biomarkers Prev . 2010 ; 19 ( 1 ): 59 – 64 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Chang ET , Frøslev T, Sørensen HT, et al. . A nationwide study of aspirin, other non-steroidal anti-inflammatory drugs, and Hodgkin lymphoma risk in Denmark . Br J Cancer . 2011 ; 105 ( 11 ): 1776 – 1782 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Li J , Shi L, Liang H, et al. . Urban-rural disparities in health care utilization among Chinese adults from 1993 to 2011 . BMC Health Serv Res . 2018 ; 18 : 102 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018.

Tasevska, Natasha; Pettinger, Mary; Kipnis, Victor; Midthune, Douglas; Tinker, Lesley F; Potischman, Nancy; Neuhouser, Marian L; Beasley, Jeannette M; Van Horn, Linda; Howard, Barbara V; Liu, Simin; Manson, JoAnn E; Shikany, James M; Thomson, Cynthia A; Prentice, Ross L

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy115pmid: 29868784

Abstract The inconsistent findings from epidemiologic studies relating total sugars (TS) consumption to cardiovascular disease (CVD) or type 2 diabetes (T2D) risk may be partly due to measurement error in self-reported intake. Using regression calibration equations developed based on the predictive biomarker for TS and recovery biomarker for energy, we examined the association of TS with T2D and CVD risk, before and after dietary calibration, in 82,254 postmenopausal women participating in the Women’s Health Initiative Observational Study. After up to 16 years of follow-up (1993–2010), 6,621 T2D and 5,802 CVD incident cases were identified. The hazard ratio for T2D per 20% increase in calibrated TS was 0.94 (95% confidence interval (CI): 0.77, 1.15) in multivariable energy substitution, and 1.00 (95% CI: 0.85, 1.18) in energy partition models. Multivariable hazard ratios for total CVD were 0.97 (95% CI: 0.87, 1.09) from energy substitution, and 0.91 (95% CI: 0.80, 1.04) from energy partition models. Uncalibrated TS generated a statistically significant inverse association with T2D and total CVD risk in multivariable energy substitution and energy partition models. The lack of conclusive findings from our calibrated analyses may be due to the low explanatory power of the calibration equations for TS, which could have led to incomplete deattenuation of the risk estimates. calibration, cardiovascular disease, diabetes, diet, measurement error, total sugars, Women’s Health Initiative The development of a predictive biomarker for total sugars (TS) consumption (1) has enabled the study of measurement error (ME) in self-reported TS intake (2) and ME correction in disease association studies (3). Predictive biomarkers are a distinct group of dietary biomarkers that can predict (i.e., estimate) intake after being calibrated for their biases (2). Hence, these biomarkers can be used as reference validation instruments similar to recovery biomarkers, provided their biases have been quantified in an appropriate feeding study. Type 2 diabetes (T2D) and cardiovascular disease (CVD) are 2 of the most prevalent, largely preventable chronic diseases worldwide (4). Approximately 30.2 million adults in the United States have diabetes; of these cases, 90%–95% are T2D (5), whereas more than 1 in 3 US adults have CVD (including hypertension) (6). Dietary sugars (e.g., TS, sucrose, fructose, glucose) or sources of sugars (e.g., sugar-sweetened beverages) in relation to T2D or CVD risk have long been investigated (7–13). Plausible mechanisms include hyperinsulinemia, insulin resistance, inflammation, and oxidative stress promoted by the glycemic effect of diets high in refined sugars (14). More specifically, de novo lipogenesis in the liver induced by high levels of fructose consumption results in dyslipidemia, insulin resistance, and hyperuricemia (15, 16). In addition, potential excess in energy intake associated with TS consumption elicits indirect effects of sugars on these 2 outcomes mediated by overweight and obesity (17). Although establishing the role of TS or individual sugars as nutrients in the etiology of these 2 outcomes poses a challenge (7, 8, 18–20), the adverse evidence for sugar-sweetened beverages, a liquid form of added sugars (21), in relation to CVD (9) and T2D (10) is more consistent. This may result from the fact that sugars in beverages may have more pronounced adverse metabolic effects related to rapid metabolism as compared with sugars in solid foods, which are embedded within the food matrix and thus are slower to enter metabolic pathways (22). Sugars from liquids may also have weaker satiating effect, leading to consumption of larger portion sizes (23). Furthermore, it is plausible that inconsistencies in the evidence arise from differential misreporting of sugars-related intake. Sugar-sweetened beverages come in common, predefined serving sizes; hence, they are generally cognitively easier to report on any self-reporting instrument, whereas estimates of individual sugars are generated through food and nutrient databases from reported intake occasions of multiple food items associated with various levels of ME (24–26). Using the predictive biomarker of TS, it has been found in several validation studies that self-reports underestimated TS consumption (2, 3, 27) and were associated with attenuation factors that could bias disease relative risks toward the null (i.e., ranging from 1.1 to 1.5 for a true relative risk = 2) (2, 3), even when they are assumed to be free of systematic biases. When dietary validation studies are incorporated within cohorts, adjustment for ME in self-reported intake or observed relative risks can be made (28). When a validation study with objective dietary biomarkers (24) was used in the Women’s Health Initiative (WHI) Observational Study (OS) (29), after ME correction, energy intake was associated with increased risk of breast cancer (30), all cancer (31, 32), CVD (33), and T2D (34), and protein intake was associated with increased risk of T2D (34) and decreased risk of frailty (35). In the current study, we used data from a WHI validation study with the predictive biomarker of TS intake (3) to correct for ME (i.e., to calibrate) self-reported TS in all WHI OS participants and to explore the associations of TS intake with CVD and T2D risk before and after dietary calibration. These associations were explored through 2 energy adjustment methods: energy substitution (ES) and energy partition (EP) models (36, 37). METHODS WHI OS Study The WHI OS is a prospective study involving 93,676 postmenopausal women aged 50–79 years, enrolled during 1993–1998 from 40 clinical centers across the United States. The study design has been described in detail elsewhere (29, 38, 39). Briefly, all participants completed baseline questionnaires inquiring about demographic characteristics, and personal and family medical history. The WHI semiquantitative food frequency questionnaire was used to assess participants’ usual diet over the previous 3 months. The food frequency questionnaire included a list of 122 foods or food groups, questions about frequency of intake and portion size, 19 adjustment questions on how foods were prepared, and 4 summary questions on usual intake of fruit, vegetables, and fat added to foods and used in cooking (40). Daily energy and nutrient intake were calculated using the University of Minnesota’s Nutrition Data System for Research, version 2005 (Nutrition Coordinating Center, Minneapolis, Minnesota). TS represent the sum of monosaccharides (i.e., glucose, fructose, and galactose) and disaccharides (i.e., sucrose, lactose, and maltose). Participants’ smoking and alcohol habits and recreational physical activity were assessed using the WHI Personal Habit Questionnaire (41). To generate activity-related energy expenditure, estimates of recreational physical activity were combined with estimates of housework, yardwork, sitting, and sleeping reported on other WHI questionnaires in metabolic equivalents per week and computed into kilocalories per day, according to the following calculation: total metabolic equivalents per week multiplied by body weight (in kilograms) divided by 7. At baseline, body height and weight, and waist circumference were measured, and body mass index (BMI) was calculated by dividing weight (in kilograms) by height (in meters) squared. Nutrition and Physical Activity Assessment Study The Nutrition and Physical Activity Assessment Study (NPAAS) was an ancillary study nested within the WHI OS in which detailed dietary and physical activity measurements were collected from a representative subsample of the WHI OS participants (24). All participants completed a doubly labeled water protocol, a 24-hour urine collection, and indirect calorimetry, as previously described in detail (24). NPAAS included 450 women participating in WHI OS who were ages 60–91 years at NPAAS baseline (2007–2009) and recruited from 9 WHI clinical centers. Biomarker-based estimates of TS (3), energy, and protein intake (24), ratio of sodium to potassium intake (42), and activity-related energy expenditure (43) were derived as previously reported. Furthermore, for each of these exposures, a calibration equation that regresses log of biomarker-predicted exposure on log of self-reported estimate, along with other personal characteristics, was developed (3, 24, 42, 43). Ascertainment of outcomes Follow-up for cases of CVD and T2D was calculated from baseline (1993–1998) until diagnosis and, for noncases, until censoring on September 30, 2010, last follow-up, or death, whichever came first. CVD incident cases were reported annually by a self-administered questionnaire. Reports were then reviewed by local WHI physician adjudicators, who assigned diagnoses on the basis of medical records, death certificates, and autopsy reports, which were then forwarded to central physician adjudicators for independent confirmation (44). CVD outcomes included incident cases from any CVD (total CVD); coronary heart disease (CHD), and stroke. Findings on nonfatal myocardial infarction, coronary death, heart failure, coronary artery bypass graft, percutaneous coronary intervention, ischemic and hemorrhagic stroke are reported in Web Table 1 (available at https://academic.oup.com/aje). T2D included self-reported incident cases identified via annual mailed questionnaires when participants were asked to report if a doctor ever prescribed pills or insulin shots for diabetes after the participant had been enrolled in the study. Substantial agreement between self-report and medication inventory or medical record-verified T2D was demonstrated in earlier work in this cohort (45, 46). Analytical data set From 93,676 participants, we excluded women with implausible self-reported energy intake (<600 or >5,000 kcal/day) on the food frequency questionnaire or missing data on diet (n = 3,662), BMI (n = 1,105), physical activity level (n = 2,861), smoking status (n = 1,351), educational level (n = 767), marital status, postmenopausal hormone therapy use, history of hypertension or hypercholesterolemia (n = 2,111), and those with no follow-up (n = 471). To generate the CVD analytical cohort (n = 64,751), from 82,254 women, we further excluded those with history of CVD at baseline (n = 16,301) and with missing data on CVD-specific model covariates (i.e., history of treated T2D, statin use, aspirin use, family history of CVD) (n = 1,507). For T2D analysis, from 82,254 women, we further excluded prevalent cases of T2D at baseline (n = 3,238) and women with missing data on T2D-specific model covariates (i.e., history of CVD and family history of diabetes) (n = 3,957) for a final T2D analytical cohort of 75,320 women. (Some participants were excluded based on more than one exclusion criterion.) Statistical analysis The calibration equations for TS, energy, protein, and ratio of sodium to potassium intake, and activity-related energy expenditure were redeveloped using the NPAAS data by regressing log-transformed biomarker-based values on log-transformed self-reported estimates, along with covariates included in the original respective calibration equations (3, 33, 42, 43) and covariates from the respective disease risk models, in accordance with the standard regression calibration methodology (47, 48). We developed individual calibration equations for age- and energy-adjusted (basic), and multivariable ES and EP models for T2D and CVD (see Web Tables 2–5). Hazard ratios and 95% confidence intervals for a 20% increase in calibrated or uncalibrated TS intake were estimated by a Cox proportional hazards regression model stratified on age in 5-year categories to allow for different baseline hazards by age category in basic and multivariable models. These hazard ratio estimates were based on linear modeling of the log of hazard ratios on the logarithm of calibrated intake. Based on median intake, a 20% increase corresponds to 18.0 g/1,000 kcal for calibrated and 12.6 g/1,000 kcal for uncalibrated TS. We report findings from 2 different modeling approaches for energy intake adjustment (36, 37). We used the ES model to investigate the association between TS and outcomes when substituting TS (g/1,000 kcal) for other energy-contributing nutrients not included in the model while keeping total energy intake constant (kcal/day). We used the EP model to investigate the association between TS and outcomes when adding TS along with energy from sugars (g/day) while keeping nonsugars and nonalcohol energy constant (kcal/day), calculated as total energy minus energy from alcohol and energy from TS intake. The standard errors for hazard ratios from the models with calibrated estimates were estimated by a bootstrap procedure with 1,000 bootstrap samples, to account for the random variation in calibration equation coefficient estimates. We conducted stratified analyses by BMI (<25.0, 25.0–29.9, or ≥30.0) using multivariable EP models with calibrated TS intake, and we only present findings for the composite CVD outcomes (i.e., total CVD and CHD) because of the limited number of cases from individual outcomes in some strata. Because participants with hypercholesterolemia, history of hypertension or CVD, or family history of diabetes may have changed their diet because of their increased risk for T2D, the models were repeated after excluding these women (n = 47,109). Furthermore, all CVD models were repeated after excluding cases diagnosed during the first year of follow-up. All analyses were conducted using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina) and R, version 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria). The P values for statistical tests were 2 tailed and considered statistically significant at a level of less than 0.05. RESULTS Baseline characteristics of the study population in the WHI OS and NPAAS are listed in Tables 1 and 2 (for more detail see Web Table 5). WHI OS participants were older, predominantly white, and had lower BMI compared with participants in NPAAS, which oversampled participants of younger age, race or ethnicities other than non-Hispanic white, and higher BMI (33). The median TS density intake ranged from 60 to 62 g/1,000 kcal before calibration (Web Table 5), and from 79 to 95 g/1,000 kcal after calibration, in WHI OS and NPAAS populations (Table 2). Table 1. Baseline Characteristics of Participants in the Women’s Health Initiative Observational Study Enrolled During 1993–1998 and Nutrition and Physical Activity Assessment Study Enrolled During 2007–2009 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . No. . % . No. . % . No. . % . No. . % . Age group at screening, years  ≤59 22,300 34.5 239 69.9 24,399 32.4 260 67.9  60–69 28,426 43.9 86 25.1 33,120 44.0 99 25.8  ≥70 14,025 21.7 17 5.0 17,801 23.6 24 6.3 White race 55,132 85.1 226 66.1 65,198 86.6 261 68.1 College degree or higher 28,670 44.3 176 51.5 33,314 44.2 205 53.5 Family history of T2D 19,925 30.8 106 31.0 23,632 31.4 123 32.1 Family history of CVD 43,170 66.7 217 63.5 50,803 67.4 237 61.9 Treated hypertension 14,229 22.0 48 14.0 17,829 23.7 57 14.9 Current smokers 3,864 6.0 17 5.0 4,502 6.0 17 4.4 Alcohol intake  Never or past 17,807 27.5 85 24.9 20,576 27.3 91 23.8  1–6 drinks/week 38,214 59.0 216 63.2 44,530 59.1 244 63.7  ≥7 drinks/week 8,730 13.5 41 12.0 10,214 13.6 48 12.5 Use of hormone therapya 30,010 46.3 174 50.9 34,774 46.2 199 51.9 Treated high cholesterol 8,003 12.4 24 7.0 10,549 14.0 27 7.0 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . No. . % . No. . % . No. . % . No. . % . Age group at screening, years  ≤59 22,300 34.5 239 69.9 24,399 32.4 260 67.9  60–69 28,426 43.9 86 25.1 33,120 44.0 99 25.8  ≥70 14,025 21.7 17 5.0 17,801 23.6 24 6.3 White race 55,132 85.1 226 66.1 65,198 86.6 261 68.1 College degree or higher 28,670 44.3 176 51.5 33,314 44.2 205 53.5 Family history of T2D 19,925 30.8 106 31.0 23,632 31.4 123 32.1 Family history of CVD 43,170 66.7 217 63.5 50,803 67.4 237 61.9 Treated hypertension 14,229 22.0 48 14.0 17,829 23.7 57 14.9 Current smokers 3,864 6.0 17 5.0 4,502 6.0 17 4.4 Alcohol intake  Never or past 17,807 27.5 85 24.9 20,576 27.3 91 23.8  1–6 drinks/week 38,214 59.0 216 63.2 44,530 59.1 244 63.7  ≥7 drinks/week 8,730 13.5 41 12.0 10,214 13.6 48 12.5 Use of hormone therapya 30,010 46.3 174 50.9 34,774 46.2 199 51.9 Treated high cholesterol 8,003 12.4 24 7.0 10,549 14.0 27 7.0 Abbreviations: CVD, cardiovascular disease; NPAAS, Nutrition and Physical Activity Assessment Study; OS, Observational Study; T2D, type 2 diabetes; WHI, Women’s Health Initiative. a Estrogen alone or estrogen plus progestin user. Open in new tab Table 1. Baseline Characteristics of Participants in the Women’s Health Initiative Observational Study Enrolled During 1993–1998 and Nutrition and Physical Activity Assessment Study Enrolled During 2007–2009 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . No. . % . No. . % . No. . % . No. . % . Age group at screening, years  ≤59 22,300 34.5 239 69.9 24,399 32.4 260 67.9  60–69 28,426 43.9 86 25.1 33,120 44.0 99 25.8  ≥70 14,025 21.7 17 5.0 17,801 23.6 24 6.3 White race 55,132 85.1 226 66.1 65,198 86.6 261 68.1 College degree or higher 28,670 44.3 176 51.5 33,314 44.2 205 53.5 Family history of T2D 19,925 30.8 106 31.0 23,632 31.4 123 32.1 Family history of CVD 43,170 66.7 217 63.5 50,803 67.4 237 61.9 Treated hypertension 14,229 22.0 48 14.0 17,829 23.7 57 14.9 Current smokers 3,864 6.0 17 5.0 4,502 6.0 17 4.4 Alcohol intake  Never or past 17,807 27.5 85 24.9 20,576 27.3 91 23.8  1–6 drinks/week 38,214 59.0 216 63.2 44,530 59.1 244 63.7  ≥7 drinks/week 8,730 13.5 41 12.0 10,214 13.6 48 12.5 Use of hormone therapya 30,010 46.3 174 50.9 34,774 46.2 199 51.9 Treated high cholesterol 8,003 12.4 24 7.0 10,549 14.0 27 7.0 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . No. . % . No. . % . No. . % . No. . % . Age group at screening, years  ≤59 22,300 34.5 239 69.9 24,399 32.4 260 67.9  60–69 28,426 43.9 86 25.1 33,120 44.0 99 25.8  ≥70 14,025 21.7 17 5.0 17,801 23.6 24 6.3 White race 55,132 85.1 226 66.1 65,198 86.6 261 68.1 College degree or higher 28,670 44.3 176 51.5 33,314 44.2 205 53.5 Family history of T2D 19,925 30.8 106 31.0 23,632 31.4 123 32.1 Family history of CVD 43,170 66.7 217 63.5 50,803 67.4 237 61.9 Treated hypertension 14,229 22.0 48 14.0 17,829 23.7 57 14.9 Current smokers 3,864 6.0 17 5.0 4,502 6.0 17 4.4 Alcohol intake  Never or past 17,807 27.5 85 24.9 20,576 27.3 91 23.8  1–6 drinks/week 38,214 59.0 216 63.2 44,530 59.1 244 63.7  ≥7 drinks/week 8,730 13.5 41 12.0 10,214 13.6 48 12.5 Use of hormone therapya 30,010 46.3 174 50.9 34,774 46.2 199 51.9 Treated high cholesterol 8,003 12.4 24 7.0 10,549 14.0 27 7.0 Abbreviations: CVD, cardiovascular disease; NPAAS, Nutrition and Physical Activity Assessment Study; OS, Observational Study; T2D, type 2 diabetes; WHI, Women’s Health Initiative. a Estrogen alone or estrogen plus progestin user. Open in new tab Table 2. Geometric Means for Anthropometric and Dietary Characteristics of Participants in the Women’s Health Initiative Observational Study Enrolled During 1993–1998 and Nutrition and Physical Activity Assessment Study Enrolled During 2007–2009 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . GM . 95% CIa . GM . 95% CI . GM . 95% CI . GM . 95% CI . Body mass indexb 26.50 26.46, 26.54 27.24 26.64, 27.86 26.45 26.41, 26.49 27.00 26.42, 27.59 Waist circumference, cm 83.17 83.07, 83.27 82.97 81.55, 84.41 83.10 83.01, 83.18 82.42 81.08, 83.78 Calibrated total energy, kcal 2,156 2,155, 2,158 2,250 2,223, 2,277 2,173 2,172, 2,175 2,273 2,246, 2,300 Calibrated total sugars density, g/1,000 kcal 95.00 94.60, 95.30 86.40 82.30, 90.70 84.30 84.10, 84.60 78.70 76.20, 81.40 Calibrated protein density, g/1,000 kcal 36.40 36.30, 36.40 35.80 35.30, 36.40 34.70 34.70, 34.70 34.20 33.90, 34.50 Calibrated Na/K 1.33 1.32, 1.33 1.40 1.36, 1.44 Calibrated AREE, kcal 860 858, 861 978 952, 1,004 854 852, 356 986 962, 1,011 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . GM . 95% CIa . GM . 95% CI . GM . 95% CI . GM . 95% CI . Body mass indexb 26.50 26.46, 26.54 27.24 26.64, 27.86 26.45 26.41, 26.49 27.00 26.42, 27.59 Waist circumference, cm 83.17 83.07, 83.27 82.97 81.55, 84.41 83.10 83.01, 83.18 82.42 81.08, 83.78 Calibrated total energy, kcal 2,156 2,155, 2,158 2,250 2,223, 2,277 2,173 2,172, 2,175 2,273 2,246, 2,300 Calibrated total sugars density, g/1,000 kcal 95.00 94.60, 95.30 86.40 82.30, 90.70 84.30 84.10, 84.60 78.70 76.20, 81.40 Calibrated protein density, g/1,000 kcal 36.40 36.30, 36.40 35.80 35.30, 36.40 34.70 34.70, 34.70 34.20 33.90, 34.50 Calibrated Na/K 1.33 1.32, 1.33 1.40 1.36, 1.44 Calibrated AREE, kcal 860 858, 861 978 952, 1,004 854 852, 356 986 962, 1,011 Abbreviations: AREE, activity-related energy expenditure; CI, confidence interval; CVD, cardiovascular disease; GM, geometric mean; Na/K, ratio of sodium to potassium; NPAAS, Nutrition and Physical Activity Assessment Study; T2D, type 2 diabetes; WHI-OS, Women’s Health Initiative Observational Study. a Naïve 95% confidence interval for calibrated estimates. b Weight (kg)/height (m)2. Open in new tab Table 2. Geometric Means for Anthropometric and Dietary Characteristics of Participants in the Women’s Health Initiative Observational Study Enrolled During 1993–1998 and Nutrition and Physical Activity Assessment Study Enrolled During 2007–2009 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . GM . 95% CIa . GM . 95% CI . GM . 95% CI . GM . 95% CI . Body mass indexb 26.50 26.46, 26.54 27.24 26.64, 27.86 26.45 26.41, 26.49 27.00 26.42, 27.59 Waist circumference, cm 83.17 83.07, 83.27 82.97 81.55, 84.41 83.10 83.01, 83.18 82.42 81.08, 83.78 Calibrated total energy, kcal 2,156 2,155, 2,158 2,250 2,223, 2,277 2,173 2,172, 2,175 2,273 2,246, 2,300 Calibrated total sugars density, g/1,000 kcal 95.00 94.60, 95.30 86.40 82.30, 90.70 84.30 84.10, 84.60 78.70 76.20, 81.40 Calibrated protein density, g/1,000 kcal 36.40 36.30, 36.40 35.80 35.30, 36.40 34.70 34.70, 34.70 34.20 33.90, 34.50 Calibrated Na/K 1.33 1.32, 1.33 1.40 1.36, 1.44 Calibrated AREE, kcal 860 858, 861 978 952, 1,004 854 852, 356 986 962, 1,011 Characteristics . CVD Analytical Cohort . T2D Analytical Cohort . WHI OS (n = 64,751) . NPAAS (n = 342) . WHI OS (n = 75,320) . NPAAS (n = 383) . GM . 95% CIa . GM . 95% CI . GM . 95% CI . GM . 95% CI . Body mass indexb 26.50 26.46, 26.54 27.24 26.64, 27.86 26.45 26.41, 26.49 27.00 26.42, 27.59 Waist circumference, cm 83.17 83.07, 83.27 82.97 81.55, 84.41 83.10 83.01, 83.18 82.42 81.08, 83.78 Calibrated total energy, kcal 2,156 2,155, 2,158 2,250 2,223, 2,277 2,173 2,172, 2,175 2,273 2,246, 2,300 Calibrated total sugars density, g/1,000 kcal 95.00 94.60, 95.30 86.40 82.30, 90.70 84.30 84.10, 84.60 78.70 76.20, 81.40 Calibrated protein density, g/1,000 kcal 36.40 36.30, 36.40 35.80 35.30, 36.40 34.70 34.70, 34.70 34.20 33.90, 34.50 Calibrated Na/K 1.33 1.32, 1.33 1.40 1.36, 1.44 Calibrated AREE, kcal 860 858, 861 978 952, 1,004 854 852, 356 986 962, 1,011 Abbreviations: AREE, activity-related energy expenditure; CI, confidence interval; CVD, cardiovascular disease; GM, geometric mean; Na/K, ratio of sodium to potassium; NPAAS, Nutrition and Physical Activity Assessment Study; T2D, type 2 diabetes; WHI-OS, Women’s Health Initiative Observational Study. a Naïve 95% confidence interval for calibrated estimates. b Weight (kg)/height (m)2. Open in new tab After up to 16 years of follow-up, a total of 6,621 incident cases of T2D and 5,802 cases of CVD were identified. In calibrated basic and multivariable ES models, TS intake was not associated with T2D risk (Table 3). The hazard ratio estimates for 20% increase in calibrated TS intake remained almost unchanged after adding BMI and waist circumference to the model. In the basic EP model, we detected a statistically significant 22% increase in T2D risk for 20% increase in calibrated TS intake (hazard ratio (HR) = 1.22, 95% confidence interval (CI): 1.09, 1.37). However, the hazard ratio was markedly attenuated toward null in multivariable model (HR = 1.00, 95% CI: 0.83, 1.18) and became lower than 1.0 when BMI and waist circumference were added to the model (Table 3). When using uncalibrated TS intake, we found a statistically significant decrease in T2D risk in multivariable ES (HR = 0.92, 95% CI: 0.90, 0.93) and EP (HR = 0.94, 95% CI: 0.93, 0.95) models. The risk estimates remained statistically significant after the models were adjusted for BMI and waist circumference. In sensitivity analysis, excluding participants with hypercholesterolemia, hypertension, history of CVD, or family history of diabetes did not appreciably change any of the findings (data not shown). Table 3. Hazard Ratios for Type 2 Diabetes for a 20% Increase in Calibrated and Uncalibrated Intakes of Total Sugars, From Baseline (1993–1998) Through September 30, 2010 (n = 75,320), Women’s Health Initiative Observational Studya,b Model . Calibrated Total Sugarsc . Uncalibrated Total Sugarsc . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Age- and energy-adjustedd 0.99 0.92, 1.07 1.22 1.09, 1.37 0.93 0.92, 0.95 0.94 0.93, 0.96 Multivariable 1e 0.94 0.76, 1.15 1.00 0.85, 1.18 0.92 0.90, 0.93 0.94 0.93, 0.95 Multivariable 2f 0.93 0.67, 1.31 0.94 0.87, 1.01 0.95 0.94, 0.97 0.96 0.95, 0.98 Model . Calibrated Total Sugarsc . Uncalibrated Total Sugarsc . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Age- and energy-adjustedd 0.99 0.92, 1.07 1.22 1.09, 1.37 0.93 0.92, 0.95 0.94 0.93, 0.96 Multivariable 1e 0.94 0.76, 1.15 1.00 0.85, 1.18 0.92 0.90, 0.93 0.94 0.93, 0.95 Multivariable 2f 0.93 0.67, 1.31 0.94 0.87, 1.01 0.95 0.94, 0.97 0.96 0.95, 0.98 Abbreviations: AREE, activity-related energy expenditure; CI, confidence interval; HR, hazard ratio; T2D, type 2 diabetes. a Findings from energy substitution and energy partition models. bn = 6,621 T2D cases. c Models with calibrated total sugars included calibrated estimates of energy intake and AREE, whereas models with uncalibrated total sugars included uncalibrated estimates of those exposures. d Cox models were stratified by 5-year age groups and adjusted for age as a continuous variable and energy intake (total energy intake in energy substitution models; nonsugars and nonalcohol energy in energy partition models). e Additionally adjusted for race and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), marital status (never married, divorced or separated, presently married or living as married, or widowed), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), and AREE. f Multivariable model 1 plus BMI plus waist circumference. Open in new tab Table 3. Hazard Ratios for Type 2 Diabetes for a 20% Increase in Calibrated and Uncalibrated Intakes of Total Sugars, From Baseline (1993–1998) Through September 30, 2010 (n = 75,320), Women’s Health Initiative Observational Studya,b Model . Calibrated Total Sugarsc . Uncalibrated Total Sugarsc . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Age- and energy-adjustedd 0.99 0.92, 1.07 1.22 1.09, 1.37 0.93 0.92, 0.95 0.94 0.93, 0.96 Multivariable 1e 0.94 0.76, 1.15 1.00 0.85, 1.18 0.92 0.90, 0.93 0.94 0.93, 0.95 Multivariable 2f 0.93 0.67, 1.31 0.94 0.87, 1.01 0.95 0.94, 0.97 0.96 0.95, 0.98 Model . Calibrated Total Sugarsc . Uncalibrated Total Sugarsc . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Age- and energy-adjustedd 0.99 0.92, 1.07 1.22 1.09, 1.37 0.93 0.92, 0.95 0.94 0.93, 0.96 Multivariable 1e 0.94 0.76, 1.15 1.00 0.85, 1.18 0.92 0.90, 0.93 0.94 0.93, 0.95 Multivariable 2f 0.93 0.67, 1.31 0.94 0.87, 1.01 0.95 0.94, 0.97 0.96 0.95, 0.98 Abbreviations: AREE, activity-related energy expenditure; CI, confidence interval; HR, hazard ratio; T2D, type 2 diabetes. a Findings from energy substitution and energy partition models. bn = 6,621 T2D cases. c Models with calibrated total sugars included calibrated estimates of energy intake and AREE, whereas models with uncalibrated total sugars included uncalibrated estimates of those exposures. d Cox models were stratified by 5-year age groups and adjusted for age as a continuous variable and energy intake (total energy intake in energy substitution models; nonsugars and nonalcohol energy in energy partition models). e Additionally adjusted for race and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), marital status (never married, divorced or separated, presently married or living as married, or widowed), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), and AREE. f Multivariable model 1 plus BMI plus waist circumference. Open in new tab In basic and multivariable ES models, no association between calibrated TS intake and total CVD, CHD, or stroke was detected (Table 4). In multivariable EP models, we found an inverse association between calibrated TS intake and total CVD (per each 20% increase, HR = 0.90, 95% CI: 0.84, 0.97) and CHD risk (HR = 0.89, 95% CI: 0.81, 0.96) only after adjusting for BMI. With regard to other CVD outcomes, we observed inverse association with coronary death and heart failure in basic ES models, and this association became attenuated toward null in multivariable models (Web Table 1). In basic EP models, a 20% increase in TS intake was associated with a statistically significant increase in risk for coronary artery bypass graft (HR = 1.14, 95% CI: 1.02, 1.27). This association dissipated after adding other covariates (multivariable 1), and become significantly inverse with BMI in the model (multivariable 2). Statistically significant inverse association was also observed for nonfatal myocardial infarction and percutaneous coronary intervention, which remained unchanged after adjusting for BMI. For uncalibrated sugars, we found a weak inverse association for several CVD outcomes in the multivariable ES and EP models (Table 4, Web Table 1). Excluding CVD cases diagnosed within the first year of follow-up did not appreciably change any of the findings (data not shown). Table 4. Hazard Ratios for Cardiovascular Disease for 20% Increase of Calibrated and Uncalibrated Intakes of Total Sugars From Energy Substitution and Energy Partition Models, From Baseline (1993–1998) Through September 30, 2010 (n = 64,751), Women’s Health Initiative Observational Study Model . Calibrated Total Sugarsa . Uncalibrated Total Sugarsa . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Total CVDb  Age- and energy-adjustedc 0.98 0.94, 1.03 1.03 0.95, 1.12 0.96 0.94, 0.97 0.96 0.95, 0.98  Multivariable 1d 0.97 0.87, 1.09 0.91 0.80, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2e 0.97 0.85, 1.12 0.90 0.84, 0.97 0.98 0.96, 1.00 0.98 0.97, 1.00 Total CHDf  Age- and energy-adjusted 0.99 0.94, 1.04 1.05 0.95, 1.15 0.95 0.93, 0.97 0.96 0.94, 0.97  Multivariable 1 0.96 0.86, 1.07 0.90 0.78, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2 0.96 0.83, 1.11 0.89 0.81, 0.96 0.97 0.95, 1.00 0.98 0.96, 1.00 Total strokeg  Age- and energy-adjusted 0.96 0.92, 1.01 0.98 0.91, 1.05 0.98 0.95, 1.01 0.98 0.96, 1.00  Multivariable 1 1.00 0.85, 1.18 0.97 0.85, 1.10 0.99 0.95, 1.03 0.99 0.96, 1.02  Multivariable 2 1.00 0.84, 1.20 0.95 0.86, 1.06 1.00 0.96, 1.03 0.99 0.96, 1.02 Model . Calibrated Total Sugarsa . Uncalibrated Total Sugarsa . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Total CVDb  Age- and energy-adjustedc 0.98 0.94, 1.03 1.03 0.95, 1.12 0.96 0.94, 0.97 0.96 0.95, 0.98  Multivariable 1d 0.97 0.87, 1.09 0.91 0.80, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2e 0.97 0.85, 1.12 0.90 0.84, 0.97 0.98 0.96, 1.00 0.98 0.97, 1.00 Total CHDf  Age- and energy-adjusted 0.99 0.94, 1.04 1.05 0.95, 1.15 0.95 0.93, 0.97 0.96 0.94, 0.97  Multivariable 1 0.96 0.86, 1.07 0.90 0.78, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2 0.96 0.83, 1.11 0.89 0.81, 0.96 0.97 0.95, 1.00 0.98 0.96, 1.00 Total strokeg  Age- and energy-adjusted 0.96 0.92, 1.01 0.98 0.91, 1.05 0.98 0.95, 1.01 0.98 0.96, 1.00  Multivariable 1 1.00 0.85, 1.18 0.97 0.85, 1.10 0.99 0.95, 1.03 0.99 0.96, 1.02  Multivariable 2 1.00 0.84, 1.20 0.95 0.86, 1.06 1.00 0.96, 1.03 0.99 0.96, 1.02 Abbreviations: AREE, activity-related energy expenditure; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio. a Models with calibrated total sugars included calibrated estimates of energy, protein, and ratio of sodium to potassium intake, and AREE, whereas models with uncalibrated total sugars included uncalibrated estimates of those exposures. bn = 5,802. c Cox models were stratified by 5-year age groups and adjusted for age as a continuous variable and energy intake (total energy intake in energy substitution models; nonsugars and nonalcohol energy in energy partition models). d Additionally adjusted for race and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), AREE, and ratio of sodium-to-potassium intake. e Multivariable 1 plus BMI. fn = 4,291. gn = 1,868. Open in new tab Table 4. Hazard Ratios for Cardiovascular Disease for 20% Increase of Calibrated and Uncalibrated Intakes of Total Sugars From Energy Substitution and Energy Partition Models, From Baseline (1993–1998) Through September 30, 2010 (n = 64,751), Women’s Health Initiative Observational Study Model . Calibrated Total Sugarsa . Uncalibrated Total Sugarsa . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Total CVDb  Age- and energy-adjustedc 0.98 0.94, 1.03 1.03 0.95, 1.12 0.96 0.94, 0.97 0.96 0.95, 0.98  Multivariable 1d 0.97 0.87, 1.09 0.91 0.80, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2e 0.97 0.85, 1.12 0.90 0.84, 0.97 0.98 0.96, 1.00 0.98 0.97, 1.00 Total CHDf  Age- and energy-adjusted 0.99 0.94, 1.04 1.05 0.95, 1.15 0.95 0.93, 0.97 0.96 0.94, 0.97  Multivariable 1 0.96 0.86, 1.07 0.90 0.78, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2 0.96 0.83, 1.11 0.89 0.81, 0.96 0.97 0.95, 1.00 0.98 0.96, 1.00 Total strokeg  Age- and energy-adjusted 0.96 0.92, 1.01 0.98 0.91, 1.05 0.98 0.95, 1.01 0.98 0.96, 1.00  Multivariable 1 1.00 0.85, 1.18 0.97 0.85, 1.10 0.99 0.95, 1.03 0.99 0.96, 1.02  Multivariable 2 1.00 0.84, 1.20 0.95 0.86, 1.06 1.00 0.96, 1.03 0.99 0.96, 1.02 Model . Calibrated Total Sugarsa . Uncalibrated Total Sugarsa . Energy Substitution . Energy Partition . Energy Substitution . Energy Partition . HR . 95% CI . HR . 95% CI . HR . 95% CI . HR . 95% CI . Total CVDb  Age- and energy-adjustedc 0.98 0.94, 1.03 1.03 0.95, 1.12 0.96 0.94, 0.97 0.96 0.95, 0.98  Multivariable 1d 0.97 0.87, 1.09 0.91 0.80, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2e 0.97 0.85, 1.12 0.90 0.84, 0.97 0.98 0.96, 1.00 0.98 0.97, 1.00 Total CHDf  Age- and energy-adjusted 0.99 0.94, 1.04 1.05 0.95, 1.15 0.95 0.93, 0.97 0.96 0.94, 0.97  Multivariable 1 0.96 0.86, 1.07 0.90 0.78, 1.04 0.97 0.95, 0.99 0.98 0.96, 0.99  Multivariable 2 0.96 0.83, 1.11 0.89 0.81, 0.96 0.97 0.95, 1.00 0.98 0.96, 1.00 Total strokeg  Age- and energy-adjusted 0.96 0.92, 1.01 0.98 0.91, 1.05 0.98 0.95, 1.01 0.98 0.96, 1.00  Multivariable 1 1.00 0.85, 1.18 0.97 0.85, 1.10 0.99 0.95, 1.03 0.99 0.96, 1.02  Multivariable 2 1.00 0.84, 1.20 0.95 0.86, 1.06 1.00 0.96, 1.03 0.99 0.96, 1.02 Abbreviations: AREE, activity-related energy expenditure; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio. a Models with calibrated total sugars included calibrated estimates of energy, protein, and ratio of sodium to potassium intake, and AREE, whereas models with uncalibrated total sugars included uncalibrated estimates of those exposures. bn = 5,802. c Cox models were stratified by 5-year age groups and adjusted for age as a continuous variable and energy intake (total energy intake in energy substitution models; nonsugars and nonalcohol energy in energy partition models). d Additionally adjusted for race and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), AREE, and ratio of sodium-to-potassium intake. e Multivariable 1 plus BMI. fn = 4,291. gn = 1,868. Open in new tab In Table 5, we report hazard ratio estimates for T2D and CVD from EP models with calibrated TS by BMI category. No association was found between calibrated TS intake and T2D risk in any of the BMI strata. There was no association between calibrated TS intake and CVD risk among normal-weight and obese participants; there was some evidence of an inverse association among overweight women only (for total CVD, per each 20% increase, HR = 0.90, 95% CI: 0.81, 1.01; for total CHD, HR = 0.87, 95% CI: 0.76, 0.99). Table 5. Multivariable Hazard Ratios for Total Cardiovascular Disease, Coronary Heart Disease, and Type 2 Diabetes for a 20% Increase in Calibrated Total Sugars From Energy Partition Models, by Body Mass Index Category, From Baseline (1993–1998) Through September 30, 2010, Women’s Health Initiative Observational Study Disease Outcome . Total No. . BMIa . <25.0b . 25.0–29.9b . ≥30.0b . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . Total CVDc 64,751 1,986 0.95 0.82, 1.11 2,064 0.90 0.81, 1.01 1,752 0.95 0.86, 1.07 CHDc 64,751 1,416 0.93 0.77, 1.11 1,511 0.87 0.76, 0.99 1,364 0.95 0.82, 1.10 T2Dd 75,320 1,318 0.91 0.79, 1.04 2,126 0.90 0.78, 1.04 3,177 0.94 0.86, 1.03 Disease Outcome . Total No. . BMIa . <25.0b . 25.0–29.9b . ≥30.0b . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . Total CVDc 64,751 1,986 0.95 0.82, 1.11 2,064 0.90 0.81, 1.01 1,752 0.95 0.86, 1.07 CHDc 64,751 1,416 0.93 0.77, 1.11 1,511 0.87 0.76, 0.99 1,364 0.95 0.82, 1.10 T2Dd 75,320 1,318 0.91 0.79, 1.04 2,126 0.90 0.78, 1.04 3,177 0.94 0.86, 1.03 Abbreviations: AREE, activity-related energy expenditure; BMI, body mass index; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; T2D, type 2 diabetes. a Body weight (kg)/height (m)2. b Total number of participants by BMI category: BMI <25.0: CVD/CHD cohort, n = 27,396, T2D cohort, n = 32,093; BMI = 25.0–29.9: CVD/CHD cohort, n = 21,806, T2D cohort, n = 25,379; BMI ≥30.0: CVD/CHD cohort, n = 15,549, T2D cohort, n = 17,848. c Multivariable models were stratified by 5-year age groups and adjusted for age as a continuous variable, calibrated nonsugars and nonalcohol energy (kcal/day), race and ethnicity (white, black, Hispanic, or other races), educational level (high school or less, more than high school, or college degree or higher), smoking status (never, past smoker, or current smoker), history of treated hypertension (yes or no), treated hypercholesterolemia (yes or no), family history of CVD (yes or no), hormone therapy use (never, estrogen alone, or estrogen plus progestin), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), calibrated AREE, and calibrated ratio of sodium to potassium intake. d Multivariable models were stratified by 5-year age groups and adjusted for age as continuous variable, calibrated nonsugars and nonalcohol energy (kcal/day), ace and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), marital status (never married, divorced or separated, presently married or living as married, and widowed), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), calibrated AREE, and calibrated protein intake (g/day). Open in new tab Table 5. Multivariable Hazard Ratios for Total Cardiovascular Disease, Coronary Heart Disease, and Type 2 Diabetes for a 20% Increase in Calibrated Total Sugars From Energy Partition Models, by Body Mass Index Category, From Baseline (1993–1998) Through September 30, 2010, Women’s Health Initiative Observational Study Disease Outcome . Total No. . BMIa . <25.0b . 25.0–29.9b . ≥30.0b . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . Total CVDc 64,751 1,986 0.95 0.82, 1.11 2,064 0.90 0.81, 1.01 1,752 0.95 0.86, 1.07 CHDc 64,751 1,416 0.93 0.77, 1.11 1,511 0.87 0.76, 0.99 1,364 0.95 0.82, 1.10 T2Dd 75,320 1,318 0.91 0.79, 1.04 2,126 0.90 0.78, 1.04 3,177 0.94 0.86, 1.03 Disease Outcome . Total No. . BMIa . <25.0b . 25.0–29.9b . ≥30.0b . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . No. of Cases . HR . 95% CI . Total CVDc 64,751 1,986 0.95 0.82, 1.11 2,064 0.90 0.81, 1.01 1,752 0.95 0.86, 1.07 CHDc 64,751 1,416 0.93 0.77, 1.11 1,511 0.87 0.76, 0.99 1,364 0.95 0.82, 1.10 T2Dd 75,320 1,318 0.91 0.79, 1.04 2,126 0.90 0.78, 1.04 3,177 0.94 0.86, 1.03 Abbreviations: AREE, activity-related energy expenditure; BMI, body mass index; CHD, coronary heart disease; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; T2D, type 2 diabetes. a Body weight (kg)/height (m)2. b Total number of participants by BMI category: BMI <25.0: CVD/CHD cohort, n = 27,396, T2D cohort, n = 32,093; BMI = 25.0–29.9: CVD/CHD cohort, n = 21,806, T2D cohort, n = 25,379; BMI ≥30.0: CVD/CHD cohort, n = 15,549, T2D cohort, n = 17,848. c Multivariable models were stratified by 5-year age groups and adjusted for age as a continuous variable, calibrated nonsugars and nonalcohol energy (kcal/day), race and ethnicity (white, black, Hispanic, or other races), educational level (high school or less, more than high school, or college degree or higher), smoking status (never, past smoker, or current smoker), history of treated hypertension (yes or no), treated hypercholesterolemia (yes or no), family history of CVD (yes or no), hormone therapy use (never, estrogen alone, or estrogen plus progestin), alcohol consumption (never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), calibrated AREE, and calibrated ratio of sodium to potassium intake. d Multivariable models were stratified by 5-year age groups and adjusted for age as continuous variable, calibrated nonsugars and nonalcohol energy (kcal/day), ace and ethnicity (white, black, Hispanic, American Indians, Asian/Pacific Islanders, or other or unknown), marital status (never married, divorced or separated, presently married or living as married, and widowed), educational level (0–8 years, some high school, high school diploma or General Educational Development diploma, school after high school, or college degree or higher), smoking status (never, past smoker, and current smoker), hormone therapy use (never, estrogen alone, and estrogen plus progestin user), history of treated hypertension (yes or no), history of cardiovascular disease (yes or no), family history of T2D (yes or no), history of treated hypercholesterolemia (yes or no), alcohol consumption never drinker, past drinker, <1 per month, 1–3 per month, 1–6 per week, and ≥7 per week), calibrated AREE, and calibrated protein intake (g/day). Open in new tab DISCUSSION In this analysis in which biomarker-based ME correction was applied to self-reports of TS and energy by WHI OS participants, we found no statistically significant association between TS intake and either T2D or CVD risk. In contrast, analyses with uncalibrated exposures appeared to generate an inverse association with T2D risk and some evidence of inverse association with CVD risk. In a meta-analysis of 12 prospective cohort studies, no association was found between self-reported TS intake and T2D risk with a pooled relative risk of 0.91 (95% CI: 0.76, 1.09) for participants with highest versus lowest level of intake (11). Only 1 among these cohorts reported an inverse association with T2D risk (49), whereas no association was reported by others (12, 50–53). Similar to our study, Ahmadi-Abhari et al. (50) used ES and EP models, and with neither model did they observe an association between TS consumption and T2D risk. Although we included calibrated protein in our analyses, the lack of biomarkers for fat or complex carbohydrates prevented us from exploring any potential confounding effect from the latter 2 macronutrients, because combining calibrated with uncalibrated energy sources would not have allowed for a meaningful interpretation. In our uncalibrated analyses, TS intake was inversely associated with T2D risk across all the models, whereas this association was no longer evident after calibration. Moreover, there was a statistically significant increase in T2D risk in the model testing the association when adding calibrated TS intake while keeping nonsugars energy constant (EP basic model), though this association was attenuated in multivariable models, and especially when BMI and waist circumference were added to the model. Yet, that, in contrast, we observed no association in the ES models suggests the association observed in the EP model was mediated by energy and, similar to other energy sources, TS intake may be a risk factor for T2D. In our cohort, family history of T2D and personal history of hypercholesterolemia are strong correlates of BMI (a potential mediator in the observed association), which may have led to underestimation of the association between sugars intake and T2D in the multivariable model (without BMI) due to possible overadjustment. An interesting observation was the opposite direction in the association of calibrated compared with uncalibrated TS consumption with T2D in the basic EP models, which suggests ME does not always lead to attenuation but can even cause a change of direction in the association. We observed no association between TS intake and CVD risk in ES models, which suggests other energy sources may be equally important in relation to CVD risk. There was some evidence of a weak inverse association with total CVD and CHD in EP models confined to overweight women only, which may have been due to confounding from other nutrients derived from nutrient-dense foods high in naturally occurring sugars (e.g., fruits, vegetables). The difference in hazard ratio estimates derived from uncalibrated versus calibrated basic EP models implies that increased sugars intake along with increases in energy may increase CVD risk (e.g., coronary artery bypass graft), yet these associations were no longer evident when other confounders were included in the models. In 2 studies, among few prospective studies of European populations, no association was found between sugars intake and total CVD (7) or CHD risk (18), in fully adjusted ES models with BMI. In 1 cohort, borderline increased risk was found for CHD risk (per 29.5 g TS, HR = 1.15, 95% CI: 0.97, 1.36) among men, no association was found in women, and no association with stroke risk was found in either sex (19). A major strength of our analysis was the use of biomarker-based, ME-corrected estimates of self-reported TS intake and other exposures, which dampens the ME in the main exposure and the effect of residual confounding from important, poorly measured confounders. The prospective design of our analysis prevented recall bias and limited the potential for reverse causality. We explored association using 2 energy adjustment models, which allowed investigation of the association between TS intake and outcomes when substituting TS for isocaloric amount of other macronutrients (ES model), and when increasing TS intake while keeping the amount of other macronutrients fixed (EP model). The hazard ratios were estimated for a 20% increase in TS consumption, which translates into modest changes in diet (i.e., 18 g/1,000 kcal); however, this would mean that even small hazard ratio estimates would still be important at a population level. We report findings from different models on multiple outcomes, thus some of the findings may have occurred due to a chance. Yet, this is an exploratory, rather than confirmatory, analysis of the effect of ME on the investigated associations. We acknowledge that some selection bias may have occurred, if data were not missing completely at random, which is unlikely, however, given the prospective nature of the analysis. Although we used ME-corrected estimates for some important confounders, we still lacked calibrated intake for other nutrients for which no biomarkers were available (e.g., fat, dietary fiber), hence we could not control for those. The original calibration equation for TS developed in NPAAS explained only small proportion of variation in “true” sugars intake (6%–18% for absolute TS and 29%–40% for TS density) (3), possibly resulting in incomplete ME correction. In addition, the relatively small size of NPAAS decreased the precision of the risk estimates. The TS biomarker was developed in 2 highly controlled feeding studies conducted in the United Kingdom (1, 54). Although the biomarker was sensitive to intake, and had good reproducibility and high predictive potential, its biases were estimated on the basis of 13 participants consuming their usual diet under controlled conditions in a UK-based study (2). In this application, therefore, we assumed that the biomarker’s biases do not substantially change from 1 population to another, thus the equation for biomarker correction or calibration is transferrable and can be applied to a US population (2, 3). This assumption has yet to be investigated under controlled conditions (55). Energy intake was a strong risk factor for T2D and CVD in this cohort when using calibrated energy (per 20% increase, for T2D, multivariate HR = 4.17, 95% CI: 2.68, 6.49; for total CVD, HR = 1.49, 95% CI: 1.23, 1.81) but not when using uncalibrated intake (for T2D, HR = 1.06, 95% CI: 1.04, 1.07; for total CVD, HR = 1.00, 95% CI: 0.99, 1.01) (32). Hence, sugars could be contributing to disease risk through provision of unnecessary energy, as suggested from our basic EP models. The lack of associations in multivariable models may be due to incomplete ME correction, and population-specific biomarker calibration equations may be needed to correct for biomarker’s biases. Finally, the biomarker measures total, rather than added sugars; hence, the negative confounding from beneficial micronutrients and bioactive compounds derived from fruit and vegetables, sources of naturally occurring sugars is very likely, and may have counterbalanced the influence of TS intake per se. Furthermore, sugars encapsulated within the food cellular structure may have different metabolic effects than sugars in processed foods high in energy density and depleted of micronutrients (22, 56). In conclusion, using biomarker-based calibrated intake estimates, no association was observed between TS intake and either T2D or CVD risk in the postmenopausal women in this study, though we cannot rule out that sugars could be contributing to T2D and CVD risk through provision of excess energy. Low explanatory power of the calibration equations for TS intake may have led to incomplete ME correction and incomplete deattenuation of the risk estimates. Additional research on the performance of the sugars biomarker in the US population is needed to better characterize its use and to verify the calibration equations applied here. ACKNOWLEDGMENTS Author affiliations: School of Nutrition and Health Promotion, College of Health Solutions, Arizona State University, Phoenix, Arizona (Natasha Tasevska); Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington (Mary Pettinger, Lesley F. Tinker, Marian L. Neuhouser, Ross L. Prentice); Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland (Victor Kipnis, Douglas Midthune); Population Studies Program, National Institutes of Health Office of Dietary Supplements, Bethesda, Maryland (Nancy Potischman); Division of General Internal Medicine and Clinical Innovation, New York University School of Medicine, New York, New York (Jeannette M. Beasley); Department of Preventive Medicine, Northwestern University, Chicago, Illinois (Linda Van Horn); Medstar Health Research Institute, Hyattsville, Maryland (Barbara V. Howard); Brown University, Providence, Rhode Island (Simin Liu); Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts (JoAnn E. Manson); University of Alabama at Birmingham, Birmingham, Alabama (James M. Shikany); and Department of Health Promotion Sciences, Mel and Enid Zuckerman College of Public Health, University of Arizona, Tucson, Arizona (Cynthia A. Thomson). This work was supported by the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health, US Department of Health and Human Services (contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C) and the National Cancer Institute (grant R01 CA119171). We acknowledge the following investigators in the Women’s Health Initiative (WHI) Program: Program Office at the National Heart, Lung, and Blood Institute (NHLBI), Bethesda, Maryland: Jacques Rossouw, Shari Ludlam, Dale Burwen, Joan McGowan, Leslie Ford, and Nancy Geller; Clinical Coordinating Center at Fred Hutchinson Cancer Research Center, Seattle, Washington: Garnet Anderson, Ross Prentice, Andrea LaCroix, and Charles Kooperberg. We thank the following investigators and academic centers: Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts: JoAnn E. Manson; MedStar Health Research Institute and Howard University, Washington, DC: Barbara V. Howard; Stanford Prevention Research Center, Stanford, California: Marcia L. Stefanick; The Ohio State University, Columbus, Ohio: Rebecca Jackson; the Tucson and Phoenix, Arizona, campuses of University of Arizona: Cynthia A. Thomson; University at Buffalo, Buffalo, New York: Jean Wactawski-Wende; the Gainesville and Jacksonville, Florida, campuses of University of Florida: Marian Limacher; the Iowa City and Davenport, Iowa, campuses of University of Iowa: Jennifer Robinson; University of Pittsburgh, Pittsburgh, Pennsylvania: Lewis Kuller; Wake Forest University School of Medicine, Winston-Salem, North Carolina: Sally Shumaker; University of Nevada, Reno, Nevada: Robert Brunner; University of Minnesota, Minneapolis, Minnesota: Karen L. Margolis; and with the WHI Memory Study at Wake Forest University School of Medicine, Winston-Salem, North Carolina: Mark Espeland. For a list of all the investigators who have contributed to WHI science, please visit: https://www.whi.org/researchers/Documents%20%20Write%20a%20Paper/WHI%20Investigator%20Long%20List.pdf. Decisions concerning study design, data collection and analysis, interpretation of the results, the preparation of the manuscript, and the decision to submit the manuscript for publication resided with committees comprising WHI investigators, who included NHLBI representatives. The contents of the paper are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health. Conflict of interest: none declared. Abbreviations BMI body mass index CHD coronary heart disease CI confidence interval CVD cardiovascular disease EP energy partition ES energy substitution HR hazard ratio ME measurement error NPAAS Nutrition and Physical Activity Assessment Study OS Observational Study T2D type 2 diabetes TS total sugars WHI Women’s Health Initiative. REFERENCES 1 Tasevska N , Runswick SA, McTaggart A, et al. . Urinary sucrose and fructose as biomarkers for sugar consumption . Cancer Epidemiol Biomarkers Prev . 2005 ; 14 ( 5 ): 1287 – 1294 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Tasevska N , Midthune D, Potischman N, et al. . Use of the predictive sugars biomarker to evaluate self-reported total sugars intake in the Observing Protein and Energy Nutrition (OPEN) Study . Cancer Epidemiol Biomarkers Prev . 2011 ; 20 ( 3 ): 490 – 500 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Tasevska N , Midthune D, Tinker LF, et al. . Use of a urinary sugars biomarker to assess measurement error in self-reported sugars intake in the Nutrition and Physical Activity Assessment Study (NPAAS) . Cancer Epidemiol Biomarkers Prev . 2014 ; 23 ( 12 ): 2874 – 2883 . Google Scholar Crossref Search ADS PubMed WorldCat 4 GBD 2015 Disease and Injury Incidence and Prevalence Collaborators . Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015 . Lancet . 2016 ; 388 ( 10053 ): 1545 – 1602 . Crossref Search ADS PubMed WorldCat 5 Centers for Disease Control and Prevention . National Diabetes Statistics Report. Estimates of Diabetes and Its Burden in the United States. Atlanta, GA: Centers for Disease Control and Prevention; 2017 . (National Diabetes Statistics Report). https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf. Accessed May 23, 2018. 6 Benjamin EJ , Blaha MJ, Chiuve SE, et al. . Heart Disease and Stroke Statistics-2017 update: a report from the American Heart Association . Circulation . 2017 ; 135 ( 10 ): e146 – e603 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Beulens JW , de Bruijne LM, Stolk RP, et al. . High dietary glycemic load and glycemic index increase risk of cardiovascular disease among middle-aged women: a population-based follow-up study . J Am Coll Cardiol . 2007 ; 50 ( 1 ): 14 – 21 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Liu S , Willett WC, Stampfer MJ, et al. . A prospective study of dietary glycemic load, carbohydrate intake, and risk of coronary heart disease in US women . Am J Clin Nutr . 2000 ; 71 ( 6 ): 1455 – 1461 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Xi B , Huang Y, Reilly KH, et al. . Sugar-sweetened beverages and risk of hypertension and CVD: a dose-response meta-analysis . Br J Nutr . 2015 ; 113 ( 5 ): 709 – 717 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Imamura F , O’Connor L, Ye Z, et al. . Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction . BMJ . 2015 ; 351 : h3576 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Tsilas CS , de Souza RJ, Mejia SB, et al. . Relation of total sugars, fructose and sucrose with incident type 2 diabetes: a systematic review and meta-analysis of prospective cohort studies . CMAJ . 2017 ; 189 ( 20 ): E711 – E720 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Janket SJ , Manson JE, Sesso H, et al. . A prospective study of sugar intake and risk of type 2 diabetes in women . Diabetes Care . 2003 ; 26 ( 4 ): 1008 – 1015 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Gross LS , Li L, Ford ES, et al. . Increased consumption of refined carbohydrates and the epidemic of type 2 diabetes in the United States: an ecologic assessment . Am J Clin Nutr . 2004 ; 79 ( 5 ): 774 – 779 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Malik VS , Hu FB. Sweeteners and risk of obesity and type 2 diabetes: the role of sugar-sweetened beverages . Curr Diab Rep . 2012 ; 12 ( 2 ): 195 – 203 . Google Scholar Crossref Search ADS WorldCat 15 Fried SK , Rao SP. Sugars, hypertriglyceridemia, and cardiovascular disease . Am J Clin Nutr . 2003 ; 78 ( 4 ): 873S – 880S . Google Scholar Crossref Search ADS PubMed WorldCat 16 Nakagawa T , Hu H, Zharikov S, et al. . A causal role for uric acid in fructose-induced metabolic syndrome . Am J Physiol Renal Physiol . 2006 ; 290 ( 3 ): F625 – F631 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Stanhope KL . Sugar consumption, metabolic disease and obesity: the state of the controversy . Crit Rev Clin Lab Sci . 2016 ; 53 ( 1 ): 52 – 67 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Sieri S , Krogh V, Berrino F, et al. . Dietary glycemic load and index and risk of coronary heart disease in a large Italian cohort: the EPICOR Study . Arch Intern Med . 2010 ; 170 ( 7 ): 640 – 647 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Burger KN , Beulens JW, Boer JM, et al. . Dietary glycemic load and glycemic index and risk of coronary heart disease and stroke in Dutch men and women: the EPIC-MORGEN Study . PLoS One . 2011 ; 6 ( 10 ): e25955 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Sluijs I , van der Schouw YT, van der A DL, et al. . Carbohydrate quantity and quality and risk of type 2 diabetes in the European Prospective Investigation into Cancer and Nutrition-Netherlands (EPIC-NL) study . Am J Clin Nutr . 2010 ; 92 ( 4 ): 905 – 911 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Bowman SA . Added sugars: definition and estimation in the USDA Food Patterns Equivalents Databases . J Food Comp Anal . 2017 ; 64 ( 1 ): 64 – 67 . Google Scholar Crossref Search ADS WorldCat 22 Englyst KN , Englyst HN. Carbohydrate bioavailability . Br J Nutr . 2005 ; 94 ( 1 ): 1 – 11 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Almiron-Roig E , Palla L, Guest K, et al. . Factors that determine energy compensation: a systematic review of preload studies . Nutr Rev . 2013 ; 71 ( 7 ): 458 – 473 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Prentice RL , Mossavar-Rahmani Y, Huang Y, et al. . Evaluation and comparison of food records, recalls, and frequencies for energy and protein assessment by using recovery biomarkers . Am J Epidemiol . 2011 ; 174 ( 5 ): 591 – 603 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Schatzkin A , Kipnis V, Carroll RJ, et al. . A comparison of a food frequency questionnaire with a 24-hour recall for use in an epidemiological cohort study: results from the biomarker-based Observing Protein and Energy Nutrition (OPEN) Study . Int J Epidemiol . 2003 ; 32 ( 6 ): 1054 – 1062 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Neuhouser ML , Tinker L, Shaw PA, et al. . Use of recovery biomarkers to calibrate nutrient consumption self-reports in the Women’s Health Initiative . Am J Epidemiol . 2008 ; 167 ( 10 ): 1247 – 1259 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Beasley JM , Jung M, Tasevska N, et al. . Biomarker-predicted sugars intake compared with self-reported measures in US Hispanics/Latinos: results from the HCHS/SOL SOLNAS Study . Public Health Nutr . 2016 ; 19 ( 18 ): 3256 – 3264 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Freedman LS , Schatzkin A, Midthune D, et al. . Dealing with dietary measurement error in nutritional cohort studies . J Natl Cancer Inst . 2011 ; 103 ( 14 ): 1086 – 1092 . Google Scholar Crossref Search ADS PubMed WorldCat 29 The Women’s Health Initiative Study Group . Design of the Women’s Health Initiative Clinical Trial and Observational Study . Control Clin Trials . 1998 ; 19 ( 1 ): 61 – 109 . Crossref Search ADS PubMed WorldCat 30 Prentice RL , Pettinger M, Tinker LF, et al. . Regression calibration in nutritional epidemiology: example of fat density and total energy in relationship to postmenopausal breast cancer . Am J Epidemiol . 2013 ; 178 ( 11 ): 1663 – 1672 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Prentice RL , Shaw PA, Bingham SA, et al. . Biomarker-calibrated energy and protein consumption and increased cancer risk among postmenopausal women . Am J Epidemiol . 2009 ; 169 ( 8 ): 977 – 989 . Google Scholar Crossref Search ADS PubMed WorldCat 32 Zheng C , Beresford SA, Van Horn L, et al. . Simultaneous association of total energy consumption and activity-related energy expenditure with risks of cardiovascular disease, cancer, and diabetes among postmenopausal women . Am J Epidemiol . 2014 ; 180 ( 5 ): 526 – 535 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Prentice RL , Huang Y, Kuller LH, et al. . Biomarker-calibrated energy and protein consumption and cardiovascular disease risk among postmenopausal women . Epidemiology . 2011 ; 22 ( 2 ): 170 – 179 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Tinker LF , Sarto GE, Howard BV, et al. . Biomarker-calibrated dietary energy and protein intake associations with diabetes risk among postmenopausal women from the Women’s Health Initiative . Am J Clin Nutr . 2011 ; 94 ( 6 ): 1600 – 1606 . Google Scholar Crossref Search ADS PubMed WorldCat 35 Beasley JM , LaCroix AZ, Neuhouser ML, et al. . Protein intake and incident frailty in the Women’s Health Initiative observational study . J Am Geriatr Soc . 2010 ; 58 ( 6 ): 1063 – 1071 . Google Scholar Crossref Search ADS PubMed WorldCat 36 Kipnis V , Freedman LS, Brown CC, et al. . Interpretation of energy adjustment models for nutritional epidemiology . Am J Epidemiol . 1993 ; 137 ( 12 ): 1376 – 1380 . Google Scholar Crossref Search ADS PubMed WorldCat 37 Willett WC , Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic studies . Am J Clin Nutr . 1997 ; 65 ( 4 suppl ): 1220S – 1228S . Google Scholar Crossref Search ADS PubMed WorldCat 38 Langer RD , White E, Lewis CE, et al. . The Women’s Health Initiative Observational Study: baseline characteristics of participants and reliability of baseline measures . Ann Epidemiol . 2003 ; 13 ( 9 suppl ): S107 – S121 . Google Scholar Crossref Search ADS PubMed WorldCat 39 Hays J , Hunt JR, Hubbell FA, et al. . The Women’s Health Initiative recruitment methods and results . Ann Epidemiol . 2003 ; 13 ( 9 suppl ): S18 – S77 . Google Scholar Crossref Search ADS PubMed WorldCat 40 Patterson RE , Kristal AR, Tinker LF, et al. . Measurement characteristics of the Women’s Health Initiative food frequency questionnaire . Ann Epidemiol . 1999 ; 9 ( 3 ): 178 – 187 . Google Scholar Crossref Search ADS PubMed WorldCat 41 Johnson-Kozlow M , Rock CL, Gilpin EA, et al. . Validation of the WHI brief physical activity questionnaire among women diagnosed with breast cancer . Am J Health Behav . 2007 ; 31 ( 2 ): 193 – 202 . Google Scholar Crossref Search ADS PubMed WorldCat 42 Huang Y , Van Horn L, Tinker LF, et al. . Measurement error corrected sodium and potassium intake estimation using 24-hour urinary excretion . Hypertension . 2014 ; 63 ( 2 ): 238 – 244 . Google Scholar Crossref Search ADS PubMed WorldCat 43 Neuhouser ML , Di C, Tinker LF, et al. . Physical activity assessment: biomarkers and self-report of activity-related energy expenditure in the WHI . Am J Epidemiol . 2013 ; 177 ( 6 ): 576 – 585 . Google Scholar Crossref Search ADS PubMed WorldCat 44 Curb JD , McTiernan A, Heckbert SR, et al. . Outcomes ascertainment and adjudication methods in the Women’s Health Initiative . Ann Epidemiol . 2003 ; 13 ( 9 suppl ): S122 – S128 . Google Scholar Crossref Search ADS PubMed WorldCat 45 Jackson JM , DeFor TA, Crain AL, et al. . Validity of diabetes self-reports in the Women’s Health Initiative . Menopause . 2014 ; 21 ( 8 ): 861 – 868 . Google Scholar Crossref Search ADS PubMed WorldCat 46 Margolis KL , Qi L, Brzyski R, et al. . Validity of diabetes self-reports in the Women’s Health Initiative: comparison with medication inventories and fasting glucose measurements . Clin Trials . 2008 ; 5 ( 3 ): 240 – 247 . Google Scholar Crossref Search ADS PubMed WorldCat 47 Prentice R . Covariate measurement errors and parameter estimation in a failure time regression model . Biometrika . 1982 ; 69 ( 2 ): 331 – 342 . Google Scholar Crossref Search ADS WorldCat 48 Wang CY , Hsu L, Feng ZD, et al. . Regression calibration in failure time regression . Biometrics . 1997 ; 53 ( 1 ): 131 – 145 . Google Scholar Crossref Search ADS PubMed WorldCat 49 Hodge AM , English DR, O’Dea K, et al. . Glycemic index and dietary fiber and the risk of type 2 diabetes . Diabetes Care . 2004 ; 27 ( 11 ): 2701 – 2706 . Google Scholar Crossref Search ADS PubMed WorldCat 50 Ahmadi-Abhari S , Luben RN, Powell N, et al. . Dietary intake of carbohydrates and risk of type 2 diabetes: the European Prospective Investigation into Cancer-Norfolk Study . Br J Nutr . 2014 ; 111 ( 2 ): 342 – 352 . Google Scholar Crossref Search ADS PubMed WorldCat 51 Sluijs I , Beulens JW, van der Schouw YT, et al. . Dietary glycemic index, glycemic load, and digestible carbohydrate intake are not associated with risk of type 2 diabetes in eight European countries . J Nutr . 2013 ; 143 ( 1 ): 93 – 99 . Google Scholar Crossref Search ADS PubMed WorldCat 52 Montonen J , Järvinen R, Knekt P, et al. . Consumption of sweetened beverages and intakes of fructose and glucose predict type 2 diabetes occurrence . J Nutr . 2007 ; 137 ( 6 ): 1447 – 1454 . Google Scholar Crossref Search ADS PubMed WorldCat 53 Barclay AW , Flood VM, Rochtchina E, et al. . Glycemic index, dietary fiber, and risk of type 2 diabetes in a cohort of older Australians . Diabetes Care . 2007 ; 30 ( 11 ): 2811 – 2813 . Google Scholar Crossref Search ADS PubMed WorldCat 54 Tasevska N . Urinary sugars–a biomarker of total sugars intake . Nutrients . 2015 ; 7 ( 7 ): 5816 – 5833 . Google Scholar Crossref Search ADS PubMed WorldCat 55 Lampe JW , Huang Y, Neuhouser ML, et al. . Dietary biomarker evaluation in a controlled feeding study in women from the Women’s Health Initiative cohort . Am J Clin Nutr . 2017 ; 105 ( 2 ): 466 – 475 . Google Scholar Crossref Search ADS PubMed WorldCat 56 Johnson IT , Southgate DA, Durnin JV. Intrinsic and non-milk extrinsic sugars: does the distinction have analytical or physiological validity? Int J Food Sci Nutr . 1996 ; 47 ( 2 ): 131 – 140 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018.

Rivera, Anna C; Powell, Teresa M; Boyko, Edward J; Lee, Rachel U; Faix, Dennis J; Luxton, David D; Rull, Rudolph P; Millennium Cohort Study Team

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy112pmid: 29893775

Abstract Recent reports suggest US military service members who deployed in support of the recent conflicts in Iraq and Afghanistan have higher rates of new-onset asthma than those who did not deploy. However, it is unknown whether combat experiences, in addition to deployment, contribute to new-onset asthma risk. This study aimed to longitudinally determine the risk factors for developing asthma, including combat deployment (categorized as deployed with combat experience, deployed without combat experience, or nondeployed), among participants in the Millennium Cohort Study from 2001 to 2013. A total of 75,770 participants completed a baseline survey and at least 1 triennial follow-up survey on deployment experiences, lifestyle characteristics, and health outcomes. Complementary log-log models stratified by sex were used to estimate the relative risk of developing asthma among participants who reported no history of asthma at baseline. In models with adjustments, those who deployed with combat experience were 24%–30% more likely to develop asthma than those who did not deploy. Deployed personnel without combat experience were not at a higher risk for new-onset asthma compared with nondeployers. Further research is needed to identify specific features of combat that are associated with greater asthma risk to inform prevention strategies. asthma, combat disorders, longitudinal studies, military personnel, occupational exposure, respiratory system, veterans Over 2 million US service members have deployed in support of Operation Enduring Freedom (OEF), Operation Iraqi Freedom (OIF), and Operation New Dawn (1). Self-reported respiratory illnesses accounted for a large proportion of non-combat-related diagnoses, second only to diarrhea, in troops deployed in support of OEF/OIF (2). Asthma can have a negative impact on combat readiness by reducing service members’ ability to perform military duties. For example, symptoms may interfere with essential activities (such as wearing a protective mask), require medications and treatments that are limited in a combat environment, and contribute to absence due to the need for medical attention, including potential evacuation and redeployment (2–5). OEF, OIF, and Operation New Dawn are not the only conflicts during which respiratory complaints were reported by service members. Respiratory conditions have been a topic of interest among Gulf War veterans (6–13) and continue to be a relevant concern. Several studies have been conducted, with mixed results, to determine whether deployment in support of OEF/OIF was associated with new-onset asthma (3, 4, 13–23). Those who deployed in support of the recent conflicts in Iraq and Afghanistan were found to have a significantly higher risk of new-onset asthma than their stateside counterparts (4, 21). On the other hand, one study of military medical records reported that the incidence of asthma diagnosis decreased by 6.3 cases per 10,000 person-years from 2001 to 2013 in all branches of the military, regardless of deployment status (24). However, this study was limited to active-duty service members and may have missed asthma diagnoses among personnel in the Reserve and National Guard components, as well as veterans. Prospective epidemiologic studies investigating the incidence of asthma among US service members and veterans of the current conflicts in Iraq and Afghanistan are lacking. The Millennium Cohort Study is a longitudinal study, including participants from all branches of the US military who are active-duty and Reserve/National Guard personnel. The survey includes a wide range of outcomes, exposures, and covariates that other studies have not been able to account for. Leveraging this large longitudinal population study, the present study investigated the relationship between combat deployment in support of OEF/OIF and new-onset asthma over a 12-year follow-up period. METHODS Study population and survey methods The Millennium Cohort Study, the largest prospective cohort study in a military population, was launched in 2001 with the primary goal of evaluating the potential consequences that deployment and other military exposures may have on health (25, 26). Of the invited sample, the initial panel (Panel 1) enrolled 77,047 (35.9%) participants, who were randomly chosen from active US military rosters in October 2000, with oversampling of Reserve and National Guard personnel, women, and those who had been recently deployed. The second accession, Panel 2, enrolled 31,110 (25.3%) participants, who were randomly selected military personnel with 1–2 years of service as of October 2003. Panel 3, the third accession, enrolled 43,440 (28.2%) participants, who were randomly selected military personnel with 1–3 years of service as of October 2006. Panels 2 and 3 were oversampled for women and Marines. The Millennium Cohort Study survey consists of questions on lifestyle characteristics, military experiences, and health outcomes and behaviors. In addition to the baseline survey, participants are asked to complete a follow-up survey approximately every 3 years. Additional participant data were provided by the Defense Manpower Data Center, including sex, birth date, race/ethnicity, deployment in support of the recent operations in Iraq and Afghanistan, pay grade, service component, service branch, primary and duty occupations, date of separation from military service, and previous deployment experience to Southwest Asia, Bosnia, or Kosovo between 1998 and 2001 (27). A more detailed description of the study’s sampling and methodology has been described elsewhere (25, 26). All participants provided informed consent, and the study was approved by the Naval Health Research Center Institutional Review Board. Study design Our study included Millennium Cohort Study participants from Panels 1, 2, and 3, from 2001 to 2013, who completed the baseline survey and at least the first triennial follow-up survey, and who were not missing covariate, exposure, or outcome data. Participants who reported a prior diagnosis of asthma on the baseline survey were excluded from the study population. Participants remained in the study until their final survey, defined as the survey at which new-onset asthma was reported or the last survey completed—whichever came first. Fixed covariates were measured at baseline, and time-varying covariates were measured at the survey prior to the final survey. The main exposure, combat deployment, was measured over the entire follow-up period (from earliest available data until the follow-up survey prior to the final survey). The outcome of interest, new-onset asthma, was obtained from the Millennium Cohort Study survey. At baseline, asthma was assessed with the following yes/no question, with asthma listed as one of several health conditions: “Has your doctor or other health professional ever told you that you have any of the following conditions?” At triennial follow-up, a similar question was asked, but the time frame was restricted to “the last 3 years” as opposed to “ever.” New-onset asthma was defined by participants not reporting being diagnosed with asthma at baseline and then reporting being diagnosed with asthma in a subsequent survey. The exposure of interest was combat deployment in support of OEF/OIF during the entire follow-up period, categorized as nondeployed, deployed without combat, and deployed with combat. Combat deployment was assessed using a combination of combat experiences endorsed on the Millennium Cohort Study survey, and deployment dates were obtained from the Contingency Tracking System (CTS) database maintained by the Defense Manpower Data Center. Combat experience was determined by asking participants whether they were personally exposed to any of the following combat items (excluding television, video, movies, computers, or theater): witnessing death, witnessing physical abuse, dead and/or decomposing bodies, maimed soldiers or civilians, or prisoners of war or refugees. The CTS provided in- and out-of-theater dates in support of OEF/OIF since 2001. Deployments in support of OEF/OIF included deployments to Iraq and Afghanistan as well as non-combat zones (e.g., Japan and Germany) and sea locations (e.g., Persian Gulf). A deployment based on the CTS in conjunction with endorsement of experiencing combat was defined as “deployed with combat.” A deployment reported in the CTS in the absence of endorsement of any of the combat items was categorized as “deployed without combat.” Participants who did not deploy during the study period, according to CTS data, were categorized as “nondeployed.” Combat deployment was assessed at baseline and each follow-up prior to the final survey (follow-up period). Once a participant was categorized as deployed with combat, this status was retained for the entire follow-up period. Similarly, in the absence of being deployed with combat, being deployed without combat was carried forward. This carry-forward method was used to ensure that the association between exposure and new-onset asthma could be measured without restricting the exposure to only the 3 years prior to new-onset asthma. Additional covariates included demographic factors, military characteristics, smoking status, environmental exposures, enrollment panel, number of life stressor events, and posttraumatic stress disorder (PTSD) status. Demographic characteristics included birth year, sex, race/ethnicity, marital status, and education. Body mass index (BMI) was calculated as self-reported weight in kilograms divided by height in meters, squared (kg/m2). Service branch, service component, military occupation, military service status, and prior deployment completed the set of military characteristics. Smoking status was classified as nonsmoker, former smoker, or current smoker. Current smokers were those who reported smoking at least 100 cigarettes (5 packs) in their lifetime and had not quit successfully, while former smokers were those who reported smoking at least 100 cigarettes in their lifetime and successfully quit. Nonsmokers were those who answered “no” to smoking at least 100 cigarettes in their lifetime. Environmental exposures were ascertained from report of being personally exposed to: 1) occupational hazards requiring protective equipment, such as respirators or hearing protection; 2) routine skin contact with paint, solvent, and/or substances; 3) pesticides, including creams, sprays, or uniform treatments; and 4) pesticides applied in the environment or around living facilities in the last 3 years. Potential responses were “yes,” “no,” or “don’t know.” As a result of high collinearity, the pesticide questions (items 3 and 4) were combined into one variable. If a participant answered “yes” to either item, then “yes” was retained for the combined variable. If either item was answered as “don’t know,” the combined variable retained the “don’t know” response. Otherwise, the combined variable was set as “no”. Life stress events were assessed at baseline as ever experienced and at follow-up as experienced in the last 3 years. Life stress was categorized as no event, 1 event, or more than 1 event from the following list: divorce or separation, financial problems, sexual assault, sexual harassment, physical assault, or suffered a disabling illness or injury. PTSD status was based on the PTSD Checklist–Civilian Version, using the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, sensitive diagnostic criteria. In an analysis of Millennium Cohort Study data, the PTSD Checklist–Civilian Version had high internal consistency (Cronbach’s α = 0.94) (28). All covariates were assessed at baseline; BMI, military service status, smoking status, environmental exposures, number of stressful life events, and PTSD status were also updated at each available follow-up prior to the final survey. Additional deployment characteristics were assessed for those participants who had been deployed during the follow-up period. Multiple deployments was defined as having more than 1 deployment during the entire follow-up period, and duration of deployment was defined as the cumulative number of days deployed during the entire follow-up period. Statistical analysis Bivariate analyses, including χ2 tests of association, were performed to investigate the relationship between new-onset asthma and combat deployment as well as occupational, demographic, and behavioral risk factors. Analyses were stratified by sex because women are known to have a higher prevalence of adult asthma compared with men (29). Because new-onset asthma was reported at the time of follow-up survey administration rather than at the time of diagnosis, discrete-time survival analysis was used to investigate the association of combat deployment with new-onset asthma that occurred during each follow-up time interval. The analysis assumes an underlying continuous-time proportional hazards model, and a complementary log-log model was fitted. Time at risk was calculated from study entry (baseline survey completion) until the outcome was measured or the last completed survey—whichever came first. Relative risks were estimated, and 95% confidence intervals were reported. Collinearity was assessed using the variance inflation factor, where a value >4 indicates possible collinearity. All covariates remained in the model, regardless of significance, to facilitate comparisons between results among men and women. All 3 panels were pooled for these analyses. Additional analyses, restricted to those who deployed during the follow-up period, assessed whether new-onset asthma was associated with multiple deployments or the cumulative number of days deployed. Data management and statistical analyses were performed using SAS, version 9.3 (SAS Institute, Inc., Cary, North Carolina). RESULTS The eligible study population consisted of participants who submitted at least a baseline and first follow-up survey (n = 94,241) and screened negative for lifetime asthma diagnosis at baseline (n = 87,914). Participants whose surveys were missing asthma status at first follow-up (n = 2,668), had undetermined combat deployment status at baseline (n = 241), or had missing covariate data (n = 9,235) were excluded, resulting in a final study population of 75,770 participants. Population characteristics are shown in Table 1. During follow-up, 1,452 (2.7%) men and 1,055 (4.6%) women developed asthma, and those who were not deployed had the highest proportion of new-onset asthma (3.1% and 5.0%, respectively), followed by those who deployed with combat experience (2.7% and 4.4%, respectively) and those who deployed without combat experience (1.6% and 3.0%, respectively). Women consistently had higher rates of new-onset asthma than did men for all categories of characteristics listed in Table 1. Among both men and women, older participants, those who were divorced/widowed/separated, those with a higher BMI, those in the Army, those of enlisted rank, those with more than 1 stressful life event, those with PTSD, and those who did not know whether they had been personally exposed to occupational hazards requiring protective equipment had higher rates of new-onset asthma. Results also indicated higher incidence of new-onset asthma among those who self-reported routine skin contact with paints/solvents/substances and pesticides, including creams, sprays, and uniform treatments, and exposure to pesticides applied in the environment/surroundings (Table 1). Table 1. Descriptive Characteristics of the Study Population, Incidence of New-Onset Asthma, and Adjusted Relative Risk of Self-Reported New-Onset Asthma According to Sex, Millennium Cohort Study, United States, 2001–2013 Characteristic . Mena,b (n = 52,826) . Womena,c (n = 22,944) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Exposure of interest  Combat deployment during follow-upd   Not deployed 29,237 3.1 1.00 Referent 15,678 5.0 1.00 Referent   Deployed without combat experience 8,922 1.6 0.97 0.83, 1.15 3,577 3.0 0.97 0.80, 1.18   Deployed with combat experience 14,667 2.7 1.30 1.14, 1.47 3,689 4.4 1.24 1.05, 1.46 Demographic factors  Race/ethnicity   White, non-Hispanic 41,880 2.7 1.00 Referent 15,895 4.4 1.00 Referent   Black, non-Hispanic 4,484 3.1 1.01 0.85, 1.20 3,692 5.1 1.06 0.90, 1.24   Asian/Pacific Islander 1,899 2.1 0.91 0.67, 1.22 1,051 4.6 1.44 1.12, 1.86   Hispanic 3,414 3.7 1.28 1.08, 1.53 1,701 5.5 1.37 1.13, 1.66   Other 1,149 2.5 0.84 0.59, 1.20 605 5.1 1.15 0.82, 1.61  Birth year   1980 or later 14,454 2.0 1.00 Referent 9,435 3.4 1.00 Referent   1970–1979 15,078 2.8 0.94 0.78, 1.14 6,754 5.0 1.14 0.96, 1.36   Before 1970 23,294 3.2 0.96 0.77, 1.21 6,755 5.9 1.19 0.96, 1.48  Marital status   Never married 14,362 2.1 0.88 0.77, 1.01 9,257 4.4 1.13 0.99, 1.30   Married 34,164 3.0 1.00 Referent 9,841 4.3 1.00 Referent   Divorced/widowed/separated 4,300 3.2 0.89 0.75, 1.05 3,846 6.0 1.19 1.02, 1.38  Education   High school degree/GED or less 9,686 2.5 0.95 0.79, 1.15 3,296 4.1 0.9 0.71, 1.14   Some college or associate degree 26,452 3.0 1.02 0.88, 1.19 12,142 4.9 1.13 0.95, 1.34   Bachelor degree or higher 16,688 2.5 1.00 Referent 7,506 4.3 1.00 Referent  BMIe,f   Underweight/normal (<25.0) 14,872 2.2 1.00 Referent 13,182 4.1 1.00 Referent   Overweight (25.0–29.9) 28,516 2.8 1.30 1.14, 1.47 7,498 5.2 1.36 1.20, 1.53   Obese (≥30.0) 9,438 3.4 1.88 1.62, 2.18 2,264 5.6 1.68 1.42, 2.00  Enrollment panelg   2001–2003 32,852 3.2 1.00 Referent 10,878 5.8 1.00 Referent   2004–2006 8,016 2.6 0.97 0.81, 1.16 5,024 4.3 1.01 0.85, 1.20   2007–2008 11,958 1.7 0.74 0.59, 0.92 7,042 2.9 0.76 0.62, 0.95 Military service  Service branch   Army 23,206 3.5 1.00 Referent 10,464 5.5 1.00 Referent   Navy/Coast Guard 9,354 2.4 0.79 0.69, 0.91 4,309 4.4 0.96 0.82, 1.12   Marine Corps 4,877 2.0 0.68 0.55, 0.84 676 4.4 0.98 0.70, 1.39   Air Force 15,389 2.1 0.69 0.61, 0.79 7,495 3.5 0.79 0.69, 0.91  Service component   Active duty 32,986 2.7 1.00 Referent 13,354 4.5 1.00 Referent   Reserve/Guard 19,840 2.8 0.83 0.74, 0.93 9,590 4.8 0.93 0.82, 1.05  Occupation   Infantry, gun crews, seamen (combat specialist) 12,285 2.5 1.00 Referent 1,429 3.4 1.00 Referent   Electrical repair 5,703 2.4 1.11 0.92, 1.34 1,512 3.3 0.93 0.65, 1.33   Communication/intelligence 4,263 3.1 1.17 0.96, 1.43 1,951 5.4 1.56 1.14, 2.15   Health care 3,745 3.2 1.26 1.03, 1.55 5,002 4.8 1.41 1.05, 1.88   Other technical and specialty 1,475 3.3 1.54 1.18, 2.03 698 4.7 1.51 1.01, 2.25   Functional support 7,403 3.1 1.34 1.14, 1.58 6,664 4.7 1.21 0.91, 1.62   Electrical/mechanical equipment repair 9,041 2.7 1.17 0.99, 1.38 1,734 3.8 1.00 0.70, 1.42   Craft workers 1,878 2.7 1.26 0.96, 1.65 422 4.0 0.97 0.58, 1.62   Service and supply 4,503 3.2 1.30 1.08, 1.57 2,318 5.7 1.32 0.96, 1.81   Non-occupation 2,530 1.7 0.85 0.63, 1.16 1,214 4.5 1.63 1.14, 2.32  Pay grade   Enlisted 40,620 2.9 1.25 1.05, 1.49 17,984 4.8 1.14 0.93, 1.39   Officerh 12,206 2.3 1.00 Referent 4,960 4.0 1.00 Referent  Military service statusf   Currently serving 44,404 2.9 1.00 Referent 19,450 4.6 1.00 Referent   Former service member 8,422 1.9 1.00 0.87, 1.16 3,494 4.4 1.37 1.17, 1.59  Prior deploymenti   Yes 12,082 3.0 0.96 0.84, 1.09 1,429 5.6 0.98 0.79, 1.21   No 40,744 2.7 1.00 Referent 21,515 4.5 1.00 Referent Behavioral factors  Smoking statusf   Nonsmoker 30,367 2.7 1.00 Referent 14,553 4.3 1.00 Referent   Former smoker 13,540 2.8 0.93 0.83, 1.05 5,163 5.0 1.11 0.97, 1.27   Current smoker 8,919 2.9 1.01 0.88, 1.16 3,228 5.3 1.09 0.92, 1.28 Environmental exposures  Occupational hazards requiring protective equipmentf   Yes 27,680 3.0 0.97 0.86, 1.09 7,358 5.4 1.13 0.99, 1.29   No 24,104 2.5 1.00 Referent 15,139 4.2 1.00 Referent   Don’t know 1,042 3.4 1.09 0.80, 1.50 447 5.6 1.11 0.76, 1.63  Routine skin contact with paints/solvents/substancesf   Yes 13,011 3.4 1.11 0.98, 1.26 2,881 5.7 1.00 0.84, 1.19   No 37,463 2.5 1.00 Referent 19,271 4.4 1.00 Referent   Don’t know 2,352 3.7 1.17 0.93, 1.47 792 5.1 0.93) 0.68, 1.27  Pesticides, including creams, sprays, and uniform  treatments, and applied in the environment/surroundingsf   Yes 19,133 3.2 1.08 0.96, 1.21 6,238 5.7 1.18 1.03, 1.35   No 29,579 2.4 1.00 Referent 14,732 4.1 1.00 Referent   Don’t know 4,114 3.2 1.01 0.84, 1.21 1,974 5.0 0.99 0.80, 1.21 Stressors  Life stressor eventsf,j   None 40,568 2.2 1.00 Referent 14,243 3.5 1.00 Referent   1 event 9,401 4.2 1.31 1.16, 1.47 5,231 5.4 1.11 0.97, 1.27   >1 event 2,857 5.6 1.37 1.15, 1.62 3,470 8.0 1.31 1.12, 1.52  PTSDf,k   Yes 2,840 4.8 1.82 1.55, 2.15 1,506 7.8 1.56 1.30, 1.88   No 49,986 2.6 1.00 Referent 21,438 4.4 1.00 Referent Characteristic . Mena,b (n = 52,826) . Womena,c (n = 22,944) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Exposure of interest  Combat deployment during follow-upd   Not deployed 29,237 3.1 1.00 Referent 15,678 5.0 1.00 Referent   Deployed without combat experience 8,922 1.6 0.97 0.83, 1.15 3,577 3.0 0.97 0.80, 1.18   Deployed with combat experience 14,667 2.7 1.30 1.14, 1.47 3,689 4.4 1.24 1.05, 1.46 Demographic factors  Race/ethnicity   White, non-Hispanic 41,880 2.7 1.00 Referent 15,895 4.4 1.00 Referent   Black, non-Hispanic 4,484 3.1 1.01 0.85, 1.20 3,692 5.1 1.06 0.90, 1.24   Asian/Pacific Islander 1,899 2.1 0.91 0.67, 1.22 1,051 4.6 1.44 1.12, 1.86   Hispanic 3,414 3.7 1.28 1.08, 1.53 1,701 5.5 1.37 1.13, 1.66   Other 1,149 2.5 0.84 0.59, 1.20 605 5.1 1.15 0.82, 1.61  Birth year   1980 or later 14,454 2.0 1.00 Referent 9,435 3.4 1.00 Referent   1970–1979 15,078 2.8 0.94 0.78, 1.14 6,754 5.0 1.14 0.96, 1.36   Before 1970 23,294 3.2 0.96 0.77, 1.21 6,755 5.9 1.19 0.96, 1.48  Marital status   Never married 14,362 2.1 0.88 0.77, 1.01 9,257 4.4 1.13 0.99, 1.30   Married 34,164 3.0 1.00 Referent 9,841 4.3 1.00 Referent   Divorced/widowed/separated 4,300 3.2 0.89 0.75, 1.05 3,846 6.0 1.19 1.02, 1.38  Education   High school degree/GED or less 9,686 2.5 0.95 0.79, 1.15 3,296 4.1 0.9 0.71, 1.14   Some college or associate degree 26,452 3.0 1.02 0.88, 1.19 12,142 4.9 1.13 0.95, 1.34   Bachelor degree or higher 16,688 2.5 1.00 Referent 7,506 4.3 1.00 Referent  BMIe,f   Underweight/normal (<25.0) 14,872 2.2 1.00 Referent 13,182 4.1 1.00 Referent   Overweight (25.0–29.9) 28,516 2.8 1.30 1.14, 1.47 7,498 5.2 1.36 1.20, 1.53   Obese (≥30.0) 9,438 3.4 1.88 1.62, 2.18 2,264 5.6 1.68 1.42, 2.00  Enrollment panelg   2001–2003 32,852 3.2 1.00 Referent 10,878 5.8 1.00 Referent   2004–2006 8,016 2.6 0.97 0.81, 1.16 5,024 4.3 1.01 0.85, 1.20   2007–2008 11,958 1.7 0.74 0.59, 0.92 7,042 2.9 0.76 0.62, 0.95 Military service  Service branch   Army 23,206 3.5 1.00 Referent 10,464 5.5 1.00 Referent   Navy/Coast Guard 9,354 2.4 0.79 0.69, 0.91 4,309 4.4 0.96 0.82, 1.12   Marine Corps 4,877 2.0 0.68 0.55, 0.84 676 4.4 0.98 0.70, 1.39   Air Force 15,389 2.1 0.69 0.61, 0.79 7,495 3.5 0.79 0.69, 0.91  Service component   Active duty 32,986 2.7 1.00 Referent 13,354 4.5 1.00 Referent   Reserve/Guard 19,840 2.8 0.83 0.74, 0.93 9,590 4.8 0.93 0.82, 1.05  Occupation   Infantry, gun crews, seamen (combat specialist) 12,285 2.5 1.00 Referent 1,429 3.4 1.00 Referent   Electrical repair 5,703 2.4 1.11 0.92, 1.34 1,512 3.3 0.93 0.65, 1.33   Communication/intelligence 4,263 3.1 1.17 0.96, 1.43 1,951 5.4 1.56 1.14, 2.15   Health care 3,745 3.2 1.26 1.03, 1.55 5,002 4.8 1.41 1.05, 1.88   Other technical and specialty 1,475 3.3 1.54 1.18, 2.03 698 4.7 1.51 1.01, 2.25   Functional support 7,403 3.1 1.34 1.14, 1.58 6,664 4.7 1.21 0.91, 1.62   Electrical/mechanical equipment repair 9,041 2.7 1.17 0.99, 1.38 1,734 3.8 1.00 0.70, 1.42   Craft workers 1,878 2.7 1.26 0.96, 1.65 422 4.0 0.97 0.58, 1.62   Service and supply 4,503 3.2 1.30 1.08, 1.57 2,318 5.7 1.32 0.96, 1.81   Non-occupation 2,530 1.7 0.85 0.63, 1.16 1,214 4.5 1.63 1.14, 2.32  Pay grade   Enlisted 40,620 2.9 1.25 1.05, 1.49 17,984 4.8 1.14 0.93, 1.39   Officerh 12,206 2.3 1.00 Referent 4,960 4.0 1.00 Referent  Military service statusf   Currently serving 44,404 2.9 1.00 Referent 19,450 4.6 1.00 Referent   Former service member 8,422 1.9 1.00 0.87, 1.16 3,494 4.4 1.37 1.17, 1.59  Prior deploymenti   Yes 12,082 3.0 0.96 0.84, 1.09 1,429 5.6 0.98 0.79, 1.21   No 40,744 2.7 1.00 Referent 21,515 4.5 1.00 Referent Behavioral factors  Smoking statusf   Nonsmoker 30,367 2.7 1.00 Referent 14,553 4.3 1.00 Referent   Former smoker 13,540 2.8 0.93 0.83, 1.05 5,163 5.0 1.11 0.97, 1.27   Current smoker 8,919 2.9 1.01 0.88, 1.16 3,228 5.3 1.09 0.92, 1.28 Environmental exposures  Occupational hazards requiring protective equipmentf   Yes 27,680 3.0 0.97 0.86, 1.09 7,358 5.4 1.13 0.99, 1.29   No 24,104 2.5 1.00 Referent 15,139 4.2 1.00 Referent   Don’t know 1,042 3.4 1.09 0.80, 1.50 447 5.6 1.11 0.76, 1.63  Routine skin contact with paints/solvents/substancesf   Yes 13,011 3.4 1.11 0.98, 1.26 2,881 5.7 1.00 0.84, 1.19   No 37,463 2.5 1.00 Referent 19,271 4.4 1.00 Referent   Don’t know 2,352 3.7 1.17 0.93, 1.47 792 5.1 0.93) 0.68, 1.27  Pesticides, including creams, sprays, and uniform  treatments, and applied in the environment/surroundingsf   Yes 19,133 3.2 1.08 0.96, 1.21 6,238 5.7 1.18 1.03, 1.35   No 29,579 2.4 1.00 Referent 14,732 4.1 1.00 Referent   Don’t know 4,114 3.2 1.01 0.84, 1.21 1,974 5.0 0.99 0.80, 1.21 Stressors  Life stressor eventsf,j   None 40,568 2.2 1.00 Referent 14,243 3.5 1.00 Referent   1 event 9,401 4.2 1.31 1.16, 1.47 5,231 5.4 1.11 0.97, 1.27   >1 event 2,857 5.6 1.37 1.15, 1.62 3,470 8.0 1.31 1.12, 1.52  PTSDf,k   Yes 2,840 4.8 1.82 1.55, 2.15 1,506 7.8 1.56 1.30, 1.88   No 49,986 2.6 1.00 Referent 21,438 4.4 1.00 Referent Abbreviations: BMI, body mass index; CI, confidence interval; GED, General Educational Development certificate; PTSD, posttraumatic stress disorder; RR, relative risk. a Univariate analyses were performed separately for men and women. For both sexes, combat deployment, birth year, marital status, education, BMI, enrollment panel, service branch, occupation, pay grade, environmental exposures, life stressor events, and PTSD were significantly associated with new-onset asthma (α = 0.05). Models were fitted separately for men and women; both models adjusted for all variables in the table. b Among men, race/ethnicity and military service status were also significantly associated with new-onset asthma (α = 0.05). c Among women, smoking status was also significantly associated with new-onset asthma (α = 0.05). d Deployment was defined as being deployed in support of Operation Enduring Freedom or Operation Iraqi Freedom. Combat was defined as reporting personal exposure to ≥1 of the following: witnessing death, witnessing physical abuse, dead and/or decomposing bodies, maimed soldiers or civilians, or prisoners of war or refugees. Combat deployment was assessed over the entire follow-up period. e BMI was calculated as weight (kg) divided by height (m) squared. f Time-varying covariates were measured at the survey prior to the final survey (the survey at which new-onset asthma was reported or the last survey completed, whichever came first). g This study used 3 panels of Millennium Cohort Study participants. Participants completed their baseline survey during the listed years (Panel 1: 2001–2003, Panel 2: 2004–2006, Panel 3: 2007–2008). h Officer includes commissioned and warrant officers. i Prior deployments to Bosnia, Kosovo, or Southwest Asia between January 1, 1998, and September 1, 2001. j Categorized number of endorsements of the following events: divorce or separation, financial problems, sexual assault, sexual harassment, physical assault, or suffered a disabling illness or injury. k Based on the PTSD Checklist–Civilian Version, using sensitive diagnostic criteria from the Diagnostic and Statistical Manual of Mental Disorder, Fourth Edition. Open in new tab Table 1. Descriptive Characteristics of the Study Population, Incidence of New-Onset Asthma, and Adjusted Relative Risk of Self-Reported New-Onset Asthma According to Sex, Millennium Cohort Study, United States, 2001–2013 Characteristic . Mena,b (n = 52,826) . Womena,c (n = 22,944) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Exposure of interest  Combat deployment during follow-upd   Not deployed 29,237 3.1 1.00 Referent 15,678 5.0 1.00 Referent   Deployed without combat experience 8,922 1.6 0.97 0.83, 1.15 3,577 3.0 0.97 0.80, 1.18   Deployed with combat experience 14,667 2.7 1.30 1.14, 1.47 3,689 4.4 1.24 1.05, 1.46 Demographic factors  Race/ethnicity   White, non-Hispanic 41,880 2.7 1.00 Referent 15,895 4.4 1.00 Referent   Black, non-Hispanic 4,484 3.1 1.01 0.85, 1.20 3,692 5.1 1.06 0.90, 1.24   Asian/Pacific Islander 1,899 2.1 0.91 0.67, 1.22 1,051 4.6 1.44 1.12, 1.86   Hispanic 3,414 3.7 1.28 1.08, 1.53 1,701 5.5 1.37 1.13, 1.66   Other 1,149 2.5 0.84 0.59, 1.20 605 5.1 1.15 0.82, 1.61  Birth year   1980 or later 14,454 2.0 1.00 Referent 9,435 3.4 1.00 Referent   1970–1979 15,078 2.8 0.94 0.78, 1.14 6,754 5.0 1.14 0.96, 1.36   Before 1970 23,294 3.2 0.96 0.77, 1.21 6,755 5.9 1.19 0.96, 1.48  Marital status   Never married 14,362 2.1 0.88 0.77, 1.01 9,257 4.4 1.13 0.99, 1.30   Married 34,164 3.0 1.00 Referent 9,841 4.3 1.00 Referent   Divorced/widowed/separated 4,300 3.2 0.89 0.75, 1.05 3,846 6.0 1.19 1.02, 1.38  Education   High school degree/GED or less 9,686 2.5 0.95 0.79, 1.15 3,296 4.1 0.9 0.71, 1.14   Some college or associate degree 26,452 3.0 1.02 0.88, 1.19 12,142 4.9 1.13 0.95, 1.34   Bachelor degree or higher 16,688 2.5 1.00 Referent 7,506 4.3 1.00 Referent  BMIe,f   Underweight/normal (<25.0) 14,872 2.2 1.00 Referent 13,182 4.1 1.00 Referent   Overweight (25.0–29.9) 28,516 2.8 1.30 1.14, 1.47 7,498 5.2 1.36 1.20, 1.53   Obese (≥30.0) 9,438 3.4 1.88 1.62, 2.18 2,264 5.6 1.68 1.42, 2.00  Enrollment panelg   2001–2003 32,852 3.2 1.00 Referent 10,878 5.8 1.00 Referent   2004–2006 8,016 2.6 0.97 0.81, 1.16 5,024 4.3 1.01 0.85, 1.20   2007–2008 11,958 1.7 0.74 0.59, 0.92 7,042 2.9 0.76 0.62, 0.95 Military service  Service branch   Army 23,206 3.5 1.00 Referent 10,464 5.5 1.00 Referent   Navy/Coast Guard 9,354 2.4 0.79 0.69, 0.91 4,309 4.4 0.96 0.82, 1.12   Marine Corps 4,877 2.0 0.68 0.55, 0.84 676 4.4 0.98 0.70, 1.39   Air Force 15,389 2.1 0.69 0.61, 0.79 7,495 3.5 0.79 0.69, 0.91  Service component   Active duty 32,986 2.7 1.00 Referent 13,354 4.5 1.00 Referent   Reserve/Guard 19,840 2.8 0.83 0.74, 0.93 9,590 4.8 0.93 0.82, 1.05  Occupation   Infantry, gun crews, seamen (combat specialist) 12,285 2.5 1.00 Referent 1,429 3.4 1.00 Referent   Electrical repair 5,703 2.4 1.11 0.92, 1.34 1,512 3.3 0.93 0.65, 1.33   Communication/intelligence 4,263 3.1 1.17 0.96, 1.43 1,951 5.4 1.56 1.14, 2.15   Health care 3,745 3.2 1.26 1.03, 1.55 5,002 4.8 1.41 1.05, 1.88   Other technical and specialty 1,475 3.3 1.54 1.18, 2.03 698 4.7 1.51 1.01, 2.25   Functional support 7,403 3.1 1.34 1.14, 1.58 6,664 4.7 1.21 0.91, 1.62   Electrical/mechanical equipment repair 9,041 2.7 1.17 0.99, 1.38 1,734 3.8 1.00 0.70, 1.42   Craft workers 1,878 2.7 1.26 0.96, 1.65 422 4.0 0.97 0.58, 1.62   Service and supply 4,503 3.2 1.30 1.08, 1.57 2,318 5.7 1.32 0.96, 1.81   Non-occupation 2,530 1.7 0.85 0.63, 1.16 1,214 4.5 1.63 1.14, 2.32  Pay grade   Enlisted 40,620 2.9 1.25 1.05, 1.49 17,984 4.8 1.14 0.93, 1.39   Officerh 12,206 2.3 1.00 Referent 4,960 4.0 1.00 Referent  Military service statusf   Currently serving 44,404 2.9 1.00 Referent 19,450 4.6 1.00 Referent   Former service member 8,422 1.9 1.00 0.87, 1.16 3,494 4.4 1.37 1.17, 1.59  Prior deploymenti   Yes 12,082 3.0 0.96 0.84, 1.09 1,429 5.6 0.98 0.79, 1.21   No 40,744 2.7 1.00 Referent 21,515 4.5 1.00 Referent Behavioral factors  Smoking statusf   Nonsmoker 30,367 2.7 1.00 Referent 14,553 4.3 1.00 Referent   Former smoker 13,540 2.8 0.93 0.83, 1.05 5,163 5.0 1.11 0.97, 1.27   Current smoker 8,919 2.9 1.01 0.88, 1.16 3,228 5.3 1.09 0.92, 1.28 Environmental exposures  Occupational hazards requiring protective equipmentf   Yes 27,680 3.0 0.97 0.86, 1.09 7,358 5.4 1.13 0.99, 1.29   No 24,104 2.5 1.00 Referent 15,139 4.2 1.00 Referent   Don’t know 1,042 3.4 1.09 0.80, 1.50 447 5.6 1.11 0.76, 1.63  Routine skin contact with paints/solvents/substancesf   Yes 13,011 3.4 1.11 0.98, 1.26 2,881 5.7 1.00 0.84, 1.19   No 37,463 2.5 1.00 Referent 19,271 4.4 1.00 Referent   Don’t know 2,352 3.7 1.17 0.93, 1.47 792 5.1 0.93) 0.68, 1.27  Pesticides, including creams, sprays, and uniform  treatments, and applied in the environment/surroundingsf   Yes 19,133 3.2 1.08 0.96, 1.21 6,238 5.7 1.18 1.03, 1.35   No 29,579 2.4 1.00 Referent 14,732 4.1 1.00 Referent   Don’t know 4,114 3.2 1.01 0.84, 1.21 1,974 5.0 0.99 0.80, 1.21 Stressors  Life stressor eventsf,j   None 40,568 2.2 1.00 Referent 14,243 3.5 1.00 Referent   1 event 9,401 4.2 1.31 1.16, 1.47 5,231 5.4 1.11 0.97, 1.27   >1 event 2,857 5.6 1.37 1.15, 1.62 3,470 8.0 1.31 1.12, 1.52  PTSDf,k   Yes 2,840 4.8 1.82 1.55, 2.15 1,506 7.8 1.56 1.30, 1.88   No 49,986 2.6 1.00 Referent 21,438 4.4 1.00 Referent Characteristic . Mena,b (n = 52,826) . Womena,c (n = 22,944) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Exposure of interest  Combat deployment during follow-upd   Not deployed 29,237 3.1 1.00 Referent 15,678 5.0 1.00 Referent   Deployed without combat experience 8,922 1.6 0.97 0.83, 1.15 3,577 3.0 0.97 0.80, 1.18   Deployed with combat experience 14,667 2.7 1.30 1.14, 1.47 3,689 4.4 1.24 1.05, 1.46 Demographic factors  Race/ethnicity   White, non-Hispanic 41,880 2.7 1.00 Referent 15,895 4.4 1.00 Referent   Black, non-Hispanic 4,484 3.1 1.01 0.85, 1.20 3,692 5.1 1.06 0.90, 1.24   Asian/Pacific Islander 1,899 2.1 0.91 0.67, 1.22 1,051 4.6 1.44 1.12, 1.86   Hispanic 3,414 3.7 1.28 1.08, 1.53 1,701 5.5 1.37 1.13, 1.66   Other 1,149 2.5 0.84 0.59, 1.20 605 5.1 1.15 0.82, 1.61  Birth year   1980 or later 14,454 2.0 1.00 Referent 9,435 3.4 1.00 Referent   1970–1979 15,078 2.8 0.94 0.78, 1.14 6,754 5.0 1.14 0.96, 1.36   Before 1970 23,294 3.2 0.96 0.77, 1.21 6,755 5.9 1.19 0.96, 1.48  Marital status   Never married 14,362 2.1 0.88 0.77, 1.01 9,257 4.4 1.13 0.99, 1.30   Married 34,164 3.0 1.00 Referent 9,841 4.3 1.00 Referent   Divorced/widowed/separated 4,300 3.2 0.89 0.75, 1.05 3,846 6.0 1.19 1.02, 1.38  Education   High school degree/GED or less 9,686 2.5 0.95 0.79, 1.15 3,296 4.1 0.9 0.71, 1.14   Some college or associate degree 26,452 3.0 1.02 0.88, 1.19 12,142 4.9 1.13 0.95, 1.34   Bachelor degree or higher 16,688 2.5 1.00 Referent 7,506 4.3 1.00 Referent  BMIe,f   Underweight/normal (<25.0) 14,872 2.2 1.00 Referent 13,182 4.1 1.00 Referent   Overweight (25.0–29.9) 28,516 2.8 1.30 1.14, 1.47 7,498 5.2 1.36 1.20, 1.53   Obese (≥30.0) 9,438 3.4 1.88 1.62, 2.18 2,264 5.6 1.68 1.42, 2.00  Enrollment panelg   2001–2003 32,852 3.2 1.00 Referent 10,878 5.8 1.00 Referent   2004–2006 8,016 2.6 0.97 0.81, 1.16 5,024 4.3 1.01 0.85, 1.20   2007–2008 11,958 1.7 0.74 0.59, 0.92 7,042 2.9 0.76 0.62, 0.95 Military service  Service branch   Army 23,206 3.5 1.00 Referent 10,464 5.5 1.00 Referent   Navy/Coast Guard 9,354 2.4 0.79 0.69, 0.91 4,309 4.4 0.96 0.82, 1.12   Marine Corps 4,877 2.0 0.68 0.55, 0.84 676 4.4 0.98 0.70, 1.39   Air Force 15,389 2.1 0.69 0.61, 0.79 7,495 3.5 0.79 0.69, 0.91  Service component   Active duty 32,986 2.7 1.00 Referent 13,354 4.5 1.00 Referent   Reserve/Guard 19,840 2.8 0.83 0.74, 0.93 9,590 4.8 0.93 0.82, 1.05  Occupation   Infantry, gun crews, seamen (combat specialist) 12,285 2.5 1.00 Referent 1,429 3.4 1.00 Referent   Electrical repair 5,703 2.4 1.11 0.92, 1.34 1,512 3.3 0.93 0.65, 1.33   Communication/intelligence 4,263 3.1 1.17 0.96, 1.43 1,951 5.4 1.56 1.14, 2.15   Health care 3,745 3.2 1.26 1.03, 1.55 5,002 4.8 1.41 1.05, 1.88   Other technical and specialty 1,475 3.3 1.54 1.18, 2.03 698 4.7 1.51 1.01, 2.25   Functional support 7,403 3.1 1.34 1.14, 1.58 6,664 4.7 1.21 0.91, 1.62   Electrical/mechanical equipment repair 9,041 2.7 1.17 0.99, 1.38 1,734 3.8 1.00 0.70, 1.42   Craft workers 1,878 2.7 1.26 0.96, 1.65 422 4.0 0.97 0.58, 1.62   Service and supply 4,503 3.2 1.30 1.08, 1.57 2,318 5.7 1.32 0.96, 1.81   Non-occupation 2,530 1.7 0.85 0.63, 1.16 1,214 4.5 1.63 1.14, 2.32  Pay grade   Enlisted 40,620 2.9 1.25 1.05, 1.49 17,984 4.8 1.14 0.93, 1.39   Officerh 12,206 2.3 1.00 Referent 4,960 4.0 1.00 Referent  Military service statusf   Currently serving 44,404 2.9 1.00 Referent 19,450 4.6 1.00 Referent   Former service member 8,422 1.9 1.00 0.87, 1.16 3,494 4.4 1.37 1.17, 1.59  Prior deploymenti   Yes 12,082 3.0 0.96 0.84, 1.09 1,429 5.6 0.98 0.79, 1.21   No 40,744 2.7 1.00 Referent 21,515 4.5 1.00 Referent Behavioral factors  Smoking statusf   Nonsmoker 30,367 2.7 1.00 Referent 14,553 4.3 1.00 Referent   Former smoker 13,540 2.8 0.93 0.83, 1.05 5,163 5.0 1.11 0.97, 1.27   Current smoker 8,919 2.9 1.01 0.88, 1.16 3,228 5.3 1.09 0.92, 1.28 Environmental exposures  Occupational hazards requiring protective equipmentf   Yes 27,680 3.0 0.97 0.86, 1.09 7,358 5.4 1.13 0.99, 1.29   No 24,104 2.5 1.00 Referent 15,139 4.2 1.00 Referent   Don’t know 1,042 3.4 1.09 0.80, 1.50 447 5.6 1.11 0.76, 1.63  Routine skin contact with paints/solvents/substancesf   Yes 13,011 3.4 1.11 0.98, 1.26 2,881 5.7 1.00 0.84, 1.19   No 37,463 2.5 1.00 Referent 19,271 4.4 1.00 Referent   Don’t know 2,352 3.7 1.17 0.93, 1.47 792 5.1 0.93) 0.68, 1.27  Pesticides, including creams, sprays, and uniform  treatments, and applied in the environment/surroundingsf   Yes 19,133 3.2 1.08 0.96, 1.21 6,238 5.7 1.18 1.03, 1.35   No 29,579 2.4 1.00 Referent 14,732 4.1 1.00 Referent   Don’t know 4,114 3.2 1.01 0.84, 1.21 1,974 5.0 0.99 0.80, 1.21 Stressors  Life stressor eventsf,j   None 40,568 2.2 1.00 Referent 14,243 3.5 1.00 Referent   1 event 9,401 4.2 1.31 1.16, 1.47 5,231 5.4 1.11 0.97, 1.27   >1 event 2,857 5.6 1.37 1.15, 1.62 3,470 8.0 1.31 1.12, 1.52  PTSDf,k   Yes 2,840 4.8 1.82 1.55, 2.15 1,506 7.8 1.56 1.30, 1.88   No 49,986 2.6 1.00 Referent 21,438 4.4 1.00 Referent Abbreviations: BMI, body mass index; CI, confidence interval; GED, General Educational Development certificate; PTSD, posttraumatic stress disorder; RR, relative risk. a Univariate analyses were performed separately for men and women. For both sexes, combat deployment, birth year, marital status, education, BMI, enrollment panel, service branch, occupation, pay grade, environmental exposures, life stressor events, and PTSD were significantly associated with new-onset asthma (α = 0.05). Models were fitted separately for men and women; both models adjusted for all variables in the table. b Among men, race/ethnicity and military service status were also significantly associated with new-onset asthma (α = 0.05). c Among women, smoking status was also significantly associated with new-onset asthma (α = 0.05). d Deployment was defined as being deployed in support of Operation Enduring Freedom or Operation Iraqi Freedom. Combat was defined as reporting personal exposure to ≥1 of the following: witnessing death, witnessing physical abuse, dead and/or decomposing bodies, maimed soldiers or civilians, or prisoners of war or refugees. Combat deployment was assessed over the entire follow-up period. e BMI was calculated as weight (kg) divided by height (m) squared. f Time-varying covariates were measured at the survey prior to the final survey (the survey at which new-onset asthma was reported or the last survey completed, whichever came first). g This study used 3 panels of Millennium Cohort Study participants. Participants completed their baseline survey during the listed years (Panel 1: 2001–2003, Panel 2: 2004–2006, Panel 3: 2007–2008). h Officer includes commissioned and warrant officers. i Prior deployments to Bosnia, Kosovo, or Southwest Asia between January 1, 1998, and September 1, 2001. j Categorized number of endorsements of the following events: divorce or separation, financial problems, sexual assault, sexual harassment, physical assault, or suffered a disabling illness or injury. k Based on the PTSD Checklist–Civilian Version, using sensitive diagnostic criteria from the Diagnostic and Statistical Manual of Mental Disorder, Fourth Edition. Open in new tab Men and women who deployed with combat experience during the follow-up period had a higher risk of new-onset asthma compared with those who did not deploy, after adjusting for covariates (for men, relative risk = 1.30, 95% confidence interval: 1.14, 1.47; for women, relative risk = 1.24, 95% confidence interval: 1.05, 1.46) (Table 1). Among both men and women, Hispanic ethnicity, being overweight or obese, Army service, experiencing more than 1 stressful life event, PTSD, and health care or other technical and specialty (e.g., mapping, weather, ordnance disposal, or diving) occupations were risk factors for new-onset asthma in mutually adjusting models. Among men, additional risk factors included active-duty status, enlisted rank, experiencing 1 stressful life event, and occupations in functional support and service and supply. Additional risk factors among women included Asian/Pacific Islander race/ethnicity, being divorced/widowed/separated, separation from military service, non-occupations (e.g., patients, prisoners, and students) and communications/intelligence occupations, and answering “yes” to being personally exposed to pesticides, including creams, sprays, and uniform treatments, and pesticides applied in the environment/surroundings. Among the 31,152 participants who deployed, 23,797 (76.4%) were men and 7,355 (23.6%) were women (Table 2). Among this group, 547 (2.3%) men and 272 (3.7%) women developed asthma during the follow-up period. No significant association was found between multiple deployments and new-onset asthma (Table 2). A significant association was found between deployment duration and new-onset asthma, after adjusting for covariates; however, deployment duration was not independently significantly associated with new-onset asthma. No collinearity was observed between the independent variables included in the models shown in Tables 1 and 2. Table 2. Multiple Deployments, Deployment Duration, and Adjusted Relative Risk of New-Onset Asthma According to Sex Among Deployed Personnel (n = 31,152), Millennium Cohort Study, United States, 2001–2013 Characteristics . Mena (n = 23,797) . Womena (n = 7,355) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Multiple deploymentsb,c  Yes 6,452 2.1 0.99 0.82, 1.20 1,442 3.3 0.95 0.70, 1.28  No 17,345 2.4 1.00 Referent 5,913 3.8 1.00 Referent Deployment durationb,d  1–200 days 10,231 2.1 1.00 Referent 3,555 3.4 1.00 Referent  201–400 days 9,521 2.4 1.08 0.88, 1.32 2,979 4.3 1.16 0.89, 1.53  401–600 days 2,683 2.8 1.35 1.02, 1.78 575 2.3 0.71 0.41, 1.22  >600 days 1,362 1.8 0.91 0.59, 1.37 246 4.5 1.54 0.85, 2.80 Characteristics . Mena (n = 23,797) . Womena (n = 7,355) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Multiple deploymentsb,c  Yes 6,452 2.1 0.99 0.82, 1.20 1,442 3.3 0.95 0.70, 1.28  No 17,345 2.4 1.00 Referent 5,913 3.8 1.00 Referent Deployment durationb,d  1–200 days 10,231 2.1 1.00 Referent 3,555 3.4 1.00 Referent  201–400 days 9,521 2.4 1.08 0.88, 1.32 2,979 4.3 1.16 0.89, 1.53  401–600 days 2,683 2.8 1.35 1.02, 1.78 575 2.3 0.71 0.41, 1.22  >600 days 1,362 1.8 0.91 0.59, 1.37 246 4.5 1.54 0.85, 2.80 Abbreviations: CI, confidence interval; PTSD, posttraumatic stress disorder; RR, relative risk. a Models were fitted separately for men and women. Both models mutually adjusted for race/ethnicity, birth year, marital status, education, body mass index, panel, service branch, service component, occupation, pay grade, separation from the military, and prior deployment to Bosnia, Kosovo, or Southwest Asia between January 1, 1998, and September 1, 2001; smoking status, occupational hazards requiring protective equipment, routine skin contact with paints/solvents/substances, and exposure to pesticides, including creams, sprays, and uniform treatments, and pesticides applied in the environment/surroundings; life stressor events; and PTSD. b Separate models were fitted for multiple deployments and deployment duration. c Multiple deployments were defined as deploying in support of Operation Iraqi Freedom or Operation Enduring Freedom more than once during the follow-up period. d Deployment duration was defined as the cumulative number of days deployed in support of Operation Iraqi Freedom or Operation Enduring Freedom during the follow-up period. Open in new tab Table 2. Multiple Deployments, Deployment Duration, and Adjusted Relative Risk of New-Onset Asthma According to Sex Among Deployed Personnel (n = 31,152), Millennium Cohort Study, United States, 2001–2013 Characteristics . Mena (n = 23,797) . Womena (n = 7,355) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Multiple deploymentsb,c  Yes 6,452 2.1 0.99 0.82, 1.20 1,442 3.3 0.95 0.70, 1.28  No 17,345 2.4 1.00 Referent 5,913 3.8 1.00 Referent Deployment durationb,d  1–200 days 10,231 2.1 1.00 Referent 3,555 3.4 1.00 Referent  201–400 days 9,521 2.4 1.08 0.88, 1.32 2,979 4.3 1.16 0.89, 1.53  401–600 days 2,683 2.8 1.35 1.02, 1.78 575 2.3 0.71 0.41, 1.22  >600 days 1,362 1.8 0.91 0.59, 1.37 246 4.5 1.54 0.85, 2.80 Characteristics . Mena (n = 23,797) . Womena (n = 7,355) . No. . New-Onset Asthma, % . RR . 95% CI . No. . New-Onset Asthma, % . RR . 95% CI . Multiple deploymentsb,c  Yes 6,452 2.1 0.99 0.82, 1.20 1,442 3.3 0.95 0.70, 1.28  No 17,345 2.4 1.00 Referent 5,913 3.8 1.00 Referent Deployment durationb,d  1–200 days 10,231 2.1 1.00 Referent 3,555 3.4 1.00 Referent  201–400 days 9,521 2.4 1.08 0.88, 1.32 2,979 4.3 1.16 0.89, 1.53  401–600 days 2,683 2.8 1.35 1.02, 1.78 575 2.3 0.71 0.41, 1.22  >600 days 1,362 1.8 0.91 0.59, 1.37 246 4.5 1.54 0.85, 2.80 Abbreviations: CI, confidence interval; PTSD, posttraumatic stress disorder; RR, relative risk. a Models were fitted separately for men and women. Both models mutually adjusted for race/ethnicity, birth year, marital status, education, body mass index, panel, service branch, service component, occupation, pay grade, separation from the military, and prior deployment to Bosnia, Kosovo, or Southwest Asia between January 1, 1998, and September 1, 2001; smoking status, occupational hazards requiring protective equipment, routine skin contact with paints/solvents/substances, and exposure to pesticides, including creams, sprays, and uniform treatments, and pesticides applied in the environment/surroundings; life stressor events; and PTSD. b Separate models were fitted for multiple deployments and deployment duration. c Multiple deployments were defined as deploying in support of Operation Iraqi Freedom or Operation Enduring Freedom more than once during the follow-up period. d Deployment duration was defined as the cumulative number of days deployed in support of Operation Iraqi Freedom or Operation Enduring Freedom during the follow-up period. Open in new tab DISCUSSION Our study results indicate that combat deployment was associated with a 24%–30% higher risk of new-onset asthma, after adjusting for demographic and military characteristics, smoking status, environmental exposures, number of life stressor events, and PTSD status. This finding suggests an elevated risk of asthma in a population of deployed military personnel who are considered to be healthier than those who did not deploy (healthy deployer effect). In addition to combat deployment, other demographic and military characteristics, environmental exposures, and stressors were associated with new-onset asthma. Demographic characteristics significantly associated with asthma included race/ethnicity, marital status, and BMI. We observed that Hispanic men and women were 28%–37% more likely to develop asthma than non-Hispanic white personnel. We also observed, among women in our sample, that Asian/Pacific Islander women were 44% more likely to report new-onset asthma than non-Hispanic white women. This finding is inconsistent with reports of lower prevalence of asthma among Asian Americans, especially among those born outside the United States (30). However, this lower prevalence among Asian Americans may be driven by nativity, given that foreign-born Asian Americans would be less likely to enter the military than those born in the United States (31). Consistent with other clinical, epidemiologic, and systematic studies evaluating body size and asthma, the risk for new-onset asthma was associated with increasing BMI in our analysis (32–36). In our study, the highest risk for asthma among men and women, when adjusted for all covariates, was among those having a BMI of 30 or above (68%–88% higher risk). Military characteristics significantly associated with asthma included service branch, service component, occupation, pay grade, and military service status. Differences in exposure among men deployed to ground operations may explain the observed differences in risk of new-onset asthma by service branch, service component, and pay grade. Among men, Army service conferred the highest risk of new-onset asthma among the service branches; enlisted personnel had a 25% higher risk compared with officers, and Reserve/National Guard had a 17% lower risk compared with their active-duty counterparts. Certain occupations, including health care, have been associated with a higher risk of new-onset asthma in nonmilitary populations (37). Compared with combat specialists, health care was significantly associated with asthma in our study. Among women, new-onset asthma was associated with self-reported exposure to pesticides, but, due to the broad wording of the survey question, we were unable to distinguish whether this was attributable to specific uses, chemical classes, or application methods. In numerous Periodic Occupational and Environmental Monitoring Summary reports of environmental hazards at operating bases in theater, permethrin and other insecticides, rodenticides, and herbicides were used for pest control at bases (38). While not available for study participants in this analysis, in-country region of deployment or base location could serve as a proxy for environmental exposures. Several studies have investigated the association between pesticides and adult asthma but have reported inconsistent findings (39). Although smoking has been associated with exacerbation of asthma symptoms, a recent review indicated there was insufficient evidence to infer a causal relationship with new-onset asthma (40). We observed no significant association between asthma and smoking status in our study population. Prolonged exposure to stress may trigger or exacerbate the symptoms of asthma (41). Allostatic load (the wearing down of the body following repeated stress) (42) has been found to be associated with higher asthma risk (43) and may explain the observed association between combat deployment and new-onset asthma. Our results also suggest a possible dose-dependent increase in asthma risk among those experiencing more than 1 stressful life event compared with those who experienced only 1 stressful life event. The observed associations between new-onset asthma and PTSD and stressful life events may also be manifestations of allostatic load. One of the limitations of this analysis is that the findings may be susceptible to recall and reporting bias due to our reliance on self-reported survey data. However, the survey instruments employed were validated and consistently administered over follow-up surveys (28). In addition, Oksanen et al. (44) observed high validity of self-reported physician diagnoses of incident asthma (63% sensitivity and 91% specificity) and prevalent asthma (91% sensitivity and 97% specificity) when compared with health records. Electronic military medical records were not used because this information is available only for active-duty and activated Reserve or National Guard personnel; these data are not available for separated personnel and inactivated Reserve or National Guard personnel. However, we did compare self-reported asthma endorsements with medical record diagnoses among 4,477 eligible individuals. We found 50.4 positive agreement and 97.6 negative agreement, consistent with a prior Millennium Cohort analysis for acute conditions that included asthma and a Finnish public-sector employee study (45). The measurement of stressful life events is limited in the Millennium Cohort Study because the items focus on several general categories and do not measure all types of stressors or the subjective experience of stress. Because self-reported asthma diagnosis was based on the 3 years preceding completion of the follow-up survey, the exact timing of asthma development and combat exposure could not be determined. Furthermore, the 5-item combat questions were not necessarily specific to a particular deployment. However, the longitudinal nature of this study did eliminate many of the limitations that previous cross-sectional studies have encountered and allowed for investigation of the temporal associations between combat deployment and asthma. Despite these limitations, the large sample size and population-based design with up to 12 years of prospective follow-up allowed us to identify incident asthma cases among participants from all branches of the military, including active-duty and Reserve/National Guard personnel and those who separated from military service. This design also allowed us to exclude prevalent cases of asthma at baseline. In addition, important covariates, such as smoking and BMI, were also available in these prospective data. To our knowledge, this is the only longitudinal, prospective epidemiologic study that has examined the association between combat deployment and new-onset asthma with over a decade of follow-up. These findings indicate a higher risk of new-onset asthma among those who deployed and experienced combat compared with those who did not deploy, but there was no observed change in risk among those who deployed without combat experience compared with those who did not deploy. This implies that specific attributes of combat or other deployment exposures and experiences may drive the association with elevated risk of new-onset asthma. Further research is needed to identify specific aspects of combat deployment that are associated with greater asthma risk, the knowledge of which can form the basis for prevention strategies among military service members. ACKNOWLEDGMENTS Author affiliations: The Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc., Bethesda, Maryland (Anna C. Rivera, Teresa M. Powell); Seattle Epidemiologic Research and Information Center, Veterans Affairs Puget Sound Health Care System, Seattle, Washington (Edward J. Boyko); Department of Epidemiology, University of Washington School of Public Health, Seattle, Washington (Edward J. Boyko); Division of Allergy and Immunology, Department of Internal Medicine, Naval Medical Center San Diego, San Diego, California (Rachel U. Lee); and Deployment Health Research Department, Naval Health Research Center, San Diego, California (Dennis J. Faix, David D. Luxton, Rudy P. Rull). This work was supported by the Military Operational Medicine Research Program (work unit 60002). We thank the Millennium Cohort Study participants. In addition to the authors, the Millennium Cohort Study Team includes Dr. Richard Armenta; Lauren Bauer, MPH; Dr. Deborah Bookwalter; Satbir Boparai, MBA; Ania Bukowinski, MPH; Carlos Carballo, MS; Dr. Adam Cooper; James Davies; Alex Esquivel, MPH; Dr. Susan Farrish; Toni Rose Geronimo, MPH; Gia Gumbs, MPH; Isabel Jacobson, MPH; Dr. Zeina Khodr; Claire Kolaja, MPH; Cynthia LeardMann, MPH; William Lee; Gordon Lynch; Chris Lo; Denise Lovec-Jenkins; Dr. Rayna Matsuno; Dr. Chiping Nieh; Anet Petrosyan; Dr. Jacqueline Pflieger; Dr. Chris Phillips; Dr. Ben Porter; Dr. Sabrina Richardson; Kimberly Roenfeldt, MAS; Beverly Sheppard; Steven Speigle; Dr. Valerie Stander; Evelyn Sun, MPH; Dr. Daniel Trone; Daniel Vaughan; Jennifer Walstrom; Steven Warner, MPH; Dr. Marleen Welsh; and Kelly Woodall, MPH, from the Deployment Health Research Department, Naval Health Research Center, San Diego, California. We appreciate the support from the Management Information Division, US Defense Manpower Data Center, Seaside, California, and the Military Operational Medicine Research Program, US Army Medical Research and Materiel Command, Fort Detrick, Maryland. A portion of this work was presented at the American Public Health Association Annual Meeting and Expo, October 29 to November 2, 2016, Denver, Colorado; at the Military Health System Research Symposium, August 27–30 2017, Kissimmee, Florida; at the American Academy of Allergy Asthma and Immunology/World Allergy Organization Joint Congress, March 2–5, 2018, Orlando, Florida; and at the Navy and Marine Corps Public Health Conference, March 20–22, 2018, Norfolk, Virginia. I am a military service member (or employee of the US Government). This work was prepared as part of my official duties. Title 17, U.S.C. §105 provides the “Copyright protection under this title is not available for any work of the United States Government.” Title 17, U.S.C. §101 defines a US Government work as work prepared by a military service member or employee of the US Government as part of that person’s official duties. Report No. 18-38 was supported by the Military Operational Medicine Research Program under work unit no. 60002. The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Department of the Navy, Department of the Army, Department of the Air Force, Department of Veterans Affairs, Department of Defense, or the US Government. Approved for public release; distribution unlimited. Human subjects participated in this study after giving their free and informed consent. This research has been conducted in compliance with all applicable federal regulations governing the protection of human subjects in research (Protocol NHRC.2000.0007). Conflict of interest: none declared. Abbreviations BMI body mass index CTS Contingency Tracking System OEF Operation Enduring Freedom OIF Operation Iraqi Freedom PTSD posttraumatic stress disorder REFERENCES 1 Committee on the Assessment of the Readjustment Needs of Military Personnel, Veterans, and Their Families; Board on the Health of Select Populations; Institute of Medicine . Returning Home from Iraq and Afghanistan: Assessment of Readjustment Needs of Veterans, Service Members, and Their Families . Washington, DC : The National Academies Press ; 2014 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 2 Sanders JW , Putnam SD, Frankart C, et al. . Impact of illness and non-combat injury during Operations Iraqi Freedom and Enduring Freedom (Afghanistan) . Am J Trop Med Hyg . 2005 ; 73 ( 4 ): 713 – 719 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Roop SA , Niven AS, Calvin BE, et al. . The prevalence and impact of respiratory symptoms in asthmatics and nonasthmatics during deployment . Mil Med . 2007 ; 172 ( 12 ): 1264 – 1269 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Szema AM , Peters MC, Weissinger KM, et al. . New-onset asthma among soldiers serving in Iraq and Afghanistan . Allergy Asthma Proc . 2010 ; 31 ( 5 ): 67 – 71 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Sessions CK . Asthma, active component, US Armed forces, 1999–2008 . MSMR . 2009 ; 16 ( 7 ): 5 . Google Scholar OpenURL Placeholder Text WorldCat 6 Karlinsky JB , Blanchard M, Alpern R, et al. . Late prevalence of respiratory symptoms and pulmonary function abnormalities in Gulf War I Veterans . Arch Intern Med . 2004 ; 164 ( 22 ): 2488 – 2491 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Gray GC , Callahan JD, Hawksworth AW, et al. . Respiratory diseases among US military personnel: countering emerging threats . Emerg Infect Dis . 1999 ; 5 ( 3 ): 379 – 385 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Gray GC , Kaiser KS, Hawksworth AW, et al. . Increased postwar symptoms and psychological morbidity among US Navy Gulf War veterans . Am J Trop Med Hyg . 1999 ; 60 ( 5 ): 758 – 766 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Richards AL , Hyams KC, Watts DM, et al. . Respiratory disease among military personnel in Saudi Arabia during Operation Desert Shield . Am J Public Health . 1993 ; 83 ( 9 ): 1326 – 1329 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Young RC , Jr, Rachal RE, Huguley JW , 3rd. Environmental health concerns of the Persian Gulf War . J Natl Med Assoc . 1992 ; 84 ( 5 ): 417 – 424 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 11 Coker WJ , Bhatt BM, Blatchley NF, et al. . Clinical findings for the first 1000 Gulf war veterans in the Ministry of Defence’s medical assessment programme . BMJ . 1999 ; 318 ( 7179 ): 290 – 294 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Iowa Persian Gulf Study Group . Self-reported illness and health status among Gulf War veterans: a population-based study. The Iowa Persian Gulf Study Group . JAMA . 1997 ; 277 ( 3 ): 238 – 245 . Crossref Search ADS PubMed WorldCat 13 Weese CB , Abraham JH. Potential health implications associated with particulate matter exposure in deployed settings in Southwest Asia . Inhal Toxicol . 2009 ; 21 ( 4 ): 291 – 296 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Abraham JH , DeBakey SF, Reid L, et al. . Does deployment to Iraq and Afghanistan affect respiratory health of US military personnel? J Occup Environ Med . 2012 ; 54 ( 6 ): 740 – 745 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Baird CP , DeBakey S, Reid L, et al. . Respiratory health status of US Army personnel potentially exposed to smoke from 2003 Al-Mishraq Sulfur Plant fire . J Occup Environ Med . 2012 ; 54 ( 6 ): 717 – 723 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Morris MJ , Zacher LL, Jackson DA. Investigating the respiratory health of deployed military personnel . Mil Med . 2011 ; 176 ( 10 ): 1157 – 1161 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Rose CS . Military service and lung disease . Clin Chest Med . 2012 ; 33 ( 4 ): 705 – 714 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Smith B , Wong CA, Boyko EJ, et al. . The effects of exposure to documented open-air burn pits on respiratory health among deployers of the Millennium Cohort Study . J Occup Environ Med . 2012 ; 54 ( 6 ): 708 – 716 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Smith B , Wong CA, Smith TC, et al. . Newly reported respiratory symptoms and conditions among military personnel deployed to Iraq and Afghanistan: a prospective population-based study . Am J Epidemiol . 2009 ; 170 ( 11 ): 1433 – 1442 . Google Scholar Crossref Search ADS PubMed WorldCat 20 DelVecchio SP , Collen JF, Zacher LL, et al. . The impact of combat deployment on asthma diagnosis and severity . J Asthma . 2015 ; 52 ( 4 ): 363 – 369 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Abraham JH , Eick-Cost A, Clark LL, et al. . A retrospective cohort study of military deployment and postdeployment medical encounters for respiratory conditions . Mil Med . 2014 ; 179 ( 5 ): 540 – 546 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Barth SK , Dursa EK, Peterson MR, et al. . Prevalence of respiratory diseases among veterans of Operation Enduring Freedom and Operation Iraqi Freedom: results from the National Health Study for a New Generation of US Veterans . Mil Med . 2014 ; 179 ( 3 ): 241 – 245 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Piccirillo AL , Packnett ER, Cowan DN, et al. . Epidemiology of asthma-related disability in the US Armed Forces: 2007–2012 . J Asthma . 2016 ; 53 ( 7 ): 668 – 678 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Abraham JH , Clark LL, Sharkey JM, et al. . Trends in rates of chronic obstructive respiratory conditions among US military personnel, 2001–2013 . US Army Med Dep J . 2014 : 33 – 43 . Google Scholar OpenURL Placeholder Text WorldCat 25 Gray GC , Chesbrough KB, Ryan MA, et al. . The Millennium Cohort Study: a 21-year prospective cohort study of 140,000 military personnel . Mil Med . 2002 ; 167 ( 6 ): 483 – 488 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Ryan MA , Smith TC, Smith B, et al. . Millennium Cohort: enrollment begins a 21-year contribution to understanding the impact of military service . J Clin Epidemiol . 2007 ; 60 ( 2 ): 181 – 191 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Defense Manpower Data Center, US Department of Defense . Personnel data. https://www.dmdc.osd.mil/appj/dwp/personnel_data.jsp. Accessed October 01, 2016. 28 Smith TC , Smith B, Jacobson IG, et al. . Reliability of standard health assessment instruments in a large, population-based cohort study . Ann Epidemiol . 2007 ; 17 ( 7 ): 525 – 532 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Akinbami LJ , Moorman JE, Bailey C, et al. . Trends in asthma prevalence, health care use, and mortality in the united states, 2001–2010 . NCHS Data Brief . 2012 ;( 94 ): 1 – 8 . Google Scholar OpenURL Placeholder Text WorldCat 30 Bhan N , Kawachi I, Glymour MM, et al. . Time Trends in Racial and Ethnic Disparities in Asthma Prevalence in the United States From the Behavioral Risk Factor Surveillance System (BRFSS) Study (1999–2011) . Am J Public Health . 2015 ; 105 ( 6 ): 1269 – 1275 . Google Scholar Crossref Search ADS PubMed WorldCat 31 Batalova J . Immigrants in the US Armed Forces. https://www.migrationpolicy.org/article/immigrants-us-armed-forces. Published May 15, 2008 . Accessed October 18, 2016. 32 Barros R , Moreira P, Padrão P, et al. . Obesity increases the prevalence and the incidence of asthma and worsens asthma severity . Clin Nutr . 2017 ; 36 ( 4 ): 1068 – 1074 . Google Scholar Crossref Search ADS PubMed WorldCat 33 Boulet LP . Asthma and obesity . Clin Exp Allergy . 2013 ; 43 ( 1 ): 8 – 21 . Google Scholar Crossref Search ADS PubMed WorldCat 34 Leinaar E , Alamian A, Wang L. A systematic review of the relationship between asthma, overweight, and the effects of physical activity in youth . Ann Epidemiol . 2016 ; 26 ( 7 ): 504 – 510.e6 . Google Scholar Crossref Search ADS PubMed WorldCat 35 Çelebi Sözener Z , Aydın Ö, Mungan D, et al. . Obesity-asthma phenotype: effect of weight gain on asthma control in adults . Allergy Asthma Proc . 2016 ; 37 ( 4 ): 311 – 317 . Google Scholar Crossref Search ADS PubMed WorldCat 36 Urban N , Boivin MR, Cowan DN. Fitness, obesity and risk of asthma among Army trainees . Occup Med (Lond) . 2016 ; 66 ( 7 ): 551 – 557 . Google Scholar Crossref Search ADS PubMed WorldCat 37 Kim JL , Torén K, Lohman S, et al. . Respiratory symptoms and respiratory-related absence from work among health care workers in Sweden . J Asthma . 2013 ; 50 ( 2 ): 174 – 179 . Google Scholar Crossref Search ADS PubMed WorldCat 38 US Army Public Health Center . Periodic Occupational and Environmental Monitoring Summary (POEMS). https://phc.amedd.army.mil/topics/envirohealth/hrasm/Pages/POEMS.aspx. Accessed May 1, 2017. 39 Prince Edward Island Canada Health and Wellness . Pesticides and Human Health. Part 2: PEI Health and Pesticide Use. Prince Edward Island, Canada : Prince Edward Island Canada Health and Wellness ; 2015 . 40 Warren GW , Alberg AJ, Kraft AS, et al. . The 2014 Surgeon General’s report: “The health consequences of smoking—50 years of progress”: a paradigm shift in cancer care . Cancer . 2014 ; 120 ( 13 ): 1914 – 1916 . Google Scholar Crossref Search ADS PubMed WorldCat 41 Rod NH , Kristensen TS, Lange P, et al. . Perceived stress and risk of adult-onset asthma and other atopic disorders: a longitudinal cohort study . Allergy . 2012 ; 67 ( 11 ): 1408 – 1414 . Google Scholar Crossref Search ADS PubMed WorldCat 42 Juster RP , McEwen BS, Lupien SJ. Allostatic load biomarkers of chronic stress and impact on health and cognition . Neurosci Biobehav Rev . 2010 ; 35 ( 1 ): 2 – 16 . Google Scholar Crossref Search ADS PubMed WorldCat 43 Bahreinian S , Ball GD, Vander Leek TK, et al. . Allostatic load biomarkers and asthma in adolescents . Am J Respir Crit Care Med . 2013 ; 187 ( 2 ): 144 – 152 . Google Scholar Crossref Search ADS PubMed WorldCat 44 Oksanen T , Kivimäki M, Pentti J, et al. . Self-report as an indicator of incident disease . Ann Epidemiol . 2010 ; 20 ( 7 ): 547 – 554 . Google Scholar Crossref Search ADS PubMed WorldCat 45 Smith B , Chu LK, Smith TC, et al. . Challenges of self-reported medical conditions and electronic medical records among members of a large military cohort . BMC Med Res Methodol . 2008 ; 8 : 37 . Google Scholar Crossref Search ADS PubMed WorldCat © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journalspermissions@oup.com © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health.

Pettee Gabriel, Kelley; Sidney, Stephen; Jacobs, David R; Whitaker, Kara M; Carnethon, Mercedes R; Lewis, Cora E; Schreiner, Pamela J; Malkani, Raja I; Shikany, James M; Reis, Jared P; Sternfeld, Barbara

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy117pmid: 29893772

Abstract We describe 10-year changes in accelerometer-determined physical activity (PA) and sedentary time in a midlife cohort of the Coronary Artery Risk Development in Young Adults Study, within and by race and sex groups. Participants (n = 962) wore the accelerometer with valid wear (≥4 of 7 days, ≥10 hours per day) at baseline (2005–2006; ages 38–50 years) and 10-year follow-up (2015–2016; ages 48–60 years). Data were calibrated to account for accelerometer model differences. Participants (mean age = 45.0 (standard deviation, 3.5) years at baseline) had reduced accelerometer counts overall (mean = −65.5 (standard error (SE), 10.2) counts per minute/day), and within race and sex groups (all P < 0.001). Sedentary time increased overall (mean = 37.9 (SE, 3.7) minutes/day) and within race and sex groups, whereas light-intensity PA (mean = −30.6 (SE, 2.7) minutes/day) and moderate- to vigorous-intensity PA (mean = −7.5 (SE, 0.8) minutes/day) declined overall and within race and sex groups (all P < 0.001). Significant differences in 10-year change estimates were noted by race and sex groups for accelerometer counts, sedentary time, and moderate- to vigorous-intensity PA bouts; black men had the greatest reductions in PA compared with other groups. PA declines during midlife were characterized by reductions in light-intensity PA with increases in sedentary time, which may have important health consequences. Targeted efforts are needed to preserve PA, regardless of intensity level, across midlife. accelerometry, cohort study, diverse sample A key strategy to attenuate the public health burden attributable to noncommunicable diseases and mobility disability is the promotion of lifestyle behaviors, including physical activity. There is substantial evidence to suggest achieving the US Department of Health and Human Services’ physical activity guidelines (PAG) (1) of at least 150 minutes per week of moderate-intensity physical activity, at least 75 minutes per week of vigorous-intensity physical activity, or an equivalent combination of moderate- and vigorous-intensity physical activity (MVPA) reduces risk of premature death and several leading causes of disease and disability. Physical activity is also recommended for management of related conditions, including obesity, hypertension, and type 2 diabetes (1–3). Prolonged sedentary time also is discouraged in the PAG (1). Age-related declines in MVPA, based primarily on reported methods (e.g., questionnaires), are well documented in the literature (1, 4), including the Coronary Artery Risk Development in Young Adults (CARDIA) study (5–9). Furthermore, these age-related declines are likely due to physiological changes, including reductions in aerobic and muscular fitness (10, 11). However, due to previous reliance on questionnaires, there is currently limited evidence demonstrating total physical activity change and change within intensity categories with aging. This important gap in knowledge was recently highlighted in the 2018 Physical Activity Guidelines Advisory Committee Scientific Report (3). CARDIA is uniquely poised to contribute to these research gaps because accelerometer-based measures were implemented in this biracial cohort early (2005–2006) in comparison with other US-based cardiovascular observational cohort studies. Furthermore, CARDIA data include a second wave of accelerometer data collected 10 years later, which provide the opportunity to describe changes in accelerometer-based measures across 10 years during midlife. Midlife may be a particularly vulnerable time during adulthood because it corresponds to a period when risk of disease and disability escalates (12). Yet, compared with older adulthood, midlife may be a time when individuals are more willing and physically able to initiate a physical activity routine, particularly after retirement when there are more discretionary hours during which to be active (13). Therefore, the primary purpose of this study was to describe the 10-year changes in accelerometer-based physical activity and sedentary behavior measures of CARDIA participants and explore differences within and across black men and women and white men and women. A secondary objective was to describe these 10-year changes by baseline age and cardiorespiratory fitness level. METHODS Beginning in 1985–1986, CARDIA researchers recruited 5,115 participants aged 18–30 years from 4 geographical locations (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California). Baseline data for this analysis were collected at examinations in year 20 (2005–2006) and at follow-up in year 30 (2015–2016), at which times retention among survivors was 72% and 71%, respectively. At both visits, accelerometer data were collected as part of ancillary studies performed in conjunction with the core clinical examination. Standardized questionnaires were used at each visit to assess participant characteristics, including sex, race, age, and educational attainment. Reported physical activity was assessed at all visits using the CARDIA Physical Activity History, a reliable and valid measure (14, 15). At baseline, a maximal symptom-limited, graded exercise test, using a modified Balke protocol (16), was completed to assess cardiorespiratory fitness. Participants provided written informed consent, and CARDIA is approved annually by the institutional review boards of each participating center. Accelerometer data were collected using the ActiGraph 7164 and wGT3X-BT accelerometer models (ActiGraph, Pensacola, Florida) at the baseline and follow-up examinations, respectively, using identical protocols. Participants were asked to wear the accelerometer on their hip for 7 consecutive days during all waking hours (except during water-based activities). Once returned, data were downloaded and screened for wear time using the Troiano algorithm (17). These analyses included 962 participants (37.6% black women; 63.2% women overall) with valid wear time (17) (≥10 hours per day for ≥4 days) at both visits. The analytic sample was statistically significantly older (mean age = 25.1 (standard error (SE), 0.11) vs. 24.8 (SE, 0.06) years at year 0) and more likely to be female (63.2% vs. 52.5%), white (62.5% vs. 45.2%), and hold a Bachelor’s degree (40.3% vs. 33.4%) (all P < 0.001); no significant difference in reported physical activity level was observed. To optimize consistency across visits, raw data from the vertical axis of the wGT3X-BT accelerometer model were reintegrated to 60-second epochs with the low-frequency extension applied (18). Based on results of a CARDIA methodological substudy (n = 87), follow-up accelerometer count data were calibrated (counts divided by 1.088) to account for ActiGraph model differences as reported by Whitaker et al. (19). Total counts per minute and average counts per minute/day were calculated and minutes per day spent performing sedentary activity (i.e., <100 counts per minute), light-intensity physical activity (i.e., 100–1,951 counts per minute), and MVPA (i.e., ≥1,952 counts per minute) were estimated using Freedson cutpoints (20). Freedson cutpoints were selected because of their broad use in physical activity research. MVPA (calculated as every minute ≥1,952 counts per minute) and MVPA bouts (estimates only include ≥8 of 10 consecutive minutes of ≥1,952 counts per minute) (17) are reported. Although there is some controversy about using accelerometer-based data to assess behavioral targets (21, 22), at least 150 minutes/week of accumulated MVPA was used to classify participants as meeting PAG. Ten-year change estimates were computed as follow-up minus baseline data. Absolute change, rather than percent change, is reported to optimize the interpretations of the findings (i.e., minutes/day vs. percent change). Accelerometer estimates are reported as means with standard errors or proportions. Student t test was used to examine differences in baseline and follow-up accelerometer estimates, and overall and within race and sex groups. Analysis of variance and Tukey post hoc tests were used to examine differences in the 10-year change estimates between race and sex groups. To address the secondary study objective, differences in accelerometer estimates were also examined by baseline (i.e., year 20) age (<45 years vs. ≥45 years) and fitness level (less than vs. more than the median value of 452 seconds); Student t test was used to examine differences in the 10-year change estimates by the age and fitness categories. All statistical significance tests were 2-sided with the type I error level set at P < 0.05. All analyses were generated with SAS/STAT software, version 9.4 (SAS Institute, Inc., Cary, North Carolina). RESULTS Table 1 lists descriptive accelerometer data among the full analytic sample (mean age = 45.0 (standard deviation, 3.5) years at baseline). There was excellent participant compliance with wear time, at approximately 15 hours/day at each visit. Therefore, 10-year change estimates were not further adjusted for wear time. Total (counts per day) and average accelerometer counts (count per minute/day), composite measures reflecting both sedentary and physical activity (23), significantly decreased over 10 years, and included increases in sedentary time (mean = 37.9 (SE, 3.7) minutes/day), coupled with reductions in light-intensity physical activity (mean = −30.6 (SE, 2.7) minutes/day) and MVPA (mean = −7.5 (SE, 0.8) minutes/day). A slight increase in MVPA bouts (mean = 2.3 (SE, 0.7) minutes/day) was also observed. Table 1. Baseline (2005–2006), 10-Year Follow-up (2015–2016), and 10-Year Change in Accelerometer-Determined Physical Activity and Sedentary Behavior Estimates (n = 962), Coronary Artery Risk Development in Young Adults Study Variable . Baseline, Mean (SE) . Follow-up, Mean (SE) . 10-Year Change, Mean (SE) . P Valuea . Wear time, minutes/day 894.1 (2.8) 893.9 (2.9) −0.17 (3.3) 0.96 Total accelerometer counts, per 10,000 counts per day 35.1 (1.2) 28.7 (0.4) −6.3 (1.2) <0.001 Average accelerometer counts, counts per minute/day 386.2 (10.2) 320.7 (4.2) −65.5 (10.2) <0.001 Sedentary time, minutes/day 495.4 (3.3) 533.2 (3.4) 37.9 (3.7) <0.001 Light-intensity PA, minutes/day 362.1 (2.7) 331.5 (2.7) −30.6 (2.7) <0.001 MVPA, minutes/day 36.6 (0.8) 29.2 (0.8) −7.5 (0.8) <0.001 MVPA in bouts, minutes/dayb 12.5 (0.6) 14.8 (0.6) 2.3 (0.7) <0.001 Reported physical activityc, EU 353.3 (9.1) 340.3 (8.7) −13.0 (7.5) 0.09 Variable . Baseline, Mean (SE) . Follow-up, Mean (SE) . 10-Year Change, Mean (SE) . P Valuea . Wear time, minutes/day 894.1 (2.8) 893.9 (2.9) −0.17 (3.3) 0.96 Total accelerometer counts, per 10,000 counts per day 35.1 (1.2) 28.7 (0.4) −6.3 (1.2) <0.001 Average accelerometer counts, counts per minute/day 386.2 (10.2) 320.7 (4.2) −65.5 (10.2) <0.001 Sedentary time, minutes/day 495.4 (3.3) 533.2 (3.4) 37.9 (3.7) <0.001 Light-intensity PA, minutes/day 362.1 (2.7) 331.5 (2.7) −30.6 (2.7) <0.001 MVPA, minutes/day 36.6 (0.8) 29.2 (0.8) −7.5 (0.8) <0.001 MVPA in bouts, minutes/dayb 12.5 (0.6) 14.8 (0.6) 2.3 (0.7) <0.001 Reported physical activityc, EU 353.3 (9.1) 340.3 (8.7) −13.0 (7.5) 0.09 Abbreviations: EU, exercise units; MVPA, moderate- to vigorous-intensity physical activity; PA, physical activity; SE, standard error. a Differences between baseline and follow-up values were determined using a Student t test. b At least 8 of 10 consecutive minutes above the threshold of 1,952 counts per minute. cn = 956. Open in new tab Table 1. Baseline (2005–2006), 10-Year Follow-up (2015–2016), and 10-Year Change in Accelerometer-Determined Physical Activity and Sedentary Behavior Estimates (n = 962), Coronary Artery Risk Development in Young Adults Study Variable . Baseline, Mean (SE) . Follow-up, Mean (SE) . 10-Year Change, Mean (SE) . P Valuea . Wear time, minutes/day 894.1 (2.8) 893.9 (2.9) −0.17 (3.3) 0.96 Total accelerometer counts, per 10,000 counts per day 35.1 (1.2) 28.7 (0.4) −6.3 (1.2) <0.001 Average accelerometer counts, counts per minute/day 386.2 (10.2) 320.7 (4.2) −65.5 (10.2) <0.001 Sedentary time, minutes/day 495.4 (3.3) 533.2 (3.4) 37.9 (3.7) <0.001 Light-intensity PA, minutes/day 362.1 (2.7) 331.5 (2.7) −30.6 (2.7) <0.001 MVPA, minutes/day 36.6 (0.8) 29.2 (0.8) −7.5 (0.8) <0.001 MVPA in bouts, minutes/dayb 12.5 (0.6) 14.8 (0.6) 2.3 (0.7) <0.001 Reported physical activityc, EU 353.3 (9.1) 340.3 (8.7) −13.0 (7.5) 0.09 Variable . Baseline, Mean (SE) . Follow-up, Mean (SE) . 10-Year Change, Mean (SE) . P Valuea . Wear time, minutes/day 894.1 (2.8) 893.9 (2.9) −0.17 (3.3) 0.96 Total accelerometer counts, per 10,000 counts per day 35.1 (1.2) 28.7 (0.4) −6.3 (1.2) <0.001 Average accelerometer counts, counts per minute/day 386.2 (10.2) 320.7 (4.2) −65.5 (10.2) <0.001 Sedentary time, minutes/day 495.4 (3.3) 533.2 (3.4) 37.9 (3.7) <0.001 Light-intensity PA, minutes/day 362.1 (2.7) 331.5 (2.7) −30.6 (2.7) <0.001 MVPA, minutes/day 36.6 (0.8) 29.2 (0.8) −7.5 (0.8) <0.001 MVPA in bouts, minutes/dayb 12.5 (0.6) 14.8 (0.6) 2.3 (0.7) <0.001 Reported physical activityc, EU 353.3 (9.1) 340.3 (8.7) −13.0 (7.5) 0.09 Abbreviations: EU, exercise units; MVPA, moderate- to vigorous-intensity physical activity; PA, physical activity; SE, standard error. a Differences between baseline and follow-up values were determined using a Student t test. b At least 8 of 10 consecutive minutes above the threshold of 1,952 counts per minute. cn = 956. Open in new tab There were also no significant differences in accelerometer wear time at baseline and follow-up within or between race and sex groups (all P > 0.05); therefore, estimates were also left unadjusted (Table 2). In general, 10-year change patterns in the analytic sample emerged within each race and sex group. However, 10-year increases in MVPA bouts were significant only in white men and white women (mean = 4.7 (SE, 1.3) minutes/day and 2.7 (SE, 0.9) minutes/day, respectively; both P < 0.01). Significant differences in 10-year changes in accelerometer counts, sedentary time, and MVPA bouts were noted by race and sex groups (all P < 0.05). Black men, who started with the highest accelerometer counts, had the largest reductions (mean = −181.0 (SE, 77.0) counts per minute per day)—a reduction that was significantly different from all other sex and race groups. Compared with white women, black women, who started with the lowest counts, had significantly greater reductions (all P < 0.05). Ten-year changes in sedentary time and MVPA bouts also significantly differed between black men and white men (both P < 0.05). Black men also had the greatest declines in reported physical activity compared with all other race and sex groups. Finally, the proportion of participants meeting PAG was higher at baseline than at follow-up, regardless of race or sex group (Web Figure 1, available at https://academic.oup.com/aje). As shown in Web Table 1, younger participants at baseline had significantly larger increases in sedentary time compared with that of the older age group. Participants with higher cardiorespiratory fitness level at baseline had significantly larger increases in MVPA bouts (P = 0.003). No other differences in the 10-year accelerometer change estimates were noted by baseline age or fitness groups. Table 2. Baseline (2005–2006), 10-Year Follow-up (2015–2016), and 10-Year Change in Accelerometer-Determined Physical Activity and Sedentary Behavior Estimates Within and Stratified by Race and Sex Groups (n = 962), Coronary Artery Risk Development in Young Adults Study Variable . Black Men (n = 117; 12.2%), mean (SE) . White Men (n = 237; 24.6%), mean (SE) . Black Women (n = 244; 25.4%), mean (SE) . White Women (n = 364, 37.8%), mean (SE) . P Valuea . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . Wear time, minutes/day 906.3 (9.8) 909.9 (10.8) 3.7 (13.4) 905.1 (4.7) 902.3 (5.5) −2.8 (5.8) 880.1 (6.4) 889.2 (6.2) 9.1 (7.5) 892.4 (4.2) 886.5 (4.0) −5.8 (4.5) 0.34 Total accelerometer counts, per 10,000 counts per day 50.2 (9.7) 29.9 (1.2) −20.4 (9.7)b 36.4 (0.9) 32.2 (0.9) −4.3 (0.8)c 28.9 (0.7) 25.0 (0.7) −3.9 (0.7)c 33.3 (0.6) 28.6 (0.6) −4.7 (0.6) c 0.001d,e,f Average accelerometer counts, counts per minute/day 509.4 (76.5) 328.5 (13.8) −181.0 (77.0)b 402.6 (9.2) 355.5 (9.5) −47.0 (8.6)c 329.8 (7.9) 282.5 (7.5) −47.2 (7.7)c 373.8 (6.4) 321.2 (6.1) −52.6 (6.5)c 0.004d,e,f,g Sedentary time, minutes/day 488.2 (11.2) 548.8 (11.8) 60.6 (13.7)c 520.5 (6.2) 545.2 (6.7) 24.6 (6.8)c 484.0 (6.9) 526.7 (7.4) 42.7 (7.6)c 488.9 (4.9) 524.8 (4.4) 35.9 (5.3)c 0.04d Light intensity PA, minutes/day 371.5 (9.6) 327.5 (8.2) −44.0 (9.3)c 339.8 (5.2) 319.3 (5.6) −20.4 (4.9)c 370.4 (4.9) 343.0 (5.3) −27.4 (5.8)c 368.0 (4.2) 333.0 (4.3) −35.0 (4.2)c 0.054 MVPA, minutes/day 46.6 (3.8) 33.6 (2.6) −13.0 (3.8)c 44.8 (1.7) 37.7 (1.8) −7.1 (1.5)c 25.7 (1.2) 19.5 (1.2) −6.2 (1.3)c 35.5 (1.1) 28.7 (1.1) −6.8 (1.2)c 0.09 MVPA in bouts, minutes/dayh 15.9 (2.8) 14.3 (2.0) −1.6 (3.0) 14.6 (1.2) 19.3 (1.4) 4.7 (1.3)c 7.3 (0.8) 8.5 (1.0) 1.1 (0.9) 13.4 (0.8) 16.1 (1.0) 2.7 (0.9)i 0.04f Reported physical activityi, EU 483.2 (37.9) 392.9 (32.0) −90.4 (31.7)j 423.7 (17.2) 433.6 (18.4) 9.9 (15.2) 242.9 (14.8) 229.2 (14.3) −13.7 (13.4) 340.0 (12.8) 336.9 (14.4) −3.2 (10.6) 0.001d,e,f Variable . Black Men (n = 117; 12.2%), mean (SE) . White Men (n = 237; 24.6%), mean (SE) . Black Women (n = 244; 25.4%), mean (SE) . White Women (n = 364, 37.8%), mean (SE) . P Valuea . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . Wear time, minutes/day 906.3 (9.8) 909.9 (10.8) 3.7 (13.4) 905.1 (4.7) 902.3 (5.5) −2.8 (5.8) 880.1 (6.4) 889.2 (6.2) 9.1 (7.5) 892.4 (4.2) 886.5 (4.0) −5.8 (4.5) 0.34 Total accelerometer counts, per 10,000 counts per day 50.2 (9.7) 29.9 (1.2) −20.4 (9.7)b 36.4 (0.9) 32.2 (0.9) −4.3 (0.8)c 28.9 (0.7) 25.0 (0.7) −3.9 (0.7)c 33.3 (0.6) 28.6 (0.6) −4.7 (0.6) c 0.001d,e,f Average accelerometer counts, counts per minute/day 509.4 (76.5) 328.5 (13.8) −181.0 (77.0)b 402.6 (9.2) 355.5 (9.5) −47.0 (8.6)c 329.8 (7.9) 282.5 (7.5) −47.2 (7.7)c 373.8 (6.4) 321.2 (6.1) −52.6 (6.5)c 0.004d,e,f,g Sedentary time, minutes/day 488.2 (11.2) 548.8 (11.8) 60.6 (13.7)c 520.5 (6.2) 545.2 (6.7) 24.6 (6.8)c 484.0 (6.9) 526.7 (7.4) 42.7 (7.6)c 488.9 (4.9) 524.8 (4.4) 35.9 (5.3)c 0.04d Light intensity PA, minutes/day 371.5 (9.6) 327.5 (8.2) −44.0 (9.3)c 339.8 (5.2) 319.3 (5.6) −20.4 (4.9)c 370.4 (4.9) 343.0 (5.3) −27.4 (5.8)c 368.0 (4.2) 333.0 (4.3) −35.0 (4.2)c 0.054 MVPA, minutes/day 46.6 (3.8) 33.6 (2.6) −13.0 (3.8)c 44.8 (1.7) 37.7 (1.8) −7.1 (1.5)c 25.7 (1.2) 19.5 (1.2) −6.2 (1.3)c 35.5 (1.1) 28.7 (1.1) −6.8 (1.2)c 0.09 MVPA in bouts, minutes/dayh 15.9 (2.8) 14.3 (2.0) −1.6 (3.0) 14.6 (1.2) 19.3 (1.4) 4.7 (1.3)c 7.3 (0.8) 8.5 (1.0) 1.1 (0.9) 13.4 (0.8) 16.1 (1.0) 2.7 (0.9)i 0.04f Reported physical activityi, EU 483.2 (37.9) 392.9 (32.0) −90.4 (31.7)j 423.7 (17.2) 433.6 (18.4) 9.9 (15.2) 242.9 (14.8) 229.2 (14.3) −13.7 (13.4) 340.0 (12.8) 336.9 (14.4) −3.2 (10.6) 0.001d,e,f Abbreviations: BL, baseline; EU, exercise units; MVPA, moderate- to vigorous-intensity physical activity; PA, physical activity; SE, standard error. a For differences by race and sex groups based on analysis of variance. bP < 0.05. cP < 0.001. d Black men different than white men. e Black men different than black women. f Black men different than white women. g Black women different than white women (d through g based on Tukey studentized range (honestly significant difference) test for differences. h At least 8 of 10 consecutive minutes above the 1,952 counts per minute threshold. in = 956. jP < 0.01. Open in new tab Table 2. Baseline (2005–2006), 10-Year Follow-up (2015–2016), and 10-Year Change in Accelerometer-Determined Physical Activity and Sedentary Behavior Estimates Within and Stratified by Race and Sex Groups (n = 962), Coronary Artery Risk Development in Young Adults Study Variable . Black Men (n = 117; 12.2%), mean (SE) . White Men (n = 237; 24.6%), mean (SE) . Black Women (n = 244; 25.4%), mean (SE) . White Women (n = 364, 37.8%), mean (SE) . P Valuea . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . Wear time, minutes/day 906.3 (9.8) 909.9 (10.8) 3.7 (13.4) 905.1 (4.7) 902.3 (5.5) −2.8 (5.8) 880.1 (6.4) 889.2 (6.2) 9.1 (7.5) 892.4 (4.2) 886.5 (4.0) −5.8 (4.5) 0.34 Total accelerometer counts, per 10,000 counts per day 50.2 (9.7) 29.9 (1.2) −20.4 (9.7)b 36.4 (0.9) 32.2 (0.9) −4.3 (0.8)c 28.9 (0.7) 25.0 (0.7) −3.9 (0.7)c 33.3 (0.6) 28.6 (0.6) −4.7 (0.6) c 0.001d,e,f Average accelerometer counts, counts per minute/day 509.4 (76.5) 328.5 (13.8) −181.0 (77.0)b 402.6 (9.2) 355.5 (9.5) −47.0 (8.6)c 329.8 (7.9) 282.5 (7.5) −47.2 (7.7)c 373.8 (6.4) 321.2 (6.1) −52.6 (6.5)c 0.004d,e,f,g Sedentary time, minutes/day 488.2 (11.2) 548.8 (11.8) 60.6 (13.7)c 520.5 (6.2) 545.2 (6.7) 24.6 (6.8)c 484.0 (6.9) 526.7 (7.4) 42.7 (7.6)c 488.9 (4.9) 524.8 (4.4) 35.9 (5.3)c 0.04d Light intensity PA, minutes/day 371.5 (9.6) 327.5 (8.2) −44.0 (9.3)c 339.8 (5.2) 319.3 (5.6) −20.4 (4.9)c 370.4 (4.9) 343.0 (5.3) −27.4 (5.8)c 368.0 (4.2) 333.0 (4.3) −35.0 (4.2)c 0.054 MVPA, minutes/day 46.6 (3.8) 33.6 (2.6) −13.0 (3.8)c 44.8 (1.7) 37.7 (1.8) −7.1 (1.5)c 25.7 (1.2) 19.5 (1.2) −6.2 (1.3)c 35.5 (1.1) 28.7 (1.1) −6.8 (1.2)c 0.09 MVPA in bouts, minutes/dayh 15.9 (2.8) 14.3 (2.0) −1.6 (3.0) 14.6 (1.2) 19.3 (1.4) 4.7 (1.3)c 7.3 (0.8) 8.5 (1.0) 1.1 (0.9) 13.4 (0.8) 16.1 (1.0) 2.7 (0.9)i 0.04f Reported physical activityi, EU 483.2 (37.9) 392.9 (32.0) −90.4 (31.7)j 423.7 (17.2) 433.6 (18.4) 9.9 (15.2) 242.9 (14.8) 229.2 (14.3) −13.7 (13.4) 340.0 (12.8) 336.9 (14.4) −3.2 (10.6) 0.001d,e,f Variable . Black Men (n = 117; 12.2%), mean (SE) . White Men (n = 237; 24.6%), mean (SE) . Black Women (n = 244; 25.4%), mean (SE) . White Women (n = 364, 37.8%), mean (SE) . P Valuea . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . BL . Follow-up . 10-Year Change . Wear time, minutes/day 906.3 (9.8) 909.9 (10.8) 3.7 (13.4) 905.1 (4.7) 902.3 (5.5) −2.8 (5.8) 880.1 (6.4) 889.2 (6.2) 9.1 (7.5) 892.4 (4.2) 886.5 (4.0) −5.8 (4.5) 0.34 Total accelerometer counts, per 10,000 counts per day 50.2 (9.7) 29.9 (1.2) −20.4 (9.7)b 36.4 (0.9) 32.2 (0.9) −4.3 (0.8)c 28.9 (0.7) 25.0 (0.7) −3.9 (0.7)c 33.3 (0.6) 28.6 (0.6) −4.7 (0.6) c 0.001d,e,f Average accelerometer counts, counts per minute/day 509.4 (76.5) 328.5 (13.8) −181.0 (77.0)b 402.6 (9.2) 355.5 (9.5) −47.0 (8.6)c 329.8 (7.9) 282.5 (7.5) −47.2 (7.7)c 373.8 (6.4) 321.2 (6.1) −52.6 (6.5)c 0.004d,e,f,g Sedentary time, minutes/day 488.2 (11.2) 548.8 (11.8) 60.6 (13.7)c 520.5 (6.2) 545.2 (6.7) 24.6 (6.8)c 484.0 (6.9) 526.7 (7.4) 42.7 (7.6)c 488.9 (4.9) 524.8 (4.4) 35.9 (5.3)c 0.04d Light intensity PA, minutes/day 371.5 (9.6) 327.5 (8.2) −44.0 (9.3)c 339.8 (5.2) 319.3 (5.6) −20.4 (4.9)c 370.4 (4.9) 343.0 (5.3) −27.4 (5.8)c 368.0 (4.2) 333.0 (4.3) −35.0 (4.2)c 0.054 MVPA, minutes/day 46.6 (3.8) 33.6 (2.6) −13.0 (3.8)c 44.8 (1.7) 37.7 (1.8) −7.1 (1.5)c 25.7 (1.2) 19.5 (1.2) −6.2 (1.3)c 35.5 (1.1) 28.7 (1.1) −6.8 (1.2)c 0.09 MVPA in bouts, minutes/dayh 15.9 (2.8) 14.3 (2.0) −1.6 (3.0) 14.6 (1.2) 19.3 (1.4) 4.7 (1.3)c 7.3 (0.8) 8.5 (1.0) 1.1 (0.9) 13.4 (0.8) 16.1 (1.0) 2.7 (0.9)i 0.04f Reported physical activityi, EU 483.2 (37.9) 392.9 (32.0) −90.4 (31.7)j 423.7 (17.2) 433.6 (18.4) 9.9 (15.2) 242.9 (14.8) 229.2 (14.3) −13.7 (13.4) 340.0 (12.8) 336.9 (14.4) −3.2 (10.6) 0.001d,e,f Abbreviations: BL, baseline; EU, exercise units; MVPA, moderate- to vigorous-intensity physical activity; PA, physical activity; SE, standard error. a For differences by race and sex groups based on analysis of variance. bP < 0.05. cP < 0.001. d Black men different than white men. e Black men different than black women. f Black men different than white women. g Black women different than white women (d through g based on Tukey studentized range (honestly significant difference) test for differences. h At least 8 of 10 consecutive minutes above the 1,952 counts per minute threshold. in = 956. jP < 0.01. Open in new tab DISCUSSION Over 10 years, CARDIA participants experienced significant reductions in total and average accelerometer count estimates. The declines were primarily reflected as reductions in light-intensity physical activity (mean = 30.6 minutes/day) and approximately reciprocal increases in sedentary time (mean = 37.9 minutes/day). Minimal, yet significant, reductions in MVPA were also noted that support previous findings based on questionnaire responses (P < 0.001) (5–9). Significant differences in accelerometer counts, sedentary time, and MVPA in bouts were also noted over 10 years by race and sex groups (all P < 0.05), with black men having the greatest declines in average accelerometer counts compared with all other groups (P < 0.05). Differences in 10-year changes in accelerometer-determined physical activity and sedentary behavior were also noted by baseline (i.e., year 20 follow-up) age and cardiorespiratory fitness categories. The observed reductions in light-intensity physical activity across midlife are concerning, particularly within the context of increases in sedentary time (24). Although this evidence is still emerging (3), partly because of the increased capabilities and feasibility of implementing accelerometer-based measures in population-based research (23), the potential age-related health benefits of light-intensity physical activity have been demonstrated in studies conducted in older adults. More specifically, Buman et al. (25) found that replacing 30 minutes/day of sedentary time with light-intensity physical activity was associated with better reported physical health. Similarly, in older women, LaMonte et al. (26) found that greater amounts of light-intensity physical activity were associated with improvements in several cardiovascular risk factors (e.g., adiposity measures, triglyceride levels) and 10-year cardiovascular disease risk score. A significant inverse association of light-intensity physical activity with mortality risk was also found in prospective studies of older women (27) and men (28). This preliminary, supportive evidence leads to the overarching recommendation by the 2018 Physical Activity Guidelines Advisory Committee for additional research examining the role and contribution of light-intensity physical activity, alone or in combination with MVPA, relative to health outcomes (3). Given current study findings, the health consequences of age-related transitions from time spent in light-intensity physical activity to sedentary pursuits should also be considered. Regardless, as this evidence accumulates, future iterations of the PAG should consider sedentary and light-intensity physical activity targets, in addition to MVPA, for overall health benefit. Results of stratified analyses suggest important findings among CARDIA’s black participants. Although black men were the most active at baseline, black men subsequently had the greatest declines in accelerometer counts compared with all race and sex groups over 10 years. Conversely, the 10-year change profile among black women suggested consistently low levels of physical activity over time. This provides an opportunity to evaluate potential differences in subsequent health risk of individuals decreasing versus those with consistently low levels of physical activity over time. The magnitude of 10-year increases in sedentary time among blacks versus whites is also concerning, given emerging evidence supporting the associated health consequences (24). These specific physical activity behavioral-change profiles may contribute to the observed disparity in diabetes incidence (29) and cardiovascular disease–related death among blacks (30). Strengths of this study include a large biracial sample of men and women with repeated measures of accelerometry that span midlife. A limitation could be that due to continually emerging technology, different ActiGraph models were used. However, after calibration, measures were comparable for all summary estimates (19). Other limitations include possible misclassification of sedentary and light-intensity activity due to placement of the accelerometer at the hip, limited (or no) detection of certain activity types (e.g., bicycling, swimming), smaller sample size within some race or sex group strata, and no interim data collection time point. Finally, the participants represent a select sample, which limits generalizability to CARDIA and the broader US population. In summary, study findings complement those of previous studies documenting age-related decline in physical activity. Yet, our findings provide novel contextual information illustrating patterns of sedentary time and physical activity change by intensity category. To support intervention research, studies are needed to evaluate potential health consequences of age-related physical activity changes while also examining the social and health-related factors that contribute to these declines. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Human Genetics, and Environmental Sciences and Michael & Susan Dell Center for Healthy Living, UTHealth School of Public Health, Austin Campus, Austin, Texas (Kelley Pettee Gabriel, Raja I. Malkani); Dell Medical School, Department of Women’s Health, The University of Texas at Austin, Austin, Texas (Kelley Pettee Gabriel); Division of Research, Kaiser Permanente Northern California, Oakland, California (Stephen Sidney, Barbara Sternfeld); School of Public Health, Division of Epidemiology & Community Health, University of Minnesota, Minneapolis, Minnesota (David R. Jacobs, Jr., Pamela Schreiner); Division of Health and Human Physiology, University of Iowa, Iowa City, Iowa (Kara M. Whitaker); Feinberg School of Medicine, Department of Preventive Medicine, Northwestern University, Chicago, Illinois (Mercedes R. Carnethon); Department of Epidemiology, University of Alabama at Birmingham, Birmingham, Alabama (Cora E. Lewis); Department of Medicine, Division of Preventive Medicine, University of Alabama at Birmingham, Birmingham, Alabama (James M. Shikany); and Division of Cardiovascular Sciences, Program in Prevention and Population Sciences, National Heart, Lung, and Blood Institute, Bethesda, Maryland (Jared P. Reis). The Coronary Artery Risk Development in Young Adults (CARDIA) Study is supported by contracts HHSN268201300025C, HHSN268201300026C, HHSN268201300027C, HHSN268201300028C, HHSN268201300029C, and HHSN268200900041C from the National Heart, Lung, and Blood Institute (NHLBI), the Intramural Research Program of the National Institute on Aging (NIA), and an intra-agency agreement between the NIA and NHLBI (AG0005). Additional support for this work was provided by the CARDIA Fitness Study (grant R01 HL078972 to B.S. and S.S.) and CARDIA Activity Study (grant R56 HL125423 to K.P.G., B.S., S.S.). We thank the CARDIA Study participants. Conflict of interest: none declared. Abbreviations CARDIA Coronary Artery Risk Development in Young Adults MVPA moderate- to vigorous-intensity physical activity PAG physical activity guidelines SE standard error. REFERENCES 1 US Department of Health and Human Services . 2008 Physical Activity Guidelines for Americans. 2008 . www.health.gov/paguidelines. Accessed October 10, 2008. 2 Lloyd-Jones DM , Hong Y, Labarthe D, et al. . Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association’s strategic Impact Goal through 2020 and beyond . Circulation . 2010 ; 121 ( 4 ): 586 – 613 . Google Scholar Crossref Search ADS PubMed WorldCat 3 US Department of Health and Human Services . 2018 Physical Activity Guidelines Advisory Committee Scientific Report. 2018 . https://health.gov/paguidelines/second-edition/report/pdf/PAG_Advisory_Committee_Report.pdf. Accessed March 12, 2018. 4 US Department of Health and Human Services . Physical Activity Guidelines Advisory Committee Report, 2008. 2008 . https://health.gov/paguidelines/report/pdf/CommitteeReport.pdf. Accessed January 29, 2018. 5 Sternfeld B , Sidney S, Jacobs DR , Jr., et al. . Seven-year changes in physical fitness, physical activity, and lipid profile in the CARDIA study. Coronary Artery Risk Development in Young Adults . Ann Epidemiol . 1999 ; 9 ( 1 ): 25 – 33 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Andersen LB , Haraldsdóttir J. Tracking of cardiovascular disease risk factors including maximal oxygen uptake and physical activity from late teenage to adulthood. An 8-year follow-up study . J Intern Med . 1993 ; 234 ( 3 ): 309 – 315 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Schmitz KH , Jacobs DR , Jr., Leon AS, et al. . Physical activity and body weight: associations over ten years in the CARDIA study. Coronary Artery Risk Development in Young Adults . Int J Obes Relat Metab Disord . 2000 ; 24 ( 11 ): 1475 – 1487 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Hankinson AL , Daviglus ML, Bouchard C, et al. . Maintaining a high physical activity level over 20 years and weight gain . JAMA . 2010 ; 304 ( 23 ): 2603 – 2610 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Parker ED , Schmitz KH, Jacobs DR , Jr., et al. . Physical activity in young adults and incident hypertension over 15 years of follow-up: the CARDIA study . Am J Public Health . 2007 ; 97 ( 4 ): 703 – 709 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Daley MJ , Spinks WL. Exercise, mobility and aging . Sports Med . 2000 ; 29 ( 1 ): 1 – 12 . Google Scholar Crossref Search ADS PubMed WorldCat 11 Martin PE , Morgan DW. Biomechanical considerations for economical walking and running . Med Sci Sports Exerc . 1992 ; 24 ( 4 ): 467 – 474 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Merrill SS , Verbrugge LM. Health and disease in midlife. In: Willis S and Reid J, eds. Life in the Middle . New York, NY : Academic Press ; 1999 . 77 – 103 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 13 Evenson KR , Rosamond WD, Cai J, et al. . Influence of retirement on leisure-time physical activity: the atherosclerosis risk in communities study . Am J Epidemiol . 2002 ; 155 ( 8 ): 692 – 699 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Jacobs DR , Jr., Hahn LP, Haskell WL, et al. . Validity and reliability of short physical activity history: Cardia and the Minnesota Heart Health Program . J Cardiopulm Rehabil . 1989 ; 9 ( 11 ): 448 – 459 . Google Scholar Crossref Search ADS PubMed WorldCat 15 Gabriel KP , Sidney S, Jacobs DR , Jr., et al. . Convergent validity of a brief self-reported physical activity questionnaire . Med Sci Sports Exerc . 2014 ; 46 ( 8 ): 1570 – 1577 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Sidney S , Haskell WL, Crow R, et al. . Symptom-limited graded treadmill exercise testing in young adults in the CARDIA study . Med Sci Sports Exerc . 1992 ; 24 ( 2 ): 177 – 183 . Google Scholar Crossref Search ADS PubMed WorldCat 17 Troiano RP , Berrigan D, Dodd KW, et al. . Physical activity in the United States measured by accelerometer . Med Sci Sports Exerc . 2008 ; 40 ( 1 ): 181 – 188 . Google Scholar Crossref Search ADS PubMed WorldCat 18 Cain KL , Conway TL, Adams MA, et al. . Comparison of older and newer generations of ActiGraph accelerometers with the normal filter and the low frequency extension . Int J Behav Nutr Phys Act . 2013 ; 10 : 51 . Google Scholar Crossref Search ADS PubMed WorldCat 19 Whitaker KM , Pettee Gabriel K, Jacobs DR , Jr., et al. . Comparison of two generations of ActiGraph accelerometers: the CARDIA study . Med Sci Sports Exerc . 2018 ; 50 ( 6 ): 1333 – 1340 . Google Scholar Crossref Search ADS PubMed WorldCat 20 Freedson PS , Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer . Med Sci Sports Exerc . 1998 ; 30 ( 5 ): 777 – 781 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Troiano RP , Pettee Gabriel KK, Welk GJ, et al. . Reported physical activity and sedentary behavior: why do you ask? J Phys Act Health . 2012 ; 9 ( suppl 1 ): S68 – S75 . Google Scholar Crossref Search ADS PubMed WorldCat 22 Pettee Gabriel KK , Morrow JR , Jr., Woolsey AL. Framework for physical activity as a complex and multidimensional behavior . J Phys Act Health . 2012 ; 9 ( suppl 1 ): S11 – S18 . Google Scholar Crossref Search ADS PubMed WorldCat 23 Matthews CE , Hagströmer M, Pober DM, et al. . Best practices for using physical activity monitors in population-based research . Med Sci Sports Exerc . 2012 ; 44 ( 1 suppl 1 ): S68 – S76 . Google Scholar Crossref Search ADS PubMed WorldCat 24 Young DR , Hivert MF, Alhassan S, et al. . Sedentary behavior and cardiovascular morbidity and mortality: a Science Advisory from the American Heart Association . Circulation . 2016 ; 134 ( 13 ): e262 – e279 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 25 Buman MP , Hekler EB, Haskell WL, et al. . Objective light-intensity physical activity associations with rated health in older adults . Am J Epidemiol . 2010 ; 172 ( 10 ): 1155 – 1165 . Google Scholar Crossref Search ADS PubMed WorldCat 26 LaMonte MJ , Lewis CE, Buchner DM, et al. . Both light intensity and moderate-to-vigorous physical activity measured by accelerometry are favorably associated with cardiometabolic risk factors in older women: the Objective Physical Activity and Cardiovascular Health (OPACH) Study . J Am Heart Assoc . 2017 ; 6 ( 10 ): e007064 . Google Scholar Crossref Search ADS PubMed WorldCat 27 LaMonte MJ , Buchner DM, Rillamas-Sun E, et al. . Accelerometer-measured physical activity and mortality in women aged 63 to 99 . J Am Geriatr Soc . 2018 ; 66 ( 5 ): 886 – 894 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Ensrud KE , Blackwell TL, Cauley JA, et al. . Objective measures of activity level and mortality in older men . J Am Geriatr Soc . 2014 ; 62 ( 11 ): 2079 – 2087 . Google Scholar Crossref Search ADS PubMed WorldCat 29 Bancks MP , Kershaw K, Carson AP, et al. . Association of modifiable risk factors in young adulthood with racial disparity in incident type 2 diabetes during middle adulthood . JAMA . 2017 ; 318 ( 24 ): 2457 – 2465 . Google Scholar Crossref Search ADS PubMed WorldCat 30 Carnethon MR , Pu J, Howard G, et al. . Cardiovascular health in African Americans: a scientific statement from the American Heart Association . Circulation . 2017 ; 136 ( 21 ): e393 – e423 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018.

Patel, Alpa V; Maliniak, Maret L; Rees-Punia, Erika; Matthews, Charles E; Gapstur, Susan M

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy125pmid: 29947736

Abstract The majority of leisure time is spent in sedentary behaviors such as television viewing. Studies have documented that prolonged leisure-time sitting is associated with higher risk of mortality—total, cardiovascular disease, cancer, and “all other causes”—but few have examined the “other” causes of death in detail. To examine associations of leisure-time sitting with risk of specific causes of death, we analyzed data from the Cancer Prevention Study II (CPS-II) Nutrition Cohort, a prospective US cohort including 127,554 men and women who were free of major chronic disease at study entry, and among whom 48,784 died during 21 years of follow-up (1993–2014; median follow-up, 20.3 years, interquartile range, 4.6 years). After multivariable adjustment, prolonged leisure-time sitting (≥6 vs. <3 hours per day) was associated with higher risk of mortality from all causes, cardiovascular disease (including coronary heart disease and stroke-specific mortality), cancer, diabetes, kidney disease, suicide, chronic obstructive pulmonary disease, pneumonitis due to solids and liquids, liver, peptic ulcer and other digestive disease, Parkinson disease, Alzheimer disease, nervous disorders, and musculoskeletal disorders. These findings provide additional evidence for associations between a broad range of mortality outcomes and prolonged sitting time. Given the pervasive nature of sitting in the contemporary lifestyle, this study further supports the recommendation that encouraging individuals to reduce sedentary time may provide health benefits. cohort, mortality, physical activity, sedentary behavior, sitting time With technological advancements, the amount of time spent sitting has increased significantly over the past few decades. In an Australian time-use study, investigators estimated that 90% of total nonoccupational time was spent sedentary and, of that, 53% was spent on screen time (computer or television) (1). Furthermore, as individuals age, physical activity decreases while sedentary behavior and risk of chronic disease increase (2). Regardless of sedentary time metric (i.e., total, leisure time, or television viewing), numerous studies have shown that prolonged time spent sitting is associated with total mortality (3–6), type II diabetes (7), cardiovascular disease (8), and some cancers (9, 10), independent of moderate-to-vigorous intensity physical activity (MVPA). Emerging evidence supports that sitting time is a behavioral risk factor that is distinct from inadequate exercise (i.e., physical inactivity) and could be an important additional target for intervention in the effort to increase daily physical activity in the population (11). In a large meta-analysis (4), investigators reported a 34% higher total mortality risk for adults sitting 10 hours per day compared with 1 hour, after adjustment for physical activity and other potential confounders. Various studies have examined specific grouped causes of death and found associations with cardiovascular disease mortality and, to a lesser extent, with cancer mortality, as well as a statistically significant excess risk of death from “all other causes” (5, 6). The largest prospective study to date, using data from the NIH-AARP Diet and Health Study, further examined the association with “all other causes” to explore the specific causes of death that were driving the excess risk (12). Prolonged television viewing was associated with a statistically significant higher risk (per 2-hour/day increase in television-viewing time) of death from cancer, coronary heart disease, chronic obstructive pulmonary disease (COPD), diabetes, influenza/pneumonia, Parkinson disease, liver disease, and suicide (12). To our knowledge, no other study has comprehensively examined sitting time in relation to death from specific causes beyond cancer and cardiovascular disease mortality. In an earlier analysis using data from the American Cancer Society’s Cancer Prevention Study II (CPS-II) Nutrition Cohort (5), leisure time spent sitting (≥6 vs. <3 hours per day) was associated with cardiovascular disease mortality among women (relative risk (RR) = 1.33, 95% confidence interval (CI): 1.17, 1.52) and men (RR = 1.18, 95% CI: 1.08, 1.30) and with cancer mortality among women (RR = 1.30, 95% CI: 1.16, 1.46) but not men (RR = 1.04, 95% CI: 0.94, 1.15). Prolonged leisure time spent sitting was also associated with other causes of death (excluding cardiovascular disease and cancer) among women (RR = 1.41, 95% CI: 1.25, 1.60) and men (RR = 1.33, 95% CI: 1.20, 1.47). To further examine the associations of sitting time with risk of specific causes of death, a detailed analysis was conducted using data from the CPS-II Nutrition Cohort, a large prospective US cohort study of men and women. This cohort provides a tremendous opportunity to add to the singular previous study on the association between sitting time and cause-specific mortality due to its large sample size and number of deaths along with detailed exposure assessment. METHODS Study population Men and women in this analysis were drawn from the 184,185 participants in the CPS-II Nutrition Cohort, a prospective study of cancer incidence and mortality in the United States initiated by the American Cancer Society in 1992 (13). The CPS-II Nutrition Cohort is a subgroup of the approximately 1.2 million men and women in the CPS-II Mortality Cohort, a prospective study established by the American Cancer Society in 1982 (14). CPS-II Mortality Cohort participants, who resided in 21 states with population-based state tumor registries and who were 50–74 years of age in 1992, were invited to participate by completing a 10-page self-administered, mailed questionnaire that included questions on demographic, medical, behavioral, and lifestyle factors. The recruitment and characteristics of the CPS-II Nutrition Cohort are described in detail elsewhere (13), and all aspects of the study were approved by Emory University Institutional Review Board. Men and women were excluded from this analysis if they reported a personal history of cancer (n = 21,785), heart attack (n = 11,559), stroke (n = 2,513), or emphysema/lung disease (n = 9,321) at the time of enrollment; were missing sitting time (n = 2,953) or recreational physical activity (n = 4,240) data; had extreme (top and bottom 0.1%) or missing body mass index (BMI) (n = 2,121); or were missing smoking status (n = 863). Men and women who reported no physical activity in their daily life (defined as no recreational physical activity, other daily life/household activity, or light housekeeping) were also excluded (n = 871) due to the strong possibility of an underlying condition that might be causing their complete inactivity. Finally, we excluded deaths in the first year of follow-up (n = 405). After exclusions, 127,554 men and women were included in the analysis, among whom 48,784 died between baseline and the end of follow-up on December 31, 2014. Measures of time spent sitting and physical activity Time spent sitting was assessed using the baseline question “During the past year, on an average day (not counting time spent at your job), how many hours per day did you spend sitting (watching television, reading, etc.)?” Responses included none, <3, 3–5, 6–8, or >8 hours per day. For the present study, time spent sitting was categorized as <3 hours per day (reference group), 3–5 hours per day, ≥6 hours per day. Information on recreational MVPA was collected using the question “During the past year, what was the average time per week you spent at the following kinds of activities: walking, jogging/running, lap swimming, tennis or racquetball, bicycling or stationary biking, aerobics/calisthenics, and dancing?” Responses to each individual activity included none, 1–3, 4–6, or ≥7 hours/week. A MET, or metabolic equivalent of task—a measure of energy expenditure—is estimated by dividing the energy cost of a given activity by resting energy expenditure (15); summary MET-hours/week were calculated for each participant. The summary MET score for each participant was calculated by multiplying the lowest number of hours within each category by the general MET level of each activity according to the Compendium of Physical Activities to provide conservative summary measures, addressing the likelihood of overreporting physical activity and the older age of study participants (15). The MET scores assigned for various activities included 3.5 for walking, 7.0 for jogging/running, 7.0 for lap swimming, 6.0 for tennis or racquetball, 4.0 for bicycling/stationary biking, 4.5 for aerobics/calisthenics, and 3.5 for dancing. MVPA MET-hours/week was controlled for as a continuous measure in all analyses. Mortality endpoints The primary endpoint was death from any cause occurring between 1 year after the time of enrollment and December 31, 2014. Deaths were identified through biennial automated linkage of the entire cohort with the National Death Index (16). Causes of death were classified by using the International Classification of Diseases, Ninth Revision (17), for deaths occurring from 1992 to 1998 and the Tenth Revision (18) for deaths from 1999 to 2012. The following specific causes of death were examined: cancer, cardiovascular disease (grouped, as well as coronary heart disease, stroke, and other circulatory diseases separately), diabetes, other nutritional/metabolic diseases, kidney disease, other genitourinary diseases, accidents, suicide, COPD, pneumonia and influenza, pneumonitis due to solids and liquids, other respiratory diseases, liver disease, peptic ulcer and other digestive diseases, infectious diseases, dementia/mental disorders, Parkinson disease, Alzheimer disease, nervous disorders, and musculoskeletal disorders. All remaining deaths were collapsed into an “ill-defined” or “all other causes” categories. Table 1 summarizes the classification of grouped cause of death according to International Classification of Diseases codes. Table 1. Diagnosis Codes Used to Classify Specific Causes of Death, Cancer Prevention Study II Nutrition Cohort, United States, 1992–2014 Cause of Death . ICD-9 . ICD-10 . Cancer 140–239 C00–C97, D00–D49 All cardiovascular disease 390–459 I00–I99  Coronary heart disease 410–414 I20–I25  Stroke 430–438 I60–I69  Other circulatory diseases 390–405, 415–429, 440–459 I00–I19, I26–I59, I70–I99 Diabetes 250 E10–E14 Endocrine, nutritional, and metabolic diseases 240–249, 251–289 D50–E09, E15–E88 Kidney disease 580–589 N00–N07, N17–N19, N25–N27 Other genitourinary disease 590–629 N08–N16, N20–N24, N28–N39, N40–N98 Accidents E800–E929 V01–X59, Y85–Y86 Suicide 950–959 X60–X84, Y87.0 Chronic obstructive pulmonary disease 490–496 J40–J45, J47 Pneumonia and influenza 480–487 J09–J18 Pneumonitis due to solids and liquids 507 J69 Other respiratory diseases 460–466, 470–478, 495, 500–506, 508–519 J00–J08, J20–J39, J60–J68, J70–J99 Liver disease 571 K70, K73–K74 Peptic ulcer and other digestive diseases 520–537, 540–543, 550–553, 555–570, 572–579 K25–K28, K00–K24, K29–K66, K71–K72, K75–K99 Infectious and parasitic diseases (including HIV and pulmonary tuberculosis) 001–139 A00–B99 Dementia/mental disorders 290–319 F00–F99 Parkinson disease 332 G20–G21 Alzheimer disease 331.0 G30 Nervous system and sense disorders 320–330, 331.1–331.9, 333–389 G00–G19, G22–29, G31–H93 Musculoskeletal diseases 710–739 M00–M99 Ill-defined 780–799 R00–R99 Cause of Death . ICD-9 . ICD-10 . Cancer 140–239 C00–C97, D00–D49 All cardiovascular disease 390–459 I00–I99  Coronary heart disease 410–414 I20–I25  Stroke 430–438 I60–I69  Other circulatory diseases 390–405, 415–429, 440–459 I00–I19, I26–I59, I70–I99 Diabetes 250 E10–E14 Endocrine, nutritional, and metabolic diseases 240–249, 251–289 D50–E09, E15–E88 Kidney disease 580–589 N00–N07, N17–N19, N25–N27 Other genitourinary disease 590–629 N08–N16, N20–N24, N28–N39, N40–N98 Accidents E800–E929 V01–X59, Y85–Y86 Suicide 950–959 X60–X84, Y87.0 Chronic obstructive pulmonary disease 490–496 J40–J45, J47 Pneumonia and influenza 480–487 J09–J18 Pneumonitis due to solids and liquids 507 J69 Other respiratory diseases 460–466, 470–478, 495, 500–506, 508–519 J00–J08, J20–J39, J60–J68, J70–J99 Liver disease 571 K70, K73–K74 Peptic ulcer and other digestive diseases 520–537, 540–543, 550–553, 555–570, 572–579 K25–K28, K00–K24, K29–K66, K71–K72, K75–K99 Infectious and parasitic diseases (including HIV and pulmonary tuberculosis) 001–139 A00–B99 Dementia/mental disorders 290–319 F00–F99 Parkinson disease 332 G20–G21 Alzheimer disease 331.0 G30 Nervous system and sense disorders 320–330, 331.1–331.9, 333–389 G00–G19, G22–29, G31–H93 Musculoskeletal diseases 710–739 M00–M99 Ill-defined 780–799 R00–R99 Abbreviations: HIV, human immunodeficiency virus; ICD-9, International Classification of Diseases, Ninth Revision; ICD-10, International Classification of Diseases, Tenth Revision. Open in new tab Table 1. Diagnosis Codes Used to Classify Specific Causes of Death, Cancer Prevention Study II Nutrition Cohort, United States, 1992–2014 Cause of Death . ICD-9 . ICD-10 . Cancer 140–239 C00–C97, D00–D49 All cardiovascular disease 390–459 I00–I99  Coronary heart disease 410–414 I20–I25  Stroke 430–438 I60–I69  Other circulatory diseases 390–405, 415–429, 440–459 I00–I19, I26–I59, I70–I99 Diabetes 250 E10–E14 Endocrine, nutritional, and metabolic diseases 240–249, 251–289 D50–E09, E15–E88 Kidney disease 580–589 N00–N07, N17–N19, N25–N27 Other genitourinary disease 590–629 N08–N16, N20–N24, N28–N39, N40–N98 Accidents E800–E929 V01–X59, Y85–Y86 Suicide 950–959 X60–X84, Y87.0 Chronic obstructive pulmonary disease 490–496 J40–J45, J47 Pneumonia and influenza 480–487 J09–J18 Pneumonitis due to solids and liquids 507 J69 Other respiratory diseases 460–466, 470–478, 495, 500–506, 508–519 J00–J08, J20–J39, J60–J68, J70–J99 Liver disease 571 K70, K73–K74 Peptic ulcer and other digestive diseases 520–537, 540–543, 550–553, 555–570, 572–579 K25–K28, K00–K24, K29–K66, K71–K72, K75–K99 Infectious and parasitic diseases (including HIV and pulmonary tuberculosis) 001–139 A00–B99 Dementia/mental disorders 290–319 F00–F99 Parkinson disease 332 G20–G21 Alzheimer disease 331.0 G30 Nervous system and sense disorders 320–330, 331.1–331.9, 333–389 G00–G19, G22–29, G31–H93 Musculoskeletal diseases 710–739 M00–M99 Ill-defined 780–799 R00–R99 Cause of Death . ICD-9 . ICD-10 . Cancer 140–239 C00–C97, D00–D49 All cardiovascular disease 390–459 I00–I99  Coronary heart disease 410–414 I20–I25  Stroke 430–438 I60–I69  Other circulatory diseases 390–405, 415–429, 440–459 I00–I19, I26–I59, I70–I99 Diabetes 250 E10–E14 Endocrine, nutritional, and metabolic diseases 240–249, 251–289 D50–E09, E15–E88 Kidney disease 580–589 N00–N07, N17–N19, N25–N27 Other genitourinary disease 590–629 N08–N16, N20–N24, N28–N39, N40–N98 Accidents E800–E929 V01–X59, Y85–Y86 Suicide 950–959 X60–X84, Y87.0 Chronic obstructive pulmonary disease 490–496 J40–J45, J47 Pneumonia and influenza 480–487 J09–J18 Pneumonitis due to solids and liquids 507 J69 Other respiratory diseases 460–466, 470–478, 495, 500–506, 508–519 J00–J08, J20–J39, J60–J68, J70–J99 Liver disease 571 K70, K73–K74 Peptic ulcer and other digestive diseases 520–537, 540–543, 550–553, 555–570, 572–579 K25–K28, K00–K24, K29–K66, K71–K72, K75–K99 Infectious and parasitic diseases (including HIV and pulmonary tuberculosis) 001–139 A00–B99 Dementia/mental disorders 290–319 F00–F99 Parkinson disease 332 G20–G21 Alzheimer disease 331.0 G30 Nervous system and sense disorders 320–330, 331.1–331.9, 333–389 G00–G19, G22–29, G31–H93 Musculoskeletal diseases 710–739 M00–M99 Ill-defined 780–799 R00–R99 Abbreviations: HIV, human immunodeficiency virus; ICD-9, International Classification of Diseases, Ninth Revision; ICD-10, International Classification of Diseases, Tenth Revision. Open in new tab Statistical analyses Deaths and person-years were calculated according to category of leisure time spent sitting and 5-year attained-age groups, and mortality rates were then standardized to the age distribution of the CPS-II Nutrition Cohort population. Cox proportional hazards regression modeling (19) was used to compute relative risks and 95% confidence intervals, with follow-up time in days as the time axis. For leisure time spent sitting, we assessed risk in 3 models: 1) adjusted only for age (single year of age) and sex; 2) adjusted for age, sex, and other potential confounding factors; and 3) additionally adjusted for MVPA MET-hours per week. The potential confounders included were race (white, black, other/unknown), education (less than high school graduate, high school graduate or some college, college graduate or higher, unknown), smoking status (never, current, former, ever but status unknown), frequency and duration of smoking among current smokers (<20 cigarettes per day and smoking ≤35 years, <20 cigarettes per day and smoking >35 years, ≥20 cigarettes per day and smoking ≤35 years, ≥20 cigarettes per day and smoking >35 years), years since quitting among former smokers (<10, 10–19, ≥20), body mass index (calculated as weight (kg)/height (m)2) (continuous), marital status (married, widowed or divorced, never married, unknown), aspirin use (pills per month: 0, >1 and <15, 15 to <30, ≥30), alcohol consumption (drinks per day: 0, >0 and <1, 1, ≥2, unknown), occupational status (not employed/retired, employed, unknown), American Cancer Society dietary guidelines adherence score (20) (<3, 3 to <6, ≥6, unknown), and comorbidity score (0, 1, or ≥2 comorbidities; includes high blood pressure, diabetes, and high cholesterol). For some causes of death, categories of covariates were sometimes collapsed due to small numbers. Secondary analyses were conducted to test for effect modification between leisure time spent sitting and sex, MVPA, BMI, and employment status (retired/homemaker vs. employed). We also conducted a sensitivity analysis among men and women who were lifelong nonsmokers or former smokers who quit more than 15 years prior to baseline. We also examined associations for diabetes mortality, excluding prevalent diabetes at baseline, and for kidney disease mortality, excluding prevalent diabetes and history of kidney stones at baseline. While sitting time was queried on some subsequent follow-up surveys, we did not update this exposure because of the possibility of reverse causation given that individuals might become more sedentary with age and/or development of disease, leading to a potential bias away from the null. However, to further address the possibility of reverse causality, we examined the association between baseline sitting time and mortality stratified by follow-up time. Weighted Schoenfeld residuals were used to test the Cox proportional hazards assumption (21). All tests for statistical significance were 2-sided, and P < 0.05 was considered statistically significant. All analyses were conducted using R (R Foundation for Statistical Computing, Vienna, Austria) (22). RESULTS Over 2,293,860 person-years of follow-up (median, 20.3 years, interquartile range, 4.6 years), we observed 48,784 total deaths. The most common cause of death was cardiovascular disease (n = 16,083), followed by cancer (n = 14,550), dementia/mental disorders (n = 2,406), Alzheimer disease (n = 2,248), COPD (n = 1,642), accidents (n = 1,339), Parkinson disease (n = 1,153), peptic ulcer and other digestive diseases (n = 1,148), and pneumonia/influenza (n = 1,034). Participants who spent the most leisure time sitting were slightly older, had a higher BMI, and were more commonly retired/unemployed (Table 2). Additionally, the most sedentary participants more commonly had ever smoked or had more comorbidities (diabetes, hypertension, or hyperlipidemia) and less commonly followed American Cancer Society dietary guidelines. Table 2. Selected Baseline Characteristics According to Hours of Leisure Time Spent Sitting for Men and Women, Cancer Prevention Study II Nutrition Cohort, United States, 1992 Characteristic . Leisure Time Spent Sitting . <3 hours/day (n = 58,910) . 3–5 hours/day (n = 54,742) . ≥6 hours/day (n = 13,902) . No. . %a . No. . %a . No. . %a . Age, yearsb 61.5 (6.3) 63.4 (6.2) 64.2 (6.3) BMIb,c 25.5 (4.0) 26.2 (4.2) 26.8 (4.6) MVPA, MET-hours/weekb 13.1 (13.2) 12.4 (12.5) 12.8 (13.1) Male sex 24,300 41.2 24,951 45.6 7,356 52.9 Employment status  Not employed or retired 20,935 35.5 27,880 50.9 8,103 58.3  Employed 35,035 59.5 23,816 43.5 4,970 35.8 Race  White 57,339 97.3 53,328 97.4 13,528 97.3  Black 770 1.3 748 1.4 183 1.3  Other 801 1.4 666 1.2 191 1.4 Education  Less than high school 3,270 5.6 3,245 5.9 937 6.7  High school graduate 31,387 53.3 30,954 56.5 7,280 52.4  College graduate or higher 23,886 40.5 20,167 36.8 5,598 40.3 Smoking status  Never 29,990 50.9 24,004 43.8 5,524 39.7  Current 4,142 7.0 5,190 9.5 1,692 12.2  Former 24,540 41.7 25,368 46.3 6,631 47.7  Ever or unknown status 238 0.4 180 0.3 55 0.4 Alcohol intake, no. of drinks/day  0 23,208 39.4 21,103 38.5 5,502 39.6  <1 23,357 39.6 21,430 39.1 5,003 36.0  1 5,861 9.9 5,662 10.3 1,403 10.1  ≥2 4,281 7.3 4,669 8.5 1,431 10.3 Marital status  Married 52,601 89.3 48,415 88.4 12,115 87.1  Widowed or divorced 4,963 8.4 5,001 9.1 1,363 9.8  Never married 828 1.4 847 1.5 252 1.8 Comorbidity scored  0 28,317 48.1 22,752 41.6 5,475 39.4  1 21,626 36.7 21,264 38.8 5,392 38.8  ≥2 8,967 15.2 10,726 19.6 3,035 21.8 Diet scoree  0–2 11,933 20.3 13,502 24.7 3,757 27.0  3–5 25,836 43.9 24,600 44.9 5,996 43.1  6–9 16,228 27.5 12,441 22.7 2,881 20.7 Aspirin use, no. of pills/month  0 33,896 57.5 30,235 55.2 7,444 53.5  1–14 8,361 14.2 7,527 13.7 1,798 12.9  15–29 4,997 8.5 4,736 8.7 1,207 8.7  ≥30 10,176 17.3 10,858 19.8 3,070 22.1 Characteristic . Leisure Time Spent Sitting . <3 hours/day (n = 58,910) . 3–5 hours/day (n = 54,742) . ≥6 hours/day (n = 13,902) . No. . %a . No. . %a . No. . %a . Age, yearsb 61.5 (6.3) 63.4 (6.2) 64.2 (6.3) BMIb,c 25.5 (4.0) 26.2 (4.2) 26.8 (4.6) MVPA, MET-hours/weekb 13.1 (13.2) 12.4 (12.5) 12.8 (13.1) Male sex 24,300 41.2 24,951 45.6 7,356 52.9 Employment status  Not employed or retired 20,935 35.5 27,880 50.9 8,103 58.3  Employed 35,035 59.5 23,816 43.5 4,970 35.8 Race  White 57,339 97.3 53,328 97.4 13,528 97.3  Black 770 1.3 748 1.4 183 1.3  Other 801 1.4 666 1.2 191 1.4 Education  Less than high school 3,270 5.6 3,245 5.9 937 6.7  High school graduate 31,387 53.3 30,954 56.5 7,280 52.4  College graduate or higher 23,886 40.5 20,167 36.8 5,598 40.3 Smoking status  Never 29,990 50.9 24,004 43.8 5,524 39.7  Current 4,142 7.0 5,190 9.5 1,692 12.2  Former 24,540 41.7 25,368 46.3 6,631 47.7  Ever or unknown status 238 0.4 180 0.3 55 0.4 Alcohol intake, no. of drinks/day  0 23,208 39.4 21,103 38.5 5,502 39.6  <1 23,357 39.6 21,430 39.1 5,003 36.0  1 5,861 9.9 5,662 10.3 1,403 10.1  ≥2 4,281 7.3 4,669 8.5 1,431 10.3 Marital status  Married 52,601 89.3 48,415 88.4 12,115 87.1  Widowed or divorced 4,963 8.4 5,001 9.1 1,363 9.8  Never married 828 1.4 847 1.5 252 1.8 Comorbidity scored  0 28,317 48.1 22,752 41.6 5,475 39.4  1 21,626 36.7 21,264 38.8 5,392 38.8  ≥2 8,967 15.2 10,726 19.6 3,035 21.8 Diet scoree  0–2 11,933 20.3 13,502 24.7 3,757 27.0  3–5 25,836 43.9 24,600 44.9 5,996 43.1  6–9 16,228 27.5 12,441 22.7 2,881 20.7 Aspirin use, no. of pills/month  0 33,896 57.5 30,235 55.2 7,444 53.5  1–14 8,361 14.2 7,527 13.7 1,798 12.9  15–29 4,997 8.5 4,736 8.7 1,207 8.7  ≥30 10,176 17.3 10,858 19.8 3,070 22.1 Abbreviations: BMI, body mass index; MET, metabolic equivalent of task; MVPA, recreational moderate-to-vigorous physical activity. a Percentages may not sum to due to rounding or missing values for exposure. b Values are expressed as mean (standard deviation). c Weight (kg)/height (m)2. d Based on self-reported diabetes, hypertension, and high cholesterol. e Calculated based on intake of fruits and vegetables, whole/refined grains, and red and processed meat, with a score of 9 representing optimal dietary adherence. Open in new tab Table 2. Selected Baseline Characteristics According to Hours of Leisure Time Spent Sitting for Men and Women, Cancer Prevention Study II Nutrition Cohort, United States, 1992 Characteristic . Leisure Time Spent Sitting . <3 hours/day (n = 58,910) . 3–5 hours/day (n = 54,742) . ≥6 hours/day (n = 13,902) . No. . %a . No. . %a . No. . %a . Age, yearsb 61.5 (6.3) 63.4 (6.2) 64.2 (6.3) BMIb,c 25.5 (4.0) 26.2 (4.2) 26.8 (4.6) MVPA, MET-hours/weekb 13.1 (13.2) 12.4 (12.5) 12.8 (13.1) Male sex 24,300 41.2 24,951 45.6 7,356 52.9 Employment status  Not employed or retired 20,935 35.5 27,880 50.9 8,103 58.3  Employed 35,035 59.5 23,816 43.5 4,970 35.8 Race  White 57,339 97.3 53,328 97.4 13,528 97.3  Black 770 1.3 748 1.4 183 1.3  Other 801 1.4 666 1.2 191 1.4 Education  Less than high school 3,270 5.6 3,245 5.9 937 6.7  High school graduate 31,387 53.3 30,954 56.5 7,280 52.4  College graduate or higher 23,886 40.5 20,167 36.8 5,598 40.3 Smoking status  Never 29,990 50.9 24,004 43.8 5,524 39.7  Current 4,142 7.0 5,190 9.5 1,692 12.2  Former 24,540 41.7 25,368 46.3 6,631 47.7  Ever or unknown status 238 0.4 180 0.3 55 0.4 Alcohol intake, no. of drinks/day  0 23,208 39.4 21,103 38.5 5,502 39.6  <1 23,357 39.6 21,430 39.1 5,003 36.0  1 5,861 9.9 5,662 10.3 1,403 10.1  ≥2 4,281 7.3 4,669 8.5 1,431 10.3 Marital status  Married 52,601 89.3 48,415 88.4 12,115 87.1  Widowed or divorced 4,963 8.4 5,001 9.1 1,363 9.8  Never married 828 1.4 847 1.5 252 1.8 Comorbidity scored  0 28,317 48.1 22,752 41.6 5,475 39.4  1 21,626 36.7 21,264 38.8 5,392 38.8  ≥2 8,967 15.2 10,726 19.6 3,035 21.8 Diet scoree  0–2 11,933 20.3 13,502 24.7 3,757 27.0  3–5 25,836 43.9 24,600 44.9 5,996 43.1  6–9 16,228 27.5 12,441 22.7 2,881 20.7 Aspirin use, no. of pills/month  0 33,896 57.5 30,235 55.2 7,444 53.5  1–14 8,361 14.2 7,527 13.7 1,798 12.9  15–29 4,997 8.5 4,736 8.7 1,207 8.7  ≥30 10,176 17.3 10,858 19.8 3,070 22.1 Characteristic . Leisure Time Spent Sitting . <3 hours/day (n = 58,910) . 3–5 hours/day (n = 54,742) . ≥6 hours/day (n = 13,902) . No. . %a . No. . %a . No. . %a . Age, yearsb 61.5 (6.3) 63.4 (6.2) 64.2 (6.3) BMIb,c 25.5 (4.0) 26.2 (4.2) 26.8 (4.6) MVPA, MET-hours/weekb 13.1 (13.2) 12.4 (12.5) 12.8 (13.1) Male sex 24,300 41.2 24,951 45.6 7,356 52.9 Employment status  Not employed or retired 20,935 35.5 27,880 50.9 8,103 58.3  Employed 35,035 59.5 23,816 43.5 4,970 35.8 Race  White 57,339 97.3 53,328 97.4 13,528 97.3  Black 770 1.3 748 1.4 183 1.3  Other 801 1.4 666 1.2 191 1.4 Education  Less than high school 3,270 5.6 3,245 5.9 937 6.7  High school graduate 31,387 53.3 30,954 56.5 7,280 52.4  College graduate or higher 23,886 40.5 20,167 36.8 5,598 40.3 Smoking status  Never 29,990 50.9 24,004 43.8 5,524 39.7  Current 4,142 7.0 5,190 9.5 1,692 12.2  Former 24,540 41.7 25,368 46.3 6,631 47.7  Ever or unknown status 238 0.4 180 0.3 55 0.4 Alcohol intake, no. of drinks/day  0 23,208 39.4 21,103 38.5 5,502 39.6  <1 23,357 39.6 21,430 39.1 5,003 36.0  1 5,861 9.9 5,662 10.3 1,403 10.1  ≥2 4,281 7.3 4,669 8.5 1,431 10.3 Marital status  Married 52,601 89.3 48,415 88.4 12,115 87.1  Widowed or divorced 4,963 8.4 5,001 9.1 1,363 9.8  Never married 828 1.4 847 1.5 252 1.8 Comorbidity scored  0 28,317 48.1 22,752 41.6 5,475 39.4  1 21,626 36.7 21,264 38.8 5,392 38.8  ≥2 8,967 15.2 10,726 19.6 3,035 21.8 Diet scoree  0–2 11,933 20.3 13,502 24.7 3,757 27.0  3–5 25,836 43.9 24,600 44.9 5,996 43.1  6–9 16,228 27.5 12,441 22.7 2,881 20.7 Aspirin use, no. of pills/month  0 33,896 57.5 30,235 55.2 7,444 53.5  1–14 8,361 14.2 7,527 13.7 1,798 12.9  15–29 4,997 8.5 4,736 8.7 1,207 8.7  ≥30 10,176 17.3 10,858 19.8 3,070 22.1 Abbreviations: BMI, body mass index; MET, metabolic equivalent of task; MVPA, recreational moderate-to-vigorous physical activity. a Percentages may not sum to due to rounding or missing values for exposure. b Values are expressed as mean (standard deviation). c Weight (kg)/height (m)2. d Based on self-reported diabetes, hypertension, and high cholesterol. e Calculated based on intake of fruits and vegetables, whole/refined grains, and red and processed meat, with a score of 9 representing optimal dietary adherence. Open in new tab Results with or without adjustment for MVPA did not differ (Web Table 1, available at https://academic.oup.com/aje). After multivariable adjustment including MVPA, leisure time spent sitting was positively associated with all-cause mortality (RR = 1.19, 95% CI: 1.15, 1.22 for ≥6 vs. <3 hours per day) among men and women combined (Table 3). Longer leisure time spent sitting also was associated with higher risks of cardiovascular disease mortality (RR = 1.19, 95% CI: 1.13, 1.25); when further examining specific types of cardiovascular disease, associations were slightly stronger for coronary heart disease mortality (RR = 1.26, 95% CI: 1.17, 1.35) than for death from stroke or other circulatory disease (Table 3). Prolonged time spent sitting was also associated with higher risk of death from cancer (RR = 1.11, 95% CI: 1.05, 1.17) as well as from diabetes, kidney disease, suicide, COPD, pneumonitis due to solids and liquids, liver, peptic ulcer and other digestive disease, Parkinson disease, Alzheimer disease, nervous disorders, and musculoskeletal disorders (Table 3). Table 3. Associations Between Leisure Time Spent Sitting and Cause-Specific Mortality Among Men and Women in the Cancer Prevention Study II Nutrition Cohort, United States, 1992–2014 . Leisure Time Spent Sitting in 1992 . <3 hours/day (n = 1,093,303 person-years) . 3–5 hours/day (n = 967,808 person-years) . ≥6 hours/day (n = 232,749 person-years) . Cause of Death . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . All causes 18,906 1,893.10 1.00 Referent 22,938 2,279.86 1.07 1.05, 1.09 6,940 2,722.97 1.19 1.15, 1.22 Cancer 5,929 584.67 1.00 Referent 6,714 679.79 1.06 1.02, 1.09 1,907 762.98 1.11 1.05, 1.17 All CVDc 6,037 612.72 1.00 Referent 7,647 760.78 1.06 1.02, 1.10 2,399 935.75 1.19 1.13, 1.25 CHD 2,728 281.90 1.00 Referent 3,535 357.61 1.07 1.02, 1.13 1,187 462.95 1.26 1.17, 1.35 Stroke 1,286 127.53 1.00 Referent 1,510 146.06 1.04 0.96, 1.12 443 173.01 1.15 1.03, 1.28 Other circulatory disease 2,023 203.29 1.00 Referent 2,602 257.12 1.06 1.00, 1.13 769 299.79 1.13 1.03, 1.23 Diabetes 338 34.38 1.00 Referent 480 47.96 1.13 0.98, 1.30 169 66.30 1.31 1.09, 1.59 Other nutrition/metabolic disease 244 24.42 1.00 Referent 335 32.58 1.19 1.01, 1.41 84 34.31 1.12 0.87, 1.44 Kidney disease 269 28.27 1.00 Referent 392 38.49 1.15 0.99, 1.35 142 54.70 1.43 1.16, 1.77 Other genitourinary disease 124 12.39 1.00 Referent 144 14.11 0.99 0.78, 1.27 45 17.32 1.09 0.77, 1.55 Accidents 564 56.54 1.00 Referent 629 62.90 1.03 0.92, 1.16 146 56.37 0.91 0.76, 1.10 Suicide 70 7.46 1.00 Referent 67 7.37 0.93 0.66, 1.31 34 13.65 1.66 1.09, 2.54 COPD 544 54.68 1.00 Referent 814 80.08 1.16 1.04, 1.30 284 110.12 1.38 1.19, 1.60 Pneumonia and influenza 386 39.24 1.00 Referent 496 48.91 1.09 0.95, 1.24 152 58.28 1.20 0.99, 1.46 Pneumonitis due to solids and liquids 121 12.60 1.00 Referent 190 18.94 1.30 1.03, 1.64 56 21.49 1.41 1.02, 1.95 Other respiratory disease 340 34.53 1.00 Referent 394 39.18 1.02 0.88, 1.18 118 45.41 1.13 0.91, 1.40 Liver disease 95 9.32 1.00 Referent 122 12.56 1.16 0.88, 1.53 54 23.18 1.80 1.27, 2.54 Peptic ulcer and other digestive disease 425 42.10 1.00 Referent 546 53.16 1.10 0.97, 1.25 177 68.64 1.31 1.09, 1.57 Infectious diseases 374 37.55 1.00 Referent 408 40.14 0.92 0.79, 1.06 131 51.86 1.05 0.85, 1.28 Dementia/mental disorders 964 94.73 1.00 Referent 1,142 108.12 1.05 0.97, 1.15 300 114.86 1.11 0.97, 1.26 Parkinson disease 456 48.10 1.00 Referent 528 52.93 1.12 0.99, 1.27 169 63.90 1.37 1.15, 1.64 Alzheimer disease 923 90.17 1.00 Referent 1,032 97.91 1.02 0.94, 1.12 293 112.01 1.18 1.03, 1.35 Nervous disorders 331 32.78 1.00 Referent 376 37.45 1.16 1.00, 1.35 127 51.13 1.54 1.25, 1.90 Musculoskeletal disorders 140 13.36 1.00 Referent 150 14.28 1.03 0.82, 1.31 56 22.08 1.58 1.15, 2.18 Ill-defined 161 16.20 1.00 Referent 244 23.39 1.36 1.11, 1.66 70 27.12 1.49 1.12, 1.99 Other 71 6.89 1.00 Referent 88 8.84 1.17 0.85, 1.61 27 11.49 1.32 0.84, 2.09 . Leisure Time Spent Sitting in 1992 . <3 hours/day (n = 1,093,303 person-years) . 3–5 hours/day (n = 967,808 person-years) . ≥6 hours/day (n = 232,749 person-years) . Cause of Death . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . All causes 18,906 1,893.10 1.00 Referent 22,938 2,279.86 1.07 1.05, 1.09 6,940 2,722.97 1.19 1.15, 1.22 Cancer 5,929 584.67 1.00 Referent 6,714 679.79 1.06 1.02, 1.09 1,907 762.98 1.11 1.05, 1.17 All CVDc 6,037 612.72 1.00 Referent 7,647 760.78 1.06 1.02, 1.10 2,399 935.75 1.19 1.13, 1.25 CHD 2,728 281.90 1.00 Referent 3,535 357.61 1.07 1.02, 1.13 1,187 462.95 1.26 1.17, 1.35 Stroke 1,286 127.53 1.00 Referent 1,510 146.06 1.04 0.96, 1.12 443 173.01 1.15 1.03, 1.28 Other circulatory disease 2,023 203.29 1.00 Referent 2,602 257.12 1.06 1.00, 1.13 769 299.79 1.13 1.03, 1.23 Diabetes 338 34.38 1.00 Referent 480 47.96 1.13 0.98, 1.30 169 66.30 1.31 1.09, 1.59 Other nutrition/metabolic disease 244 24.42 1.00 Referent 335 32.58 1.19 1.01, 1.41 84 34.31 1.12 0.87, 1.44 Kidney disease 269 28.27 1.00 Referent 392 38.49 1.15 0.99, 1.35 142 54.70 1.43 1.16, 1.77 Other genitourinary disease 124 12.39 1.00 Referent 144 14.11 0.99 0.78, 1.27 45 17.32 1.09 0.77, 1.55 Accidents 564 56.54 1.00 Referent 629 62.90 1.03 0.92, 1.16 146 56.37 0.91 0.76, 1.10 Suicide 70 7.46 1.00 Referent 67 7.37 0.93 0.66, 1.31 34 13.65 1.66 1.09, 2.54 COPD 544 54.68 1.00 Referent 814 80.08 1.16 1.04, 1.30 284 110.12 1.38 1.19, 1.60 Pneumonia and influenza 386 39.24 1.00 Referent 496 48.91 1.09 0.95, 1.24 152 58.28 1.20 0.99, 1.46 Pneumonitis due to solids and liquids 121 12.60 1.00 Referent 190 18.94 1.30 1.03, 1.64 56 21.49 1.41 1.02, 1.95 Other respiratory disease 340 34.53 1.00 Referent 394 39.18 1.02 0.88, 1.18 118 45.41 1.13 0.91, 1.40 Liver disease 95 9.32 1.00 Referent 122 12.56 1.16 0.88, 1.53 54 23.18 1.80 1.27, 2.54 Peptic ulcer and other digestive disease 425 42.10 1.00 Referent 546 53.16 1.10 0.97, 1.25 177 68.64 1.31 1.09, 1.57 Infectious diseases 374 37.55 1.00 Referent 408 40.14 0.92 0.79, 1.06 131 51.86 1.05 0.85, 1.28 Dementia/mental disorders 964 94.73 1.00 Referent 1,142 108.12 1.05 0.97, 1.15 300 114.86 1.11 0.97, 1.26 Parkinson disease 456 48.10 1.00 Referent 528 52.93 1.12 0.99, 1.27 169 63.90 1.37 1.15, 1.64 Alzheimer disease 923 90.17 1.00 Referent 1,032 97.91 1.02 0.94, 1.12 293 112.01 1.18 1.03, 1.35 Nervous disorders 331 32.78 1.00 Referent 376 37.45 1.16 1.00, 1.35 127 51.13 1.54 1.25, 1.90 Musculoskeletal disorders 140 13.36 1.00 Referent 150 14.28 1.03 0.82, 1.31 56 22.08 1.58 1.15, 2.18 Ill-defined 161 16.20 1.00 Referent 244 23.39 1.36 1.11, 1.66 70 27.12 1.49 1.12, 1.99 Other 71 6.89 1.00 Referent 88 8.84 1.17 0.85, 1.61 27 11.49 1.32 0.84, 2.09 Abbreviations: CHD, coronary heart disease; CI, confidence interval; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; RR, relative risk. a Age and sex-standardized rate per 100,000 person-years. b Multivariable models adjusted for sex, race, education, employment status, alcohol intake, marital status, smoking status (including smoking duration and frequency among current smokers and years since quit among former smokers), number of comorbidities, American Cancer Society diet score, aspirin use, body mass index, and recreational moderate-to-vigorous physical activity. c All CVD includes coronary heart disease, stroke, and other circulatory diseases. Open in new tab Table 3. Associations Between Leisure Time Spent Sitting and Cause-Specific Mortality Among Men and Women in the Cancer Prevention Study II Nutrition Cohort, United States, 1992–2014 . Leisure Time Spent Sitting in 1992 . <3 hours/day (n = 1,093,303 person-years) . 3–5 hours/day (n = 967,808 person-years) . ≥6 hours/day (n = 232,749 person-years) . Cause of Death . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . All causes 18,906 1,893.10 1.00 Referent 22,938 2,279.86 1.07 1.05, 1.09 6,940 2,722.97 1.19 1.15, 1.22 Cancer 5,929 584.67 1.00 Referent 6,714 679.79 1.06 1.02, 1.09 1,907 762.98 1.11 1.05, 1.17 All CVDc 6,037 612.72 1.00 Referent 7,647 760.78 1.06 1.02, 1.10 2,399 935.75 1.19 1.13, 1.25 CHD 2,728 281.90 1.00 Referent 3,535 357.61 1.07 1.02, 1.13 1,187 462.95 1.26 1.17, 1.35 Stroke 1,286 127.53 1.00 Referent 1,510 146.06 1.04 0.96, 1.12 443 173.01 1.15 1.03, 1.28 Other circulatory disease 2,023 203.29 1.00 Referent 2,602 257.12 1.06 1.00, 1.13 769 299.79 1.13 1.03, 1.23 Diabetes 338 34.38 1.00 Referent 480 47.96 1.13 0.98, 1.30 169 66.30 1.31 1.09, 1.59 Other nutrition/metabolic disease 244 24.42 1.00 Referent 335 32.58 1.19 1.01, 1.41 84 34.31 1.12 0.87, 1.44 Kidney disease 269 28.27 1.00 Referent 392 38.49 1.15 0.99, 1.35 142 54.70 1.43 1.16, 1.77 Other genitourinary disease 124 12.39 1.00 Referent 144 14.11 0.99 0.78, 1.27 45 17.32 1.09 0.77, 1.55 Accidents 564 56.54 1.00 Referent 629 62.90 1.03 0.92, 1.16 146 56.37 0.91 0.76, 1.10 Suicide 70 7.46 1.00 Referent 67 7.37 0.93 0.66, 1.31 34 13.65 1.66 1.09, 2.54 COPD 544 54.68 1.00 Referent 814 80.08 1.16 1.04, 1.30 284 110.12 1.38 1.19, 1.60 Pneumonia and influenza 386 39.24 1.00 Referent 496 48.91 1.09 0.95, 1.24 152 58.28 1.20 0.99, 1.46 Pneumonitis due to solids and liquids 121 12.60 1.00 Referent 190 18.94 1.30 1.03, 1.64 56 21.49 1.41 1.02, 1.95 Other respiratory disease 340 34.53 1.00 Referent 394 39.18 1.02 0.88, 1.18 118 45.41 1.13 0.91, 1.40 Liver disease 95 9.32 1.00 Referent 122 12.56 1.16 0.88, 1.53 54 23.18 1.80 1.27, 2.54 Peptic ulcer and other digestive disease 425 42.10 1.00 Referent 546 53.16 1.10 0.97, 1.25 177 68.64 1.31 1.09, 1.57 Infectious diseases 374 37.55 1.00 Referent 408 40.14 0.92 0.79, 1.06 131 51.86 1.05 0.85, 1.28 Dementia/mental disorders 964 94.73 1.00 Referent 1,142 108.12 1.05 0.97, 1.15 300 114.86 1.11 0.97, 1.26 Parkinson disease 456 48.10 1.00 Referent 528 52.93 1.12 0.99, 1.27 169 63.90 1.37 1.15, 1.64 Alzheimer disease 923 90.17 1.00 Referent 1,032 97.91 1.02 0.94, 1.12 293 112.01 1.18 1.03, 1.35 Nervous disorders 331 32.78 1.00 Referent 376 37.45 1.16 1.00, 1.35 127 51.13 1.54 1.25, 1.90 Musculoskeletal disorders 140 13.36 1.00 Referent 150 14.28 1.03 0.82, 1.31 56 22.08 1.58 1.15, 2.18 Ill-defined 161 16.20 1.00 Referent 244 23.39 1.36 1.11, 1.66 70 27.12 1.49 1.12, 1.99 Other 71 6.89 1.00 Referent 88 8.84 1.17 0.85, 1.61 27 11.49 1.32 0.84, 2.09 . Leisure Time Spent Sitting in 1992 . <3 hours/day (n = 1,093,303 person-years) . 3–5 hours/day (n = 967,808 person-years) . ≥6 hours/day (n = 232,749 person-years) . Cause of Death . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . No. of Deaths . Ratea . RRb . 95% CI . All causes 18,906 1,893.10 1.00 Referent 22,938 2,279.86 1.07 1.05, 1.09 6,940 2,722.97 1.19 1.15, 1.22 Cancer 5,929 584.67 1.00 Referent 6,714 679.79 1.06 1.02, 1.09 1,907 762.98 1.11 1.05, 1.17 All CVDc 6,037 612.72 1.00 Referent 7,647 760.78 1.06 1.02, 1.10 2,399 935.75 1.19 1.13, 1.25 CHD 2,728 281.90 1.00 Referent 3,535 357.61 1.07 1.02, 1.13 1,187 462.95 1.26 1.17, 1.35 Stroke 1,286 127.53 1.00 Referent 1,510 146.06 1.04 0.96, 1.12 443 173.01 1.15 1.03, 1.28 Other circulatory disease 2,023 203.29 1.00 Referent 2,602 257.12 1.06 1.00, 1.13 769 299.79 1.13 1.03, 1.23 Diabetes 338 34.38 1.00 Referent 480 47.96 1.13 0.98, 1.30 169 66.30 1.31 1.09, 1.59 Other nutrition/metabolic disease 244 24.42 1.00 Referent 335 32.58 1.19 1.01, 1.41 84 34.31 1.12 0.87, 1.44 Kidney disease 269 28.27 1.00 Referent 392 38.49 1.15 0.99, 1.35 142 54.70 1.43 1.16, 1.77 Other genitourinary disease 124 12.39 1.00 Referent 144 14.11 0.99 0.78, 1.27 45 17.32 1.09 0.77, 1.55 Accidents 564 56.54 1.00 Referent 629 62.90 1.03 0.92, 1.16 146 56.37 0.91 0.76, 1.10 Suicide 70 7.46 1.00 Referent 67 7.37 0.93 0.66, 1.31 34 13.65 1.66 1.09, 2.54 COPD 544 54.68 1.00 Referent 814 80.08 1.16 1.04, 1.30 284 110.12 1.38 1.19, 1.60 Pneumonia and influenza 386 39.24 1.00 Referent 496 48.91 1.09 0.95, 1.24 152 58.28 1.20 0.99, 1.46 Pneumonitis due to solids and liquids 121 12.60 1.00 Referent 190 18.94 1.30 1.03, 1.64 56 21.49 1.41 1.02, 1.95 Other respiratory disease 340 34.53 1.00 Referent 394 39.18 1.02 0.88, 1.18 118 45.41 1.13 0.91, 1.40 Liver disease 95 9.32 1.00 Referent 122 12.56 1.16 0.88, 1.53 54 23.18 1.80 1.27, 2.54 Peptic ulcer and other digestive disease 425 42.10 1.00 Referent 546 53.16 1.10 0.97, 1.25 177 68.64 1.31 1.09, 1.57 Infectious diseases 374 37.55 1.00 Referent 408 40.14 0.92 0.79, 1.06 131 51.86 1.05 0.85, 1.28 Dementia/mental disorders 964 94.73 1.00 Referent 1,142 108.12 1.05 0.97, 1.15 300 114.86 1.11 0.97, 1.26 Parkinson disease 456 48.10 1.00 Referent 528 52.93 1.12 0.99, 1.27 169 63.90 1.37 1.15, 1.64 Alzheimer disease 923 90.17 1.00 Referent 1,032 97.91 1.02 0.94, 1.12 293 112.01 1.18 1.03, 1.35 Nervous disorders 331 32.78 1.00 Referent 376 37.45 1.16 1.00, 1.35 127 51.13 1.54 1.25, 1.90 Musculoskeletal disorders 140 13.36 1.00 Referent 150 14.28 1.03 0.82, 1.31 56 22.08 1.58 1.15, 2.18 Ill-defined 161 16.20 1.00 Referent 244 23.39 1.36 1.11, 1.66 70 27.12 1.49 1.12, 1.99 Other 71 6.89 1.00 Referent 88 8.84 1.17 0.85, 1.61 27 11.49 1.32 0.84, 2.09 Abbreviations: CHD, coronary heart disease; CI, confidence interval; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; RR, relative risk. a Age and sex-standardized rate per 100,000 person-years. b Multivariable models adjusted for sex, race, education, employment status, alcohol intake, marital status, smoking status (including smoking duration and frequency among current smokers and years since quit among former smokers), number of comorbidities, American Cancer Society diet score, aspirin use, body mass index, and recreational moderate-to-vigorous physical activity. c All CVD includes coronary heart disease, stroke, and other circulatory diseases. Open in new tab Cancer and kidney disease mortality were the only outcomes where associations differed according to sex (interaction P = 0.04 and 0.01, respectively), with stronger associations among women for both cancer (for ≥6 vs. <3 hours per day sitting time, RR = 1.20, 95% CI: 1.11, 1.30) and kidney disease (RR = 1.97, 95% CI: 1.40, 2.77) compared with men (for cancer, RR = 1.06, 95% CI: 0.99, 1.13; for kidney disease, RR = 1.18, 95% CI: 0.90, 1.54). MVPA modified the association between leisure time spent sitting and death from all cardiovascular disease (for <8.75 MET-hours/week, RR = 1.24, 95% CI: 1.16, 1.33; for ≥8.75 MET-hours/week; RR = 1.13, 95% CI: 1.05, 1.21; P for interaction = 0.003) and stroke with associations stronger in those who were less (for <8.75 MET-hours/week, RR = 1.27, 95% CI: 1.10, 1.48) compared with more active (for ≥8.75 MET-hours/week, RR = 1.01, 95% CI: 0.86, 1.20; P for interaction = 0.01). No other associations differed when stratified by sex, physical activity level, or BMI. When stratifying by occupational status, there was significant effect modification for all-cause mortality (P for interaction = 0.02) but not for any specific causes of death (data not shown). Participants who were employed at baseline had a higher risk of all-cause mortality (for ≥6 vs. <3 hours per day of leisure time spent sitting, RR = 1.23, 95% CI: 1.16, 1.29) compared with those who were unemployed or retired (for ≥6 vs. <3 hours per day of leisure time spent sitting, RR = 1.17, 95% CI: 1.13, 1.21). There was a statistical violation (global P < 0.0001) of the Cox proportional hazards assumption observed for all-cause mortality and some of the more common specific causes of death (cancer, cardiovascular disease, COPD, accidents, Parkinson disease, and Alzheimer disease). However, visual review of log-log survival curves did not show strong evidence of nonparallelism. When results were then stratified by follow-up time, associations were similar but attenuated in the second period. To further examine whether reverse causation explained any of the observed associations in the earlier years of follow-up, we also conducted an analysis with a 5-year lag, and associations were similar with very slight attenuation (e.g., overall RR = 1.38, 95% CI: 1.19, 1.60; with a 5-year lag for COPD mortality, RR = 1.37, 95% CI: 1.18, 1.59; additional data shown in Web Table 2). Results were similar to those in the overall population in the sensitivity analysis restricted to men and women who were lifelong nonsmokers or long-term (≥15 years) former smokers (data not shown). For diabetes mortality, excluding individuals with diabetes at baseline resulted in an attenuation of the association and loss of statistical significance (for ≥6 hours per day spent sitting vs. <3 hours per day among individuals with no history of diabetes at baseline, RR = 1.23, 95% CI: 0.91, 1.68). Associations among those with no personal history of diabetes or kidney stones were virtually identical to those for the overall population in relation to deaths due to kidney disease (data not shown). DISCUSSION In this large prospective study of adults who were free of major chronic diseases at baseline, prolonged leisure time spent sitting for more than 6 hours per day was associated with a 19% higher all-cause death rate when compared with sitting less than 3 hours per day. Risk was significantly higher for 14 of the 22 specific causes of death examined and, importantly, for 8 of the top 10 leading causes of death in the United States (23), including cancer, coronary heart disease, stroke, diabetes, kidney disease, suicide, COPD, pneumonitis due to solids and liquids, liver disease, peptic ulcer and other digestive disease, Parkinson disease, Alzheimer disease, nervous disorders, and musculoskeletal disorders. Our findings are consistent with previous studies that examined associations between prolonged sitting time and all-cause, cardiovascular disease, and cancer mortality (3–6), including an earlier report from the CPS-II Nutrition Cohort where leisure time spent sitting (≥6 vs. <3 hours per day) was associated with all-cause and cardiovascular disease mortality among women and men, and with cancer mortality among women but not men (5). In that study, sitting time was associated with mortality from other causes combined among women (RR = 1.41, 95% CI: 1.25, 1.60) and men (RR = 1.33, 95% CI: 1.20, 1.47). To better understand individual, less-studied cardiovascular disease mortality outcomes, we expanded the present study to examine specific types of cardiovascular disease and found that sitting time was more strongly associated with coronary heart disease mortality than with stroke or death from other circulatory diseases. These findings were similar to those from the NIH-AARP Diet and Health Study, a prospective cohort of approximately 220,000 healthy adults, in which investigators found positive associations between prolonged sitting while watching television and subsequent mortality risk from cancer and coronary heart disease but not for stroke (12). In another prospective cohort of 12,608 men and women, investigators also found that prolonged sitting was associated with incident coronary heart disease and a (statistically not significant) higher risk of incident stroke (24). Thus, these results require additional replication in other large prospective cohorts. The NIH-AARP study also found prolonged television time to be associated with higher risk of death from COPD, diabetes, influenza/pneumonia, Parkinson disease, liver disease, and suicide (12). Our findings are largely consistent with a higher risk of mortality from all of these causes of death with the exception of influenza/pneumonia. In addition, we found prolonged sitting to be associated with death from kidney disease, pneumonitis due to solids and liquids, peptic ulcer and other digestive diseases, Alzheimer disease, nervous disorders, and musculoskeletal disorders, with risk estimates ranging from 18% to 58% higher risk for ≥6 versus <3 hours per day spent sitting during leisure time. However, risk estimates in the NIH-AARP study for kidney disease and Alzheimer disease were suggestive of a positive association (RR = 1.22 and 1.46, respectively, for ≥7 hours vs. <1 hour/day of television viewing) but did not reach statistical significance (12). To our knowledge, the other causes of death associated with sitting time included in the present study have not been previously studied in relation to time spent sitting, and those findings require replication in other large prospective studies. Several possible factors could explain the positive associations between time spent sitting and higher mortality risk from various chronic diseases. Time spent sitting displaces time spent in physical activity and is associated with lower total physical activity levels (11). Adjustment for MVPA did not change results, but it is unclear whether other forms of activity might have (e.g., daily-life light activities). Alternatively, it is possible that time spent sitting, especially when engaged in specific activities such as television viewing, is associated with other unhealthy behaviors, such as excess snacking (25). This could explain why associations with time spent watching television are generally stronger in magnitude than those for leisure time spent sitting (3, 12). However, the relative contributions of poor diet and inactivity linked to prolonged television viewing and poor health are not well understood. Prolonged time spent sitting has also been shown to have important metabolic consequences, including a detrimental influence on various cardiometabolic factors (triglycerides, fasting plasma glucose, blood pressure, and insulin) and the promotion of obesity-related systemic inflammation (24, 26, 27). These pathways may explain why associations were observed with death from cardiovascular disease, cancer, diabetes, liver disease, kidney disease, and COPD, all of which have been shown to be associated with poor metabolic function and obesity. Regardless of the underlying mechanism, results from observational studies consistently suggest that reducing leisure time spent sitting could be beneficial for many health outcomes. Nevertheless, additional studies are essential to better understand how reductions in sitting time affect mortality and other outcomes. Ideally, randomized controlled trials should be conducted, but they would be costly and lengthy and thus may not be feasible. Other associations—including death from suicide, Parkinson disease, Alzheimer disease, nervous disorders, and musculoskeletal disorders—are less clearly understood. For these causes of death, it is plausible that underlying conditions that would result in excess time spent sitting may explain associations. For example, individuals who are clinically depressed may be more likely to spend time sitting and more likely to commit suicide; however, we are unable to adjust for depression in the present study. Similarly, we were unable to exclude individuals with prevalent Parkinson disease or Alzheimer disease at baseline because we did not capture this information; thus, it is possible that individuals with these conditions are sitting longer as a result of the disease, rather than sitting being a cause of death from these conditions. More research is needed to better understand the relationship between prolonged sitting and death from these outcomes. The strengths of our study include the large sample size, prospective design, and ability to control for many potential confounders. The lack of information on some prevalent diseases is a limitation of the study. While we adjusted for a wide range of potential confounders, it is possible that there is some residual confounding by another unmeasured factor. Another limitation is the lack of information on occupational physical activity and sedentary time; however, when we stratified our analyses by employment status, results for all-cause mortality were slightly stronger among employed participants compared with those who were unemployed/retired at baseline. This may be due to a higher amount of total sitting time. For all individual causes of death, results were virtually unchanged. Another limitation is the use of self-reported data, including sitting time and physical activity information, which may lead to some nondifferential misclassification. Although these questions are subject to misreporting, the sitting time and physical activity measures are very similar to those validated in the Nurses’ Health Study II, a prospective cohort with similar participant characteristics, which found a correlation of 0.79 between reported behavior on questionnaires and recalls (28). Finally, another limitation to note is that we were unable to differentiate among types of leisure-time sitting activities (i.e., sitting while watching television, reading, or driving). In conclusion, prolonged leisure time spent sitting was associated with all-cause mortality and with 14 of 22 specific causes of death—including 8 of the 10 most common—independent of moderate-to-vigorous intensity physical activity. Given the increase in leisure time spent sitting in westernized countries over the past several decades, this report broadens our understanding of the array of negative health effects associated with prolonged sitting time and again highlights the importance of sedentary behavior, in addition to exercise, as a health behavior that may be modified in our ongoing efforts to improve public health. ACKNOWLEDGMENTS Author affiliations: Behavioral and Epidemiology Research Group, American Cancer Society, Atlanta, Georgia (Alpa V. Patel, Maret L. Maliniak, Erika Rees-Punia, Susan M. Gapstur); and Metabolic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Maryland (Charles E. Matthews). The American Cancer Society funded all aspects of the data collection and analysis. We thank all the Cancer Prevention Study II participants and each member of the study and biospecimen management group. Conflict of interest: none declared. Abbreviations BMI body mass index CI confidence interval COPD chronic obstructive pulmonary disease CPS-II Cancer Prevention Study II MET metabolic equivalent of task MVPA moderate-to-vigorous physical activity RR relative risk REFERENCES 1 Chau JY , Merom D, Grunseit A, et al. . Temporal trends in non-occupational sedentary behaviours from Australian Time Use Surveys 1992, 1997 and 2006 . Int J Behav Nutr Phys Act . 2012 ; 9 : 76 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Menai M , Fezeu L, Charreire H, et al. . Changes in sedentary behaviours and associations with physical activity through retirement: a 6-year longitudinal study . PLoS One . 2014 ; 9 ( 9 ): e106850 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Ekelund E , Steene-Johannessen J, Brown WJ, et al. . Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? A harmonised meta-analysis of data from more than 1 million men and women . Lancet . 2016 ; 388 ( 10051 ): 1302 – 1310 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Chau JY , Grunseit AC, Chey T, et al. . Daily sitting time and all-cause mortality: a meta-analysis . PLoS One . 2013 ; 8 ( 11 ): e80000 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Patel AV , Bernstein L, Deka A, et al. . Leisure time spent sitting in relation to total mortality in a prospective cohort of US adults . Am J Epidemiol . 2010 ; 172 ( 4 ): 419 – 429 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Matthews CE , George SM, Moore SC, et al. . Amount of time spent in sedentary behaviors and cause-specific mortality in US adults . Am J Clin Nutr . 2012 ; 95 ( 2 ): 437 – 445 . Google Scholar Crossref Search ADS PubMed WorldCat 7 Wilmot EG , Edwardson CL, Achana FA, et al. . Sedentary time in adults and the association with diabetes, cardiovascular disease and death: systematic review and meta-analysis . Diabetologia . 2012 ; 55 ( 11 ): 2895 – 2905 . Google Scholar Crossref Search ADS PubMed WorldCat 8 Ford ES , Caspersen CJ. Sedentary behaviour and cardiovascular disease: a review of prospective studies . Int J Epidemiol . 2012 ; 41 ( 5 ): 1338 – 1353 . Google Scholar Crossref Search ADS PubMed WorldCat 9 Patel AV , Hildebrand JS, Campbell PT, et al. . Leisure-time spent sitting and site-specific cancer incidence in a large US cohort . Cancer Epidemiol Biomarkers Prev . 2015 ; 24 ( 9 ): 1350 – 1359 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Schmid D , Leitzmann MF. Television viewing and time spent sedentary in relation to cancer risk: a meta-analysis . J Natl Cancer Inst . 2014 ; 106 ( 7 ). Google Scholar OpenURL Placeholder Text WorldCat 11 Keadle SK , Conroy DE, Buman MP, et al. . Targeting reductions in sitting time to increase physical activity and improve health . Med Sci Sports Exerc . 2017 ; 49 ( 8 ): 1572 – 1582 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Keadle SK , Moore SC, Sampson JN, et al. . Causes of death associated with prolonged TV viewing: NIH-AARP Diet and Health Study . Am J Prev Med . 2015 ; 49 ( 6 ): 811 – 821 . Google Scholar Crossref Search ADS PubMed WorldCat 13 Calle EE , Rodriguez C, Jacobs EJ, et al. . The American Cancer Society Cancer Prevention Study II Nutrition Cohort: rationale, study design, and baseline characteristics . Cancer . 2002 ; 94 ( 2 ): 500 – 511 . Google Scholar Crossref Search ADS PubMed WorldCat 14 Garfinkel L . Selection, follow-up, and analysis in the American Cancer Society prospective studies . Natl Cancer Inst Monogr . 1985 ; 67 : 49 – 52 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 15 Ainsworth BE , Haskell WL, Whitt MC, et al. . Compendium of physical activities: an update of activity codes and MET intensities . Med Sci Sports Exerc . 2000 ; 32 ( 9 suppl ): S498 – S504 . Google Scholar Crossref Search ADS PubMed WorldCat 16 Calle EE , Terrell D. Utility of the National Death Index for ascertainment of mortality among cancer prevention study II participants . Am J Epidemiol . 1993 ; 137 ( 2 ): 235 – 241 . Google Scholar Crossref Search ADS PubMed WorldCat 17 World Health Organization . International Classification of Diseases: Manual of the International Statistical Classification of Disease, Injuries, and Causes of Death. Geneva, Switzerland : World Health Organization ; 1977 . (Based on the recommendations of the Ninth Revision Conference, 1975). 18 World Health Organization . International Statistical Classification of Diseases and Related Health Problems Tenth Revision. Geneva, Switzerland : World Health Organization ; 2016 . http://apps.who.int/classifications/icd10/browse/2016/en. 19 Cox D . Regression models and life tables . J R Stat Soc . 1972 ; 34 ( 2 ): 187 – 220 . Google Scholar OpenURL Placeholder Text WorldCat 20 McCullough ML , Patel AV, Kushi LH, et al. . Following cancer prevention guidelines reduces risk of cancer, cardiovascular disease, and all-cause mortality . Cancer Epidemiol Biomarkers Prev . 2011 ; 20 ( 6 ): 1089 – 1097 . Google Scholar Crossref Search ADS PubMed WorldCat 21 Grambsch PM , Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals . Biometrika . 1994 ; 81 ( 3 ): 515 – 526 . Google Scholar Crossref Search ADS WorldCat 22 R Core Team . R: A language and environment for statistical computing. Vienna, Austria : The R Foundation ; 2017 . 23 Heron M . Deaths: Leading Causes for 2014 . Natl Vital Stat Rep . 2016 ; 65 ( 5 ): 1 – 96 . Google Scholar OpenURL Placeholder Text WorldCat 24 Wijndaele K , Brage S, Besson H, et al. . Television viewing and incident cardiovascular disease: prospective associations and mediation analysis in the EPIC Norfolk Study . PLoS One . 2011 ; 6 ( 5 ): e20058 . Google Scholar Crossref Search ADS PubMed WorldCat 25 Thorp AA , McNaughton SA, Owen N, et al. . Independent and joint associations of TV viewing time and snack food consumption with the metabolic syndrome and its components; a cross-sectional study in Australian adults . Int J Behav Nutr Phys Act . 2013 ; 10 : 96 . Google Scholar Crossref Search ADS PubMed WorldCat 26 Healy GN , Winkler EA, Owen N, et al. . Replacing sitting time with standing or stepping: associations with cardio-metabolic risk biomarkers . Eur Heart J . 2015 ; 36 ( 39 ): 2643 – 2649 . Google Scholar Crossref Search ADS PubMed WorldCat 27 Lynch BM . Sedentary behavior and cancer: a systematic review of the literature and proposed biological mechanisms . Cancer Epidemiol Biomarkers Prev . 2010 ; 19 ( 11 ): 2691 – 2709 . Google Scholar Crossref Search ADS PubMed WorldCat 28 Wolf AM , Hunter DJ, Colditz GA, et al. . Reproducibility and validity of a self-administered physical activity questionnaire . Int J Epidemiol . 1994 ; 23 ( 5 ): 991 – 999 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2018.

Bin, Wang,;Jihong, Liu,;Yongjun, Zhang,;Chonghuai, Yan,;Hui, Wang,;Fan, Jiang,;Fei, Li,;Jun, Zhang,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy122pmid: 29893794

Abstract Mounting evidence has linked postnatal antibiotic use with body mass index (BMI) in children, but the influence of prenatal antibiotic use on offspring obesity risk remains unclear. We aimed to assess the association between fetal exposure to antibiotics and obesity at ages 4 and 7 years among 43,332 children using a multicenter prospective cohort of the US Collaborative Perinatal Project (1959–1976). Antibiotic use was ascertained for mothers during pregnancy. Modified Poisson regression models were used to estimate risk ratios for obesity (BMI >95th percentile), and linear mixed models were applied to assess the association with BMI z score. Repeated prenatal exposure to antibiotics was associated with childhood obesity at age 7 years, and risk of obesity tended to increase with an increasing number of antibiotic exposures (for 2–3 exposures, risk ratio (RR) = 1.22, 95% confidence interval (CI): 1.04, 1.44; for ≥4 exposures, RR = 1.34, 95% CI: 1.03, 1.73). The magnitude of association was strongest for repeated exposures in the second trimester (RR = 1.40, 95% CI: 1.16, 1.71). Prenatal antibiotic use was not associated with obesity or BMI z score at age 4 years. These findings support an increased risk of mid-childhood obesity with repeated use of antibiotics during pregnancy. antibiotic, childhood obesity, exposure, prenatal, prospective studies Childhood obesity is a challenging public health crisis worldwide (1). Between 1980 and 2013, the global prevalence of childhood overweight and obesity increased by about 50% (2). Obesity in childhood is associated with adverse cardiometabolic outcomes such as type 2 diabetes, fatty liver disease, dyslipidemia, and hypertension in adulthood (3). In addition to well-known risk factors including genetic predisposition, excess energy intake, and less physical activity (1), emerging evidence has identified that disturbances in microbial diversity and composition of the gut could affect host energy metabolism and lead to obesity (4–6). Antibiotics have been used to promote weight gain in livestock for more than half a century (7). This effect was hypothesized to be mediated by the gut microbiome. Findings from animal models suggest that alteration of the microbiota due to antibiotic exposure in early life, even if it is a short-term use, may be sufficient to induce lasting metabolic consequences (8–10). Randomized controlled trials have also shown that antibiotics also promote weight gain in humans (11–14). Epidemiologic studies of early-life exposure to antibiotics and the development of body mass in children mostly focused on postnatal antibiotic use. The majority of the studies demonstrated that antibiotic exposure in infancy was associated with an increased risk of childhood obesity (15–17). Two prior studies observed a positive association between prenatal exposure to antibiotics and obesity in children at school ages (18, 19). However, 2 recent studies examining antibiotic use during pregnancy and body mass index (BMI) in early childhood reported conflicting results (20, 21). Although such evidence suggests the potential role of prenatal antibiotic use in offspring weight gain, substantial uncertainty still remains regarding the influence of different antibiotic types and timing of the first exposure. Moreover, the sample sizes of several previous studies were relatively small. Antibiotics are among the most commonly dispensed drug classes in pregnancy (22). Therefore, a better understanding of the association between prenatal antibiotic exposure and childhood obesity may have important public health implications. The present study aimed to evaluate the impact of prenatal antibiotic use on the development of obesity among children at 4 and 7 years of age, using data from a historical prospective birth cohort, the US Collaborative Perinatal Project (CPP). METHODS Study design and participants The CPP was a prospective cohort study that recruited approximately 46,000 women who gave birth to nearly 59,000 babies at 12 medical centers across the United States during 1959–1965. During the study period, 18% of the women gave birth to more than 1 child. Women were routinely interviewed during pregnancy, and their children were followed through age 7 or 8 years with systematic assessments of neurological and cognitive development, physical growth, and other health parameters. The CPP has been described in detail previously (23). Anthropometric data captured at approximately ages 4 years (46–52 months) and 7 years (82–90 months) were employed to assess study outcomes; there was strong clustering of measurements at these times, with fewer children measured outside these age ranges (24). The primary outcome of interest for this study was childhood obesity; therefore, measurements at younger ages (e.g., in infancy) were not applicable. The present analysis focused on live singleton births with complete records of prenatal antibiotic use and child weight and height measurements at ages 4 or 7 years. Women could be enrolled again in subsequent pregnancies. The institutional review board of our hospital granted an exemption from ethics review because the CPP data are deidentified for public use. Antibiotic exposure At enrollment and each prenatal visit, use of drugs including antibiotics was ascertained by maternal self-report. Women reported the type of drugs used and the number of days the drugs were taken in each month of pregnancy. Antibiotic exposure was defined as a systemic antibacterial agent use at any time during pregnancy. Timing of first exposure to antibiotics was grouped into the first trimester (month 1–3), second trimester (month 4–7), or third trimester (month 8 to delivery) of gestation. The number of antibiotics taken was used as a surrogate for antibiotic courses because actual courses could not be discerned in the CPP data. Different types of antibiotics were identified as follows: penicillins, sulfonamides, tetracyclines, nitrofurans, aminoglycosides, chloramphenicol, macrolides, and others. Of these, penicillins were classified as narrow spectrum, and all other antibacterial agents were classified as broad spectrum. Outcome measures Children had their height and weight measured at ages 4 and 7 years by trained personnel. Height was measured to the nearest 0.5 cm using a standardized backboard, and weight was recorded in pounds to the nearest 0.25 pounds or grams to the nearest 100 g using scales calibrated semiannually (25). BMI was calculated as weight in kilograms divided by height in meters squared. Age- and sex-specific BMI z scores and percentiles were calculated based on the Centers for Disease Control and Prevention (CDC) 2000 growth charts (26). Overweight or obese status was defined as BMI ≥85th percentile and obesity as BMI ≥95th percentile. The primary outcome was childhood obesity at 2 time points. Overweight or obesity (hereafter referred to as overweight unless otherwise noted) and BMI z scores were modeled as the secondary outcomes. Statistical analysis The association between prenatal antibiotic exposure and childhood overweight and obesity was initially evaluated by log-binomial regression. Because convergence problems arose with binomial regression models, modified Poisson regression was finally used to estimate risk ratios and 95% confidence intervals (27). To account for correlation in the data resulting from sibling clusters, generalized estimating equations were used to provide robust standard errors. Linear mixed models with an unstructured correlation matrix and a random intercept were used to assess the association of antibiotic exposure with BMI z score. Separate models were fitted for each outcome at ages 4 and 7 years, respectively. We examined whether the exposure-outcome association differed across pregnancy by fitting trimester-specific models and investigated the effect estimates of cumulative exposure to antibiotics by categorizing the number of antibiotic exposures as none, 1, 2–3, or ≥4 times. Interactions between timing of first antibiotic use and the number of uses were evaluated by dividing them into 7 groups: No antibiotic (referent), single use in first trimester, repeated use in first trimester, single use in second trimester, repeated use in second trimester, single use in third trimester, and repeated use in third trimester. The impacts of different types of antibiotics were determined through similar analyses within narrow- and broad-spectrum categories as well as within specific antibiotic types. Models adjusted for potential confounders selected a priori, including maternal age (continuous), race (white, black, and other), socioeconomic index (5 categories), smoking during pregnancy (yes/no), prepregnancy BMI (continuous), and center (12 strata). Other potential confounders were considered using a change in risk ratio of ≥10%. None of the tested variables (e.g., child sex, mode of delivery) except parity (0, 1, >1 prior births) was added for adjustment. Because of missing values for prepregnancy BMI (8% of observations) and several other covariates, a multiple imputation approach (by chained equations with 5 replicates) was used for covariates with missing data (28). Predictors used in this imputation were variables listed in Table 1 and mode of delivery, study center, and childhood BMI. We evaluated effect modification by child sex, race, prepregnancy BMI status (<25, ≥25), and mode of delivery (vaginal, cesarean) by including a cross-product term for these variables. Table 1. Maternal and Child Characteristics According to Antibiotic Exposure During Pregnancy and Child Body Mass Index at Ages 4 and 7 Yearsa, Collaborative Perinatal Project, United States, 1959–1976 Characteristic Prenatal Antibiotic Exposure BMI z Scoreb Exposed (n = 10,908) Unexposed (n = 32,424) Age 4 Years (n = 33,228) Age 7 Years (n = 39,615) No. % No. % No. Mean (SD) No. Mean (SD) Maternal age group, years  <20 2,624 24.1 7,606 23.5 7,864 0.19 (1.13) 9,284 −0.06 (1.00)  20–34 7,479 68.6 22,186 68.5 22,632 0.32 (1.11) 27,135 0.01 (0.98)  ≥35 792 7.3 2,594 8.0 2,686 0.37 (1.22) 3,183 0.08 (1.09) Prepregnancy BMIc  <25.0 7,734 76.3 23,061 77.5 23,767 0.20 (1.12) 28,096 −0.10 (0.98)  25.0–29.9 1,679 16.6 4,662 15.7 4,968 0.47 (1.11) 5,801 0.23 (0.98)  ≥30.0 729 7.2 2,026 6.8 2,168 0.61 (1.15) 2,555 0.38 (1.05) Maternal race/ethnicity  White 4,425 40.6 15,100 46.6 14,555 0.54 (1.04) 18,360 0.16 (0.92)  Black 5,881 53.9 15,404 47.5 16,633 0.05 (1.14) 19,618 −0.15 (1.03)  Other 602 5.5 1,920 5.9 2,040 0.54 (1.20) 1,637 0.09 (1.07) Parity  0 3,132 28.8 9,584 29.6 9,774 0.29 (1.15) 11,529 0.02 (1.01)  1 2,311 21.2 7,441 23.0 7,433 0.34 (1.13) 8,826 0.04 (0.97)  >1 5,441 50.0 15,345 47.4 15,969 0.28 (1.12) 19,186 −0.03 (0.99) Education, years  ≤9 3,229 30.0 9,048 28.4 9,497 0.23 (1.12) 11,020 −0.07 (1.03)  10–12 6,489 60.3 19,066 59.9 19,864 0.31 (1.13) 23,510 0.01 (0.99)  >12 1,037 9.6 3,713 11.7 3,463 0.35 (1.09) 4,407 0.09 (0.89) Socioeconomic indexd  1 (lowest) 917 8.6 2,480 7.9 2,627 0.01 (1.16) 3,134 −0.21 (1.04)  2 3,545 33.3 9,705 30.8 10,271 0.18 (1.14) 11,953 −0.09 (1.03)  3 3,364 31.6 9,860 31.3 10,260 0.34 (1.13) 12,028 0.03 (0.99)  4 1,952 18.3 6,308 20.0 6,319 0.47 (1.10) 7,647 0.13 (0.95)  5 (highest) 879 8.2 3,189 10.1 3,088 0.42 (1.05) 3,806 0.10 (0.89) Gestational age, weeks  <37 1,632 15.0 4,967 15.3 4,899 0.18 (1.14) 6,025 −0.10 (1.02)  ≥37 9,273 85.0 27,441 84.7 28,316 0.32 (1.12) 33,571 0.02 (0.98) Diabetes  No 10,561 97.3 31,585 98.0 32,315 0.29 (1.13) 38,516 0.00 (0.99)  Yes 297 2.7 658 2.0 766 0.62 (1.16) 884 0.29 (1.04) Maternal smoking  No 5,611 51.5 17,455 54.0 17,853 0.25 (1.14) 20,960 −0.04 (0.99)  Yes 5,274 48.5 14,873 46.0 15,297 0.36 (1.11) 18,576 0.05 (0.99) Child’s sex  Male 5,552 50.9 16,368 50.5 16,793 0.36 (1.13) 20,060 0.06 (0.98)  Female 5,355 49.1 16,053 49.5 16,435 0.23 (1.12) 19,555 −0.06 (1.01) Birth weight, g  <2,500 1,078 9.9 2,945 9.1 3,056 −0.14 (1.25) 3,684 −0.36 (1.11)  2,500–3,999 9,257 85.1 27,689 85.6 28,363 0.31 (1.10) 33,781 0.01 (0.97)  ≥4,000 548 5.0 1,707 5.3 1,763 0.88 (0.98) 2,137 0.44 (0.91) Characteristic Prenatal Antibiotic Exposure BMI z Scoreb Exposed (n = 10,908) Unexposed (n = 32,424) Age 4 Years (n = 33,228) Age 7 Years (n = 39,615) No. % No. % No. Mean (SD) No. Mean (SD) Maternal age group, years  <20 2,624 24.1 7,606 23.5 7,864 0.19 (1.13) 9,284 −0.06 (1.00)  20–34 7,479 68.6 22,186 68.5 22,632 0.32 (1.11) 27,135 0.01 (0.98)  ≥35 792 7.3 2,594 8.0 2,686 0.37 (1.22) 3,183 0.08 (1.09) Prepregnancy BMIc  <25.0 7,734 76.3 23,061 77.5 23,767 0.20 (1.12) 28,096 −0.10 (0.98)  25.0–29.9 1,679 16.6 4,662 15.7 4,968 0.47 (1.11) 5,801 0.23 (0.98)  ≥30.0 729 7.2 2,026 6.8 2,168 0.61 (1.15) 2,555 0.38 (1.05) Maternal race/ethnicity  White 4,425 40.6 15,100 46.6 14,555 0.54 (1.04) 18,360 0.16 (0.92)  Black 5,881 53.9 15,404 47.5 16,633 0.05 (1.14) 19,618 −0.15 (1.03)  Other 602 5.5 1,920 5.9 2,040 0.54 (1.20) 1,637 0.09 (1.07) Parity  0 3,132 28.8 9,584 29.6 9,774 0.29 (1.15) 11,529 0.02 (1.01)  1 2,311 21.2 7,441 23.0 7,433 0.34 (1.13) 8,826 0.04 (0.97)  >1 5,441 50.0 15,345 47.4 15,969 0.28 (1.12) 19,186 −0.03 (0.99) Education, years  ≤9 3,229 30.0 9,048 28.4 9,497 0.23 (1.12) 11,020 −0.07 (1.03)  10–12 6,489 60.3 19,066 59.9 19,864 0.31 (1.13) 23,510 0.01 (0.99)  >12 1,037 9.6 3,713 11.7 3,463 0.35 (1.09) 4,407 0.09 (0.89) Socioeconomic indexd  1 (lowest) 917 8.6 2,480 7.9 2,627 0.01 (1.16) 3,134 −0.21 (1.04)  2 3,545 33.3 9,705 30.8 10,271 0.18 (1.14) 11,953 −0.09 (1.03)  3 3,364 31.6 9,860 31.3 10,260 0.34 (1.13) 12,028 0.03 (0.99)  4 1,952 18.3 6,308 20.0 6,319 0.47 (1.10) 7,647 0.13 (0.95)  5 (highest) 879 8.2 3,189 10.1 3,088 0.42 (1.05) 3,806 0.10 (0.89) Gestational age, weeks  <37 1,632 15.0 4,967 15.3 4,899 0.18 (1.14) 6,025 −0.10 (1.02)  ≥37 9,273 85.0 27,441 84.7 28,316 0.32 (1.12) 33,571 0.02 (0.98) Diabetes  No 10,561 97.3 31,585 98.0 32,315 0.29 (1.13) 38,516 0.00 (0.99)  Yes 297 2.7 658 2.0 766 0.62 (1.16) 884 0.29 (1.04) Maternal smoking  No 5,611 51.5 17,455 54.0 17,853 0.25 (1.14) 20,960 −0.04 (0.99)  Yes 5,274 48.5 14,873 46.0 15,297 0.36 (1.11) 18,576 0.05 (0.99) Child’s sex  Male 5,552 50.9 16,368 50.5 16,793 0.36 (1.13) 20,060 0.06 (0.98)  Female 5,355 49.1 16,053 49.5 16,435 0.23 (1.12) 19,555 −0.06 (1.01) Birth weight, g  <2,500 1,078 9.9 2,945 9.1 3,056 −0.14 (1.25) 3,684 −0.36 (1.11)  2,500–3,999 9,257 85.1 27,689 85.6 28,363 0.31 (1.10) 33,781 0.01 (0.97)  ≥4,000 548 5.0 1,707 5.3 1,763 0.88 (0.98) 2,137 0.44 (0.91) Abbreviations: BMI, body mass index; SD, standard deviation. a Among the included subjects, 51 were missing data for maternal age, 3,441 for BMI, 78 for parity, 750 for education, 1,133 for socioeconomic index, 19 for gestational age, 231 for diabetes, 119 for smoking, 4 for child sex, and 108 for birth weight. b Based on the 2000 US Centers for Disease Control and Prevention growth charts. c Weight (kg)/height (m)2. d Socioeconomic index was a combined score of maternal education, occupation, and family income, and was classified into 5 categories (44). Table 1. Maternal and Child Characteristics According to Antibiotic Exposure During Pregnancy and Child Body Mass Index at Ages 4 and 7 Yearsa, Collaborative Perinatal Project, United States, 1959–1976 Characteristic Prenatal Antibiotic Exposure BMI z Scoreb Exposed (n = 10,908) Unexposed (n = 32,424) Age 4 Years (n = 33,228) Age 7 Years (n = 39,615) No. % No. % No. Mean (SD) No. Mean (SD) Maternal age group, years  <20 2,624 24.1 7,606 23.5 7,864 0.19 (1.13) 9,284 −0.06 (1.00)  20–34 7,479 68.6 22,186 68.5 22,632 0.32 (1.11) 27,135 0.01 (0.98)  ≥35 792 7.3 2,594 8.0 2,686 0.37 (1.22) 3,183 0.08 (1.09) Prepregnancy BMIc  <25.0 7,734 76.3 23,061 77.5 23,767 0.20 (1.12) 28,096 −0.10 (0.98)  25.0–29.9 1,679 16.6 4,662 15.7 4,968 0.47 (1.11) 5,801 0.23 (0.98)  ≥30.0 729 7.2 2,026 6.8 2,168 0.61 (1.15) 2,555 0.38 (1.05) Maternal race/ethnicity  White 4,425 40.6 15,100 46.6 14,555 0.54 (1.04) 18,360 0.16 (0.92)  Black 5,881 53.9 15,404 47.5 16,633 0.05 (1.14) 19,618 −0.15 (1.03)  Other 602 5.5 1,920 5.9 2,040 0.54 (1.20) 1,637 0.09 (1.07) Parity  0 3,132 28.8 9,584 29.6 9,774 0.29 (1.15) 11,529 0.02 (1.01)  1 2,311 21.2 7,441 23.0 7,433 0.34 (1.13) 8,826 0.04 (0.97)  >1 5,441 50.0 15,345 47.4 15,969 0.28 (1.12) 19,186 −0.03 (0.99) Education, years  ≤9 3,229 30.0 9,048 28.4 9,497 0.23 (1.12) 11,020 −0.07 (1.03)  10–12 6,489 60.3 19,066 59.9 19,864 0.31 (1.13) 23,510 0.01 (0.99)  >12 1,037 9.6 3,713 11.7 3,463 0.35 (1.09) 4,407 0.09 (0.89) Socioeconomic indexd  1 (lowest) 917 8.6 2,480 7.9 2,627 0.01 (1.16) 3,134 −0.21 (1.04)  2 3,545 33.3 9,705 30.8 10,271 0.18 (1.14) 11,953 −0.09 (1.03)  3 3,364 31.6 9,860 31.3 10,260 0.34 (1.13) 12,028 0.03 (0.99)  4 1,952 18.3 6,308 20.0 6,319 0.47 (1.10) 7,647 0.13 (0.95)  5 (highest) 879 8.2 3,189 10.1 3,088 0.42 (1.05) 3,806 0.10 (0.89) Gestational age, weeks  <37 1,632 15.0 4,967 15.3 4,899 0.18 (1.14) 6,025 −0.10 (1.02)  ≥37 9,273 85.0 27,441 84.7 28,316 0.32 (1.12) 33,571 0.02 (0.98) Diabetes  No 10,561 97.3 31,585 98.0 32,315 0.29 (1.13) 38,516 0.00 (0.99)  Yes 297 2.7 658 2.0 766 0.62 (1.16) 884 0.29 (1.04) Maternal smoking  No 5,611 51.5 17,455 54.0 17,853 0.25 (1.14) 20,960 −0.04 (0.99)  Yes 5,274 48.5 14,873 46.0 15,297 0.36 (1.11) 18,576 0.05 (0.99) Child’s sex  Male 5,552 50.9 16,368 50.5 16,793 0.36 (1.13) 20,060 0.06 (0.98)  Female 5,355 49.1 16,053 49.5 16,435 0.23 (1.12) 19,555 −0.06 (1.01) Birth weight, g  <2,500 1,078 9.9 2,945 9.1 3,056 −0.14 (1.25) 3,684 −0.36 (1.11)  2,500–3,999 9,257 85.1 27,689 85.6 28,363 0.31 (1.10) 33,781 0.01 (0.97)  ≥4,000 548 5.0 1,707 5.3 1,763 0.88 (0.98) 2,137 0.44 (0.91) Characteristic Prenatal Antibiotic Exposure BMI z Scoreb Exposed (n = 10,908) Unexposed (n = 32,424) Age 4 Years (n = 33,228) Age 7 Years (n = 39,615) No. % No. % No. Mean (SD) No. Mean (SD) Maternal age group, years  <20 2,624 24.1 7,606 23.5 7,864 0.19 (1.13) 9,284 −0.06 (1.00)  20–34 7,479 68.6 22,186 68.5 22,632 0.32 (1.11) 27,135 0.01 (0.98)  ≥35 792 7.3 2,594 8.0 2,686 0.37 (1.22) 3,183 0.08 (1.09) Prepregnancy BMIc  <25.0 7,734 76.3 23,061 77.5 23,767 0.20 (1.12) 28,096 −0.10 (0.98)  25.0–29.9 1,679 16.6 4,662 15.7 4,968 0.47 (1.11) 5,801 0.23 (0.98)  ≥30.0 729 7.2 2,026 6.8 2,168 0.61 (1.15) 2,555 0.38 (1.05) Maternal race/ethnicity  White 4,425 40.6 15,100 46.6 14,555 0.54 (1.04) 18,360 0.16 (0.92)  Black 5,881 53.9 15,404 47.5 16,633 0.05 (1.14) 19,618 −0.15 (1.03)  Other 602 5.5 1,920 5.9 2,040 0.54 (1.20) 1,637 0.09 (1.07) Parity  0 3,132 28.8 9,584 29.6 9,774 0.29 (1.15) 11,529 0.02 (1.01)  1 2,311 21.2 7,441 23.0 7,433 0.34 (1.13) 8,826 0.04 (0.97)  >1 5,441 50.0 15,345 47.4 15,969 0.28 (1.12) 19,186 −0.03 (0.99) Education, years  ≤9 3,229 30.0 9,048 28.4 9,497 0.23 (1.12) 11,020 −0.07 (1.03)  10–12 6,489 60.3 19,066 59.9 19,864 0.31 (1.13) 23,510 0.01 (0.99)  >12 1,037 9.6 3,713 11.7 3,463 0.35 (1.09) 4,407 0.09 (0.89) Socioeconomic indexd  1 (lowest) 917 8.6 2,480 7.9 2,627 0.01 (1.16) 3,134 −0.21 (1.04)  2 3,545 33.3 9,705 30.8 10,271 0.18 (1.14) 11,953 −0.09 (1.03)  3 3,364 31.6 9,860 31.3 10,260 0.34 (1.13) 12,028 0.03 (0.99)  4 1,952 18.3 6,308 20.0 6,319 0.47 (1.10) 7,647 0.13 (0.95)  5 (highest) 879 8.2 3,189 10.1 3,088 0.42 (1.05) 3,806 0.10 (0.89) Gestational age, weeks  <37 1,632 15.0 4,967 15.3 4,899 0.18 (1.14) 6,025 −0.10 (1.02)  ≥37 9,273 85.0 27,441 84.7 28,316 0.32 (1.12) 33,571 0.02 (0.98) Diabetes  No 10,561 97.3 31,585 98.0 32,315 0.29 (1.13) 38,516 0.00 (0.99)  Yes 297 2.7 658 2.0 766 0.62 (1.16) 884 0.29 (1.04) Maternal smoking  No 5,611 51.5 17,455 54.0 17,853 0.25 (1.14) 20,960 −0.04 (0.99)  Yes 5,274 48.5 14,873 46.0 15,297 0.36 (1.11) 18,576 0.05 (0.99) Child’s sex  Male 5,552 50.9 16,368 50.5 16,793 0.36 (1.13) 20,060 0.06 (0.98)  Female 5,355 49.1 16,053 49.5 16,435 0.23 (1.12) 19,555 −0.06 (1.01) Birth weight, g  <2,500 1,078 9.9 2,945 9.1 3,056 −0.14 (1.25) 3,684 −0.36 (1.11)  2,500–3,999 9,257 85.1 27,689 85.6 28,363 0.31 (1.10) 33,781 0.01 (0.97)  ≥4,000 548 5.0 1,707 5.3 1,763 0.88 (0.98) 2,137 0.44 (0.91) Abbreviations: BMI, body mass index; SD, standard deviation. a Among the included subjects, 51 were missing data for maternal age, 3,441 for BMI, 78 for parity, 750 for education, 1,133 for socioeconomic index, 19 for gestational age, 231 for diabetes, 119 for smoking, 4 for child sex, and 108 for birth weight. b Based on the 2000 US Centers for Disease Control and Prevention growth charts. c Weight (kg)/height (m)2. d Socioeconomic index was a combined score of maternal education, occupation, and family income, and was classified into 5 categories (44). The robustness of our findings was tested by restricting the analysis to mother-child dyads with complete covariate data. As a sensitivity analysis, regression models were refitted with the outcomes of childhood overweight and obesity that were determined by the International Obesity Task Force age- and sex-specific cutoffs of BMI (29). To assess the potential bias of loss to follow-up, we conducted another sensitivity analysis using inverse probability weights for successful follow-up (30). The probability of successful follow-up was calculated in logistic models (separately for 4-year and 7-year visits) that included terms for the above-mentioned confounders plus maternal education, birth weight, gestational age, mode of delivery, and diabetes before and during pregnancy. All statistical analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina). RESULTS Among 53,647 live singleton children in the CPP, 43,665 completed follow-up visits at age 4 or 7 years. Exclusion of children with missing data on prenatal antibiotic exposure and valid BMI measurements resulted in a final study sample of 43,332 children (Web Figure 1, available at https://academic.oup.com/aje). Of these, 33,228 were eligible for 4-year outcome analysis and 39,615 for 7-year outcome analysis. In this study, 10,908 (25.2%) women used antibiotics during pregnancy. Generally, women with antibiotic exposure were more likely to be younger, black, overweight or obese before pregnancy, at a lower socioeconomic level, or smokers compared with those without exposure (Table 1). Of the exposed mothers, more than 40% had repeated use of antibiotics over the entire pregnancy. Penicillins were the most used (56.7%), followed by sulfonamides (39.8%), tetracyclines (13.7%), nitrofurans (4.8%), and other antibiotics were used less frequently (Web Table 1). Participants included in the analysis were more likely to be black than were those excluded, and distributions for other baseline characteristics were similar (Web Table 2). At age 4 years, the prevalence of childhood overweight and obesity was 24.7% and of obesity was 8.9%. In the crude analysis, a tendency of inverse relationship between prenatal antibiotic exposure and overweight/obesity was identified (Web Table 3). However, after adjusting for potential confounders, prenatal antibiotic use was not associated with 4-year overweight or obesity in children, either cumulatively or according to trimester (Table 2). Table 2. Adjusted Risk Ratios for Childhood Obesity at Ages 4 and 7 Years According to Prenatal Exposure to Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years No antibiotic 24,636 6,168 1.00 Referent 2,250 1.00 Referent Any antibiotics 8,592 2,026 1.01 0.97, 1.06 699 0.98 0.90, 1.06 Timing of first exposurec  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  First trimester 2,702 636 1.00 0.93, 1.07 208 0.92 0.81, 1.06  Second trimester 3,252 783 1.04 0.98, 1.11 266 0.99 0.88, 1.12  Third trimester 2,556 585 0.99 0.92, 1.07 218 1.01 0.89, 1.15 No. of exposures  0 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 5,087 1,247 1.04 0.99, 1.09 431 0.99 0.90, 1.09  2–3 2,660 579 0.99 0.92, 1.07 207 1.00 0.88, 1.15  ≥4 845 200 0.93 0.82, 1.05 61 0.79 0.62, 1.01   P for trend 0.73 0.25 Follow-up at 7 Years No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent Any antibiotics 9,974 1,300 1.04 0.98, 1.10 471 1.09 0.98, 1.20 Timing of first exposured  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  First trimester 3,150 394 0.99 0.90, 1.09 138 1.01 0.86, 1.20  Second trimester 3,764 518 1.11 1.02, 1.20 190 1.19 1.03, 1.38  Third trimester 2,930 370 1.01 0.91, 1.11 136 1.04 0.87, 1.23 No. of exposures  0 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 5,928 769 1.02 0.95, 1.09 260 0.98 0.86, 1.11  2–3 3,093 406 1.08 0.98, 1.19 157 1.22 1.04, 1.44  ≥4 953 125 1.03 0.87, 1.21 54 1.34 1.03, 1.73   P for trend 0.15 <0.01 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years No antibiotic 24,636 6,168 1.00 Referent 2,250 1.00 Referent Any antibiotics 8,592 2,026 1.01 0.97, 1.06 699 0.98 0.90, 1.06 Timing of first exposurec  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  First trimester 2,702 636 1.00 0.93, 1.07 208 0.92 0.81, 1.06  Second trimester 3,252 783 1.04 0.98, 1.11 266 0.99 0.88, 1.12  Third trimester 2,556 585 0.99 0.92, 1.07 218 1.01 0.89, 1.15 No. of exposures  0 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 5,087 1,247 1.04 0.99, 1.09 431 0.99 0.90, 1.09  2–3 2,660 579 0.99 0.92, 1.07 207 1.00 0.88, 1.15  ≥4 845 200 0.93 0.82, 1.05 61 0.79 0.62, 1.01   P for trend 0.73 0.25 Follow-up at 7 Years No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent Any antibiotics 9,974 1,300 1.04 0.98, 1.10 471 1.09 0.98, 1.20 Timing of first exposured  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  First trimester 3,150 394 0.99 0.90, 1.09 138 1.01 0.86, 1.20  Second trimester 3,764 518 1.11 1.02, 1.20 190 1.19 1.03, 1.38  Third trimester 2,930 370 1.01 0.91, 1.11 136 1.04 0.87, 1.23 No. of exposures  0 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 5,928 769 1.02 0.95, 1.09 260 0.98 0.86, 1.11  2–3 3,093 406 1.08 0.98, 1.19 157 1.22 1.04, 1.44  ≥4 953 125 1.03 0.87, 1.21 54 1.34 1.03, 1.73   P for trend 0.15 <0.01 Abbreviations: BMI, body mass index; CI, confidence interval; RR, risk ratio. a Includes children classified as overweight or obese. Overweight or obese status was defined as BMI ≥85th percentile and obesity as BMI ≥95th percentile. b Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. c Eighty-two participants with unknown exposure month were not included. d One hundred thirty participants with unknown exposure month were not included. Table 2. Adjusted Risk Ratios for Childhood Obesity at Ages 4 and 7 Years According to Prenatal Exposure to Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years No antibiotic 24,636 6,168 1.00 Referent 2,250 1.00 Referent Any antibiotics 8,592 2,026 1.01 0.97, 1.06 699 0.98 0.90, 1.06 Timing of first exposurec  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  First trimester 2,702 636 1.00 0.93, 1.07 208 0.92 0.81, 1.06  Second trimester 3,252 783 1.04 0.98, 1.11 266 0.99 0.88, 1.12  Third trimester 2,556 585 0.99 0.92, 1.07 218 1.01 0.89, 1.15 No. of exposures  0 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 5,087 1,247 1.04 0.99, 1.09 431 0.99 0.90, 1.09  2–3 2,660 579 0.99 0.92, 1.07 207 1.00 0.88, 1.15  ≥4 845 200 0.93 0.82, 1.05 61 0.79 0.62, 1.01   P for trend 0.73 0.25 Follow-up at 7 Years No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent Any antibiotics 9,974 1,300 1.04 0.98, 1.10 471 1.09 0.98, 1.20 Timing of first exposured  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  First trimester 3,150 394 0.99 0.90, 1.09 138 1.01 0.86, 1.20  Second trimester 3,764 518 1.11 1.02, 1.20 190 1.19 1.03, 1.38  Third trimester 2,930 370 1.01 0.91, 1.11 136 1.04 0.87, 1.23 No. of exposures  0 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 5,928 769 1.02 0.95, 1.09 260 0.98 0.86, 1.11  2–3 3,093 406 1.08 0.98, 1.19 157 1.22 1.04, 1.44  ≥4 953 125 1.03 0.87, 1.21 54 1.34 1.03, 1.73   P for trend 0.15 <0.01 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years No antibiotic 24,636 6,168 1.00 Referent 2,250 1.00 Referent Any antibiotics 8,592 2,026 1.01 0.97, 1.06 699 0.98 0.90, 1.06 Timing of first exposurec  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  First trimester 2,702 636 1.00 0.93, 1.07 208 0.92 0.81, 1.06  Second trimester 3,252 783 1.04 0.98, 1.11 266 0.99 0.88, 1.12  Third trimester 2,556 585 0.99 0.92, 1.07 218 1.01 0.89, 1.15 No. of exposures  0 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 5,087 1,247 1.04 0.99, 1.09 431 0.99 0.90, 1.09  2–3 2,660 579 0.99 0.92, 1.07 207 1.00 0.88, 1.15  ≥4 845 200 0.93 0.82, 1.05 61 0.79 0.62, 1.01   P for trend 0.73 0.25 Follow-up at 7 Years No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent Any antibiotics 9,974 1,300 1.04 0.98, 1.10 471 1.09 0.98, 1.20 Timing of first exposured  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  First trimester 3,150 394 0.99 0.90, 1.09 138 1.01 0.86, 1.20  Second trimester 3,764 518 1.11 1.02, 1.20 190 1.19 1.03, 1.38  Third trimester 2,930 370 1.01 0.91, 1.11 136 1.04 0.87, 1.23 No. of exposures  0 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 5,928 769 1.02 0.95, 1.09 260 0.98 0.86, 1.11  2–3 3,093 406 1.08 0.98, 1.19 157 1.22 1.04, 1.44  ≥4 953 125 1.03 0.87, 1.21 54 1.34 1.03, 1.73   P for trend 0.15 <0.01 Abbreviations: BMI, body mass index; CI, confidence interval; RR, risk ratio. a Includes children classified as overweight or obese. Overweight or obese status was defined as BMI ≥85th percentile and obesity as BMI ≥95th percentile. b Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. c Eighty-two participants with unknown exposure month were not included. d One hundred thirty participants with unknown exposure month were not included. At age 7 years, prevalence of childhood overweight and obesity was 13.1% and of obesity was 4.6%. After adjustment for covariates, there was no general association between prenatal exposure to antibiotics and age-7-years overweight or obesity (Table 2). Nevertheless, the analysis according to time at first antibiotic use demonstrated a significant association of offspring overweight (risk ratio (RR) = 1.11, 95% confidence interval (CI): 1.02, 1.20) and obesity (RR = 1.19, 95% CI: 1.03, 1.38) with exposure in the second trimester. Children of mothers with repeated antibiotic use during pregnancy had significantly higher risk of obesity compared with those of mothers without antibiotic use (2–3 times, RR = 1.22, 95% CI: 1.04, 1.44; ≥4 times, RR = 1.34, 95% CI: 1.03, 1.73). As for the type of antibiotics, the obesity risk at age 7 years increased with repeated use of narrow-spectrum (≥2 times, RR = 1.31, 95% CI: 1.09, 1.56) as well as broad-spectrum drugs (≥2 times, RR = 1.23, 95% CI: 1.04, 1.46) (Table 3). Analysis of the interaction between the timing of first exposure and the number of exposures revealed that repeated antibiotic use (≥2 times) in the second trimester was the major contributor to the risk of age-7-years overweight (RR = 1.15, 95% CI: 1.02, 1.29) and obesity (RR = 1.40, 95% CI: 1.16, 1.71) (Figure 1). Prenatal antibiotic exposure was not associated with the child’s BMI z score at the 2 time points (Table 4). Table 3. Adjusted Risk Ratios for Childhood Obesity at Ages 4 and 7 Years According to Prenatal Exposure to Different Types of Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years Narrow spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,928 750 1.05 0.99, 1.12 249 0.96 0.85, 1.08  ≥2 1,883 429 0.96 0.88, 1.04 160 1.01 0.87, 1.17 Broad spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,456 566 1.02 0.95, 1.10 206 1.04 0.91, 1.19  ≥2 2,523 535 0.97 0.90, 1.05 179 0.93 0.81, 1.08 Follow-up at 7 Years Narrow spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 3,490 452 0.99 0.90, 1.08 163 1.00 0.86, 1.17  ≥2 2,214 305 1.09 0.98, 1.22 124 1.31 1.09, 1.56 Broad spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 2,778 358 1.05 0.96, 1.16 116 0.98 0.82, 1.18  ≥2 2,887 360 1.05 0.95, 1.16 144 1.23 1.04, 1.46 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years Narrow spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,928 750 1.05 0.99, 1.12 249 0.96 0.85, 1.08  ≥2 1,883 429 0.96 0.88, 1.04 160 1.01 0.87, 1.17 Broad spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,456 566 1.02 0.95, 1.10 206 1.04 0.91, 1.19  ≥2 2,523 535 0.97 0.90, 1.05 179 0.93 0.81, 1.08 Follow-up at 7 Years Narrow spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 3,490 452 0.99 0.90, 1.08 163 1.00 0.86, 1.17  ≥2 2,214 305 1.09 0.98, 1.22 124 1.31 1.09, 1.56 Broad spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 2,778 358 1.05 0.96, 1.16 116 0.98 0.82, 1.18  ≥2 2,887 360 1.05 0.95, 1.16 144 1.23 1.04, 1.46 Abbreviations: BMI, body mass index; CI, confidence interval; RR, risk ratio. a Includes children classified as overweight or obese. Overweight or obese status was defined as BMI ≥85th percentile and obesity as BMI ≥95th percentile. b Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. Table 3. Adjusted Risk Ratios for Childhood Obesity at Ages 4 and 7 Years According to Prenatal Exposure to Different Types of Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years Narrow spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,928 750 1.05 0.99, 1.12 249 0.96 0.85, 1.08  ≥2 1,883 429 0.96 0.88, 1.04 160 1.01 0.87, 1.17 Broad spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,456 566 1.02 0.95, 1.10 206 1.04 0.91, 1.19  ≥2 2,523 535 0.97 0.90, 1.05 179 0.93 0.81, 1.08 Follow-up at 7 Years Narrow spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 3,490 452 0.99 0.90, 1.08 163 1.00 0.86, 1.17  ≥2 2,214 305 1.09 0.98, 1.22 124 1.31 1.09, 1.56 Broad spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 2,778 358 1.05 0.96, 1.16 116 0.98 0.82, 1.18  ≥2 2,887 360 1.05 0.95, 1.16 144 1.23 1.04, 1.46 Antibiotic Exposure Overall Overweight or Obesea Obese No. RRb 95% CI No. RRb 95% CI Follow-up at 4 Years Narrow spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,928 750 1.05 0.99, 1.12 249 0.96 0.85, 1.08  ≥2 1,883 429 0.96 0.88, 1.04 160 1.01 0.87, 1.17 Broad spectrum  No antibiotics 24,636 6,168 1.00 Referent 2,250 1.00 Referent  1 2,456 566 1.02 0.95, 1.10 206 1.04 0.91, 1.19  ≥2 2,523 535 0.97 0.90, 1.05 179 0.93 0.81, 1.08 Follow-up at 7 Years Narrow spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 3,490 452 0.99 0.90, 1.08 163 1.00 0.86, 1.17  ≥2 2,214 305 1.09 0.98, 1.22 124 1.31 1.09, 1.56 Broad spectrum  No antibiotics 29,641 3,883 1.00 Referent 1,340 1.00 Referent  1 2,778 358 1.05 0.96, 1.16 116 0.98 0.82, 1.18  ≥2 2,887 360 1.05 0.95, 1.16 144 1.23 1.04, 1.46 Abbreviations: BMI, body mass index; CI, confidence interval; RR, risk ratio. a Includes children classified as overweight or obese. Overweight or obese status was defined as BMI ≥85th percentile and obesity as BMI ≥95th percentile. b Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. Table 4. Estimated Differences in Child Body Mass Index z Score at Ages 4 and 7 Years According to Prenatal Exposure to Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Follow-up at 4 Years Follow-up at 7 Years βa 95% CI βa 95% CI No antibiotics 0 Referent 0 Referent Any antibiotics −0.006 −0.032, 0.020 0.006 −0.016, 0.027 Timing of first exposureb  No antibiotics 0 Referent 0 Referent  First trimester −0.014 −0.055, 0.028 0.005 −0.029, 0.039  Second trimester 0.014 −0.024, 0.052 0.020 −0.011, 0.052  Third trimester −0.023 −0.067, 0.021 −0.010 −0.047, 0.027 Number of exposures  0 0 Referent 0 Referent  1 0.002 −0.029, 0.034 0.011 −0.016, 0.038  2–3 −0.014 −0.057, 0.028 0.003 −0.032, 0.038  ≥4 −0.037 −0.105, 0.032 −0.015 −0.075, 0.046   P for trend 0.34 0.90 Narrow spectrum  No antibiotics 0 Referent 0 Referent  1 −0.013 −0.053, 0.028 0.003 −0.030, 0.037  ≥2 −0.011 −0.061, 0.039 0.018 −0.022, 0.059 Broad spectrum  No antibiotics 0 Referent 0 Referent  1 0.017 −0.027, 0.061 0.025 −0.012, 0.062  ≥2 −0.007 −0.050, 0.035 −0.006 −0.042, 0.030 Antibiotic Exposure Follow-up at 4 Years Follow-up at 7 Years βa 95% CI βa 95% CI No antibiotics 0 Referent 0 Referent Any antibiotics −0.006 −0.032, 0.020 0.006 −0.016, 0.027 Timing of first exposureb  No antibiotics 0 Referent 0 Referent  First trimester −0.014 −0.055, 0.028 0.005 −0.029, 0.039  Second trimester 0.014 −0.024, 0.052 0.020 −0.011, 0.052  Third trimester −0.023 −0.067, 0.021 −0.010 −0.047, 0.027 Number of exposures  0 0 Referent 0 Referent  1 0.002 −0.029, 0.034 0.011 −0.016, 0.038  2–3 −0.014 −0.057, 0.028 0.003 −0.032, 0.038  ≥4 −0.037 −0.105, 0.032 −0.015 −0.075, 0.046   P for trend 0.34 0.90 Narrow spectrum  No antibiotics 0 Referent 0 Referent  1 −0.013 −0.053, 0.028 0.003 −0.030, 0.037  ≥2 −0.011 −0.061, 0.039 0.018 −0.022, 0.059 Broad spectrum  No antibiotics 0 Referent 0 Referent  1 0.017 −0.027, 0.061 0.025 −0.012, 0.062  ≥2 −0.007 −0.050, 0.035 −0.006 −0.042, 0.030 Abbreviations: BMI, body mass index; CI, confidence interval. a Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. b Eighty-two and 130 participants with unknown exposure month were not included for 4-year measures and 7-year measures analysis, respectively. Table 4. Estimated Differences in Child Body Mass Index z Score at Ages 4 and 7 Years According to Prenatal Exposure to Antibiotics, Collaborative Perinatal Project, United States, 1959–1976 Antibiotic Exposure Follow-up at 4 Years Follow-up at 7 Years βa 95% CI βa 95% CI No antibiotics 0 Referent 0 Referent Any antibiotics −0.006 −0.032, 0.020 0.006 −0.016, 0.027 Timing of first exposureb  No antibiotics 0 Referent 0 Referent  First trimester −0.014 −0.055, 0.028 0.005 −0.029, 0.039  Second trimester 0.014 −0.024, 0.052 0.020 −0.011, 0.052  Third trimester −0.023 −0.067, 0.021 −0.010 −0.047, 0.027 Number of exposures  0 0 Referent 0 Referent  1 0.002 −0.029, 0.034 0.011 −0.016, 0.038  2–3 −0.014 −0.057, 0.028 0.003 −0.032, 0.038  ≥4 −0.037 −0.105, 0.032 −0.015 −0.075, 0.046   P for trend 0.34 0.90 Narrow spectrum  No antibiotics 0 Referent 0 Referent  1 −0.013 −0.053, 0.028 0.003 −0.030, 0.037  ≥2 −0.011 −0.061, 0.039 0.018 −0.022, 0.059 Broad spectrum  No antibiotics 0 Referent 0 Referent  1 0.017 −0.027, 0.061 0.025 −0.012, 0.062  ≥2 −0.007 −0.050, 0.035 −0.006 −0.042, 0.030 Antibiotic Exposure Follow-up at 4 Years Follow-up at 7 Years βa 95% CI βa 95% CI No antibiotics 0 Referent 0 Referent Any antibiotics −0.006 −0.032, 0.020 0.006 −0.016, 0.027 Timing of first exposureb  No antibiotics 0 Referent 0 Referent  First trimester −0.014 −0.055, 0.028 0.005 −0.029, 0.039  Second trimester 0.014 −0.024, 0.052 0.020 −0.011, 0.052  Third trimester −0.023 −0.067, 0.021 −0.010 −0.047, 0.027 Number of exposures  0 0 Referent 0 Referent  1 0.002 −0.029, 0.034 0.011 −0.016, 0.038  2–3 −0.014 −0.057, 0.028 0.003 −0.032, 0.038  ≥4 −0.037 −0.105, 0.032 −0.015 −0.075, 0.046   P for trend 0.34 0.90 Narrow spectrum  No antibiotics 0 Referent 0 Referent  1 −0.013 −0.053, 0.028 0.003 −0.030, 0.037  ≥2 −0.011 −0.061, 0.039 0.018 −0.022, 0.059 Broad spectrum  No antibiotics 0 Referent 0 Referent  1 0.017 −0.027, 0.061 0.025 −0.012, 0.062  ≥2 −0.007 −0.050, 0.035 −0.006 −0.042, 0.030 Abbreviations: BMI, body mass index; CI, confidence interval. a Adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy BMI, parity, and center. Multiple imputation was used for covariates with missing values. b Eighty-two and 130 participants with unknown exposure month were not included for 4-year measures and 7-year measures analysis, respectively. Figure 1. View largeDownload slide Adjusted risk ratios for childhood overweight and obesity at 7 years of age, associated with maternal trimester-specific antibiotic use and number of exposures, Collaborative Perinatal Project, United States, 1959–1976. Overweight or obese status (denoted with circles) was defined as BMI ≥85th percentile and obesity (denoted with squares) as BMI ≥95th percentile. Models adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy body mass index, parity, and center. Children born to mothers who did not take any antibiotics during pregnancy were the referent. Points indicate risk ratios, with 95% confidence intervals shown by lines. Filled circle or square: single antibiotic use during pregnancy; white circle or square: repeated use. Figure 1. View largeDownload slide Adjusted risk ratios for childhood overweight and obesity at 7 years of age, associated with maternal trimester-specific antibiotic use and number of exposures, Collaborative Perinatal Project, United States, 1959–1976. Overweight or obese status (denoted with circles) was defined as BMI ≥85th percentile and obesity (denoted with squares) as BMI ≥95th percentile. Models adjusted for maternal age, race, socioeconomic index, smoking during pregnancy, prepregnancy body mass index, parity, and center. Children born to mothers who did not take any antibiotics during pregnancy were the referent. Points indicate risk ratios, with 95% confidence intervals shown by lines. Filled circle or square: single antibiotic use during pregnancy; white circle or square: repeated use. Stratified analysis demonstrated the effect modification by mode of delivery on the association between antibiotic use and childhood obesity (Web Table 4). Compared with children born vaginally (RR = 1.18, 95% CI: 1.02, 1.38), those born by cesarean delivery presented a stronger association with 7-year obesity (RR = 1.77, 95% CI: 1.21, 2.60). There was no evidence of the child’s sex, maternal prepregnancy BMI, or race as the effect modifiers. Assessments of impacts of specific antibiotic types indicated that maternal exposure to tetracyclines (RR = 1.26, 95% CI: 1.01, 1.57) and aminoglycosides (RR = 1.58, 95% CI: 1.02, 2.45) was related to obesity at age 7 years (Web Figure 2). The results did not change materially after restricting the analysis to observations with complete covariate data (Web Table 5). When the models were refitted with outcomes of overweight and obesity defined by the cutpoints recommended by the International Obesity Task Force, results were consistent, confirming the association of repeated antibiotic exposures during pregnancy with mid-childhood obesity (Web Table 6). The results were similar when the data were weighted by the inverse probability of follow-up (or being examined) at each time point (Web Table 7). DISCUSSION Within the context of the longitudinal CPP data, we examined prenatal exposure to antibiotics in relation to childhood overweight and obesity at ages 4 and 7 years. Our study did not observe a significant general association between prenatal antibiotic use and childhood obesity. Repeated antibiotic use, however, was associated with age-7-years obesity, and the risk tended to increase with an increasing number of antibiotic uses. Antibiotic exposure during pregnancy was not associated with age-4-years obesity risk. Similarly, no associations with childhood BMI z score at the 2 time points were observed. There has been little information on prenatal antibiotic use and childhood obesity or BMI. Two prior studies indicated an association between prenatal exposure to antibiotics and an increased risk of offspring obesity at ages 7–16 years (18, 19). Consistent with our study, an increasing risk of obesity with higher numbers of maternal antibiotic prescriptions has been demonstrated (19). However, other investigators have recently published studies with conflicting results in relation to BMI in early childhood. One group reported that prenatal antibiotic use was associated with larger BMI at age 2 years (20), whereas the other failed to observe such a relationship with BMI at age 3 years (21). In comparisons with prior studies, the strength of the association observed in our study was relatively weak. However, the robustness and reliability are supported by serial sensitivity analyses. These findings may suggest a potential role for prenatal antibiotic use, particularly repeated use, in childhood adiposity at school age. The underlying mechanisms by which prenatal antibiotic exposure has growth-promotion effects may be related to antibiotic-induced perturbations of maternal microbiota. Altered maternal gut and vaginal microbiota would directly affect the postnatal metabolism for newborns by changing the composition of the “pioneer” microbiota upon delivery (31, 32). Mounting evidence has also emerged to indicate that the fetus may be exposed to maternal gut microbes prenatally; studies have found a diverse range of microbes in umbilical cord blood, placenta, amniotic fluid, meconium, and fetal membranes in healthy normal pregnancies (33). Importantly, the microbes in meconium had a distinct microbiome that included similarities with the amniotic fluid and placenta (34). These results suggest that maternal-fetal exchange of commensal microbiota might exist before birth (33–35). Intrauterine maternal microbiota can drive intestinal transcriptional programming in offspring that are consistent with adapting early postnatal immune and metabolic functions (36). In this light, antibiotic use, particular repeated use, in pregnancy is likely to alter maternal microbial communities and disrupt normal colonization of the fetus intestinal microbiome, which might have long-lasting metabolic consequences later in life (8–10). Antibiotic exposure during pregnancy has been associated with low birth weight and methylation of imprinted genes involved in early growth and development (37). Similarly, a nonsignificant inverse association was observed between prenatal antibiotic exposure and birth weight in the present study. Further adjustment for birth weight did not alter the associations between antibiotics and childhood obesity. It remains to be elucidated whether somatic epigenetic changes in response to prenatal antibiotic exposure also contribute to adiposity in childhood. Repeated exposure to antibiotics in the second trimester showed the most pronounced association with childhood obesity risk. The fetal intestine undergoes rapid development during mid-to-late gestation (38) and, therefore, this period could be a window of particular susceptibility to antibiotics. Additionally, evidence suggests that repeated antibiotic use could preclude the recovery of gut microbiota from antibiotic-induced perturbations, resulting in a persistent, antibiotic-altered microbiota composition (39). Our study identified mode of delivery as a potential effect modifier. Because antibiotics are routinely administered prophylactically in cesarean delivery, it is possible that intrapartum antibiotic exposure constitutes an additional perturbation to these children, which is manifested as a stronger association in this group. The present study has, to our knowledge, by far the largest sample size of the studies published on this topic. Height and weight of children from the CPP were measured by trained personnel who were unaware of the child’s exposure status; thus, differential misclassification of the outcomes was unlikely. Furthermore, CPP collected detailed demographic and medical information as well as patterns of antibiotic use at each prenatal visit, which made it possible for us to control for multiple confounders. The prospective design and long follow-up period enabled us to assess the associations in both early and mid-childhood. On the other hand, the CPP was conducted a few decades ago, creating a major limitation to the potential applicability of the results to today’s conditions. Some antibiotics have passed out of use, and the patterns of antibiotic use during pregnancy might have also changed. For example, sulfonamides and tetracyclines accounted for more than 80% of the broad-spectrum antibiotics issued during pregnancy in the CPP, and macrolides were the least used. Today, macrolides are the most commonly prescribed broad-spectrum agents (40). Currently, although sulfonamides and tetracyclines are not officially recommended for pregnant women unless indicated due to potential fetal risk (41), both have still been commonly dispensed (more than 2% of the time) due to lack of safety evidence to inform treatment decisions or inappropriate use in the United States during the past two decades (40, 42). Moreover, penicillins remain the most used antibiotics in pregnancy (22, 40, 42). As such, it is still relevant to assess the metabolic consequences of these antibiotics to better understand their long-term safety. Admittedly, not only the prevalence but also the nature of childhood obesity has changed over the past several decades (43). For example, excess energy intake among mothers and children, which was uncommon then, is one of the main contributors now. Lack of physical activity is another major driving factor for obesity. Unfortunately, we had no data from the CPP on maternal or childhood diets or physical activities. However, if these unmeasured factors were not associated with prenatal antibiotic use, they should not have significant confounding effects on the association between prenatal exposure to antibiotics and childhood obesity. Confounding by the indications for which antibiotics were used was another concern in this study. The CPP did not record information on the specific infection that led to the use of antibiotics. It is, therefore, possible that the observed association was confounded by the underlying infection. But this may not explain our results, as the most common infection diagnoses for antibiotic administration during pregnancy include acute upper respiratory infections and urinary tract infections, which suggests that antibiotics prescribed for common and acute infections are unlikely to affect offspring weight (21). Second, information on postnatal exposure to antibiotics was not available in the CPP. Thus, if maternal prenatal antibiotic use is correlated with infant use of antibiotics, the association in our study was at least partly due to the influence of postnatal antibiotic exposure. To the best of our knowledge, however, there is no evidence to support a strong correlation between maternal and infant antibiotic use. Last, information on antibiotic use during pregnancy relied on self-report. Even so, recall bias is unlikely to occur because the data were captured through multiple prenatal visits in different stages of pregnancy. If some women did underreport their antibiotic use and were misclassified as unexposed, the association between prenatal antibiotic use and child obesity would tend to be attenuated. In conclusion, our results suggest that repeated antibiotic use during pregnancy was associated with risk of childhood obesity. The obesity risk was greatest for children of mothers who used antibiotics repeatedly in the second trimester. Given the high antibiotic prescription rates for pregnant women and the obesity epidemic in children, future studies are warranted to confirm the relationship between prenatal exposure to antibiotics and childhood obesity in contemporary populations, and to determine how intrauterine exposure could induce metabolic disorders of offspring. ACKNOWLEDGMENTS Author affiliations: Ministry of Education–Shanghai Key Laboratory of Children’s Environmental Health, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China (Bin Wang, Chonghuai Yan, Fan Jiang, Fei Li, Jun Zhang); Department of Epidemiology and Biostatistics, University of South Carolina, Columbia, South Carolina (Jihong Liu); Department of Neonatology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China (Yongjun Zhang); and School of Public Health, Shanghai Jiao Tong University School of Medicine, Shanghai, China (Hui Wang). This work was supported by the National Basic Science Research Program, Ministry of Science and Technology of China (grant 2014CB943300); Ministry of Science and Technology of China (grant 2014DFG31460); and Shanghai Health and Family Planning Commission (grant GWIII-26, 15GWZK0401). We thank the US National Archives and Records Administration for providing data from the Collaborative Perinatal Project. Conflict of interest: none declared. Abbreviations BMI body mass index CI confidence interval CPP Collaborative Perinatal Project RR risk ratio REFERENCES 1 Lobstein T , Baur L , Uauy R , et al. . Obesity in children and young people: a crisis in public health . Obes Rev . 2004 ; 5 ( 1 suppl ): 4 – 104 . Google Scholar Crossref Search ADS PubMed 2 Ng M , Fleming T , Robinson M , et al. . Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: a systematic analysis for the Global Burden of Disease Study 2013 . Lancet . 2014 ; 384 ( 9945 ): 766 – 781 . Google Scholar Crossref Search ADS PubMed 3 Lloyd LJ , Langley-Evans SC , McMullen S . Childhood obesity and risk of the adult metabolic syndrome: a systematic review . Int J Obes (Lond) . 2012 ; 36 : 1 – 11 . Google Scholar Crossref Search ADS PubMed 4 Tilg H , Adolph TE . Influence of the human intestinal microbiome on obesity and metabolic dysfunction . Curr Opin Pediatr . 2015 ; 27 ( 4 ): 496 – 501 . Google Scholar Crossref Search ADS PubMed 5 Turnbaugh PJ , Ley RE , Mahowald MA , et al. . An obesity-associated gut microbiome with increased capacity for energy harvest . Nature . 2006 ; 444 ( 7122 ): 1027 – 1031 . Google Scholar Crossref Search ADS PubMed 6 Tremaroli V , Bäckhed F . Functional interactions between the gut microbiota and host metabolism . Nature . 2012 ; 489 ( 7415 ): 242 – 249 . Google Scholar Crossref Search ADS PubMed 7 Gaskins HR , Collier CT , Anderson DB . Antibiotics as growth promotants: mode of action . Anim Biotechnol . 2002 ; 13 ( 1 ): 29 – 42 . Google Scholar Crossref Search ADS PubMed 8 Cho I , Yamanishi S , Cox L , et al. . Antibiotics in early life alter the murine colonic microbiome and adiposity . Nature . 2012 ; 488 ( 7413 ): 621 – 626 . Google Scholar Crossref Search ADS PubMed 9 Cox LM , Yamanishi S , Sohn J , et al. . Altering the intestinal microbiota during a critical developmental window has lasting metabolic consequences . Cell . 2014 ; 158 ( 4 ): 705 – 721 . Google Scholar Crossref Search ADS PubMed 10 Nobel YR , Cox LM , Kirigin FF , et al. . Metabolic and metagenomic outcomes from early-life pulsed antibiotic treatment . Nat Commun . 2015 ; 6 : 7486 . Google Scholar Crossref Search ADS PubMed 11 Mansi Y , Abdelaziz N , Ezzeldin Z , et al. . Randomized controlled trial of a high dose of oral erythromycin for the treatment of feeding intolerance in preterm infants . Neonatology . 2011 ; 100 ( 3 ): 290 – 294 . Google Scholar Crossref Search ADS PubMed 12 Guzman MA , Scrimshaw NS , Monroe RJ . Growth and development of central American children. I. Growth responses of rural Guatemalan school children to daily administration of penicillin and aureomycin . Am J Clin Nutr . 1958 ; 6 ( 4 ): 430 – 438 . Google Scholar Crossref Search ADS PubMed 13 Trehan I , Goldbach HS , LaGrone LN , et al. . Antibiotics as part of the management of severe acute malnutrition . N Engl J Med . 2013 ; 368 ( 5 ): 425 – 435 . Google Scholar Crossref Search ADS PubMed 14 Haight TH , Pierce WE . Effect of prolonged antibiotic administration of the weight of healthy young males . J Nutr . 1955 ; 56 ( 1 ): 151 – 161 . Google Scholar Crossref Search ADS PubMed 15 Bailey LC , Forrest CB , Zhang P , et al. . Association of antibiotics in infancy with early childhood obesity . JAMA Pediatr . 2014 ; 168 ( 11 ): 1063 – 1069 . Google Scholar Crossref Search ADS PubMed 16 Azad MB , Bridgman SL , Becker AB , et al. . Infant antibiotic exposure and the development of childhood overweight and central adiposity . Int J Obes (Lond) . 2014 ; 38 ( 10 ): 1290 – 1298 . Google Scholar Crossref Search ADS PubMed 17 Saari A , Virta LJ , Sankilampi U , et al. . Antibiotic exposure in infancy and risk of being overweight in the first 24 months of life . Pediatrics . 2015 ; 135 ( 4 ): 617 – 626 . Google Scholar Crossref Search ADS PubMed 18 Mueller NT , Whyatt R , Hoepner L , et al. . Prenatal exposure to antibiotics, cesarean section and risk of childhood obesity . Int J Obes (Lond) . 2015 ; 39 ( 4 ): 665 – 670 . Google Scholar Crossref Search ADS PubMed 19 Mor A , Antonsen S , Kahlert J , et al. . Prenatal exposure to systemic antibacterials and overweight and obesity in Danish schoolchildren: a prevalence study . Int J Obes (Lond) . 2015 ; 39 ( 10 ): 1450 – 1455 . Google Scholar Crossref Search ADS PubMed 20 Cassidy-Bushrow AE , Burmeister C , Havstad S , et al. . Prenatal antimicrobial use and early-childhood body mass index . Int J Obes (Lond) . 2018 ; 42 : 1 – 7 . Google Scholar Crossref Search ADS PubMed 21 Poulsen MN , Pollak J , Bailey-Davis L , et al. . Associations of prenatal and childhood antibiotic use with child body mass index at age 3 years . Obesity (Silver Spring) . 2017 ; 25 ( 2 ): 438 – 444 . Google Scholar Crossref Search ADS PubMed 22 Mitchell AA , Gilboa SM , Werler MM , et al. . Medication use during pregnancy, with particular focus on prescription drugs: 1976–2008 . Am J Obstet Gynecol . 2011 ; 205 ( 1 ): 51.e1 – 51.e8 . Google Scholar Crossref Search ADS 23 Niswander KR , Gordon M . The Women and Their Pregnancies: the Collaborative Perinatal Study of the National Institute of Neurological Diseases and Stroke . Washington, DC : National Institutes of Health ; 1972 . 24 Klebanoff MA , Keim SA . Maternal serum paraxanthine during pregnancy and offspring body mass index at ages 4 and 7 years . Epidemiology . 2015 ; 26 ( 2 ): 185 – 191 . Google Scholar Crossref Search ADS PubMed 25 Branum AM , Parker JD , Keim SA , et al. . Prepregnancy body mass index and gestational weight gain in relation to child body mass index among siblings . Am J Epidemiol . 2011 ; 174 ( 10 ): 1159 – 1165 . Google Scholar Crossref Search ADS PubMed 26 US Centers for Disease Control and Prevention . A SAS Program for the 2000 CDC Growth Charts (ages 0 to <20 years). http://www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Accessed February 15, 2018. 27 Zou G . A modified Poisson regression approach to prospective studies with binary data . Am J Epidemiol . 2004 ; 159 ( 7 ): 702 – 706 . Google Scholar Crossref Search ADS PubMed 28 White IR , Royston P , Wood AM . Multiple imputation using chained equations: issues and guidance for practice . Stat Med . 2011 ; 30 ( 4 ): 377 – 399 . Google Scholar Crossref Search ADS PubMed 29 Cole TJ , Bellizzi MC , Flegal KM , et al. . Establishing a standard definition for child overweight and obesity worldwide: international survey . BMJ . 2000 ; 320 ( 7244 ): 1240 – 1243 . Google Scholar Crossref Search ADS PubMed 30 Hernán MA , Hernández-Díaz S , Robins JM . A structural approach to selection bias . Epidemiology . 2004 ; 15 ( 5 ): 615 – 625 . Google Scholar Crossref Search ADS PubMed 31 Dominguez-Bello MG , Costello EK , Contreras M , et al. . Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns . Proc Natl Acad Sci USA . 2010 ; 107 ( 26 ): 11971 – 11975 . Google Scholar Crossref Search ADS PubMed 32 Palmer C , Bik EM , DiGiulio DB , et al. . Development of the human infant intestinal microbiota . PLoS Biol . 2007 ; 5 ( 7 ): e177 . Google Scholar Crossref Search ADS PubMed 33 Stinson LF , Payne MS , Keelan JA . Planting the seed: origins, composition, and postnatal health significance of the fetal gastrointestinal microbiota . Crit Rev Microbiol . 2017 ; 43 ( 3 ): 352 – 369 . Google Scholar Crossref Search ADS PubMed 34 Collado MC , Rautava S , Aakko J , et al. . Human gut colonization may be initiated in utero by distinct microbial communities in the placenta and amniotic fluid . Sci Rep . 2016 ; 6 : 23129 . Google Scholar Crossref Search ADS PubMed 35 Jiménez E , Marín ML , Martín R , et al. . Is meconium from healthy newborns actually sterile? Res Microbiol . 2008 ; 159 ( 3 ): 187 – 193 . Google Scholar Crossref Search ADS PubMed 36 Gomez de Agüero M , Ganal-Vonarburg SC , Fuhrer T , et al. . The maternal microbiota drives early postnatal innate immune development . Science . 2016 ; 351 ( 6279 ): 1296 – 1302 . Google Scholar Crossref Search ADS PubMed 37 Vidal AC , Murphy SK , Murtha AP , et al. . Associations between antibiotic exposure during pregnancy, birth weight and aberrant methylation at imprinted genes among offspring . Int J Obes (Lond) . 2013 ; 37 ( 7 ): 907 – 913 . Google Scholar Crossref Search ADS PubMed 38 Malas MA , Aslankoç R , Ungör B , et al. . The development of large intestine during the fetal period . Early Hum Dev . 2004 ; 78 ( 1 ): 1 – 13 . Google Scholar Crossref Search ADS PubMed 39 Korpela K , Salonen A , Virta LJ , et al. . Intestinal microbiome is related to lifetime antibiotic use in Finnish pre-school children . Nat Commun . 2016 ; 7 : 10410 . Google Scholar Crossref Search ADS PubMed 40 Palmsten K , Hernández-Díaz S , Chambers CD , et al. . The most commonly dispensed prescription medications among pregnant women enrolled in the US Medicaid program . Obstet Gynecol . 2015 ; 126 ( 3 ): 465 – 473 . Google Scholar Crossref Search ADS PubMed 41 Briggs GG , Freeman RK , Yaffe SJ . A Reference Guide to Fetal and Neonatal Risk—Drugs in Pregnancy and Lactation . 5th ed . Philadephia, USA : Lippincott Williams & Wilkins ; 1998 . 42 Andrade SE , Gurwitz JH , Davis RL , et al. . Prescription drug use in pregnancy . Am J Obstet Gynecol . 2004 ; 191 ( 2 ): 398 – 407 . Google Scholar Crossref Search ADS PubMed 43 Skinner AC , Skelton JA . Prevalence and trends in obesity and severe obesity among children in the United States, 1999–2012 . JAMA Pediatr . 2014 ; 168 ( 6 ): 561 – 566 . Google Scholar Crossref Search ADS PubMed 44 Myrianthopoulos NC , French KS . An application of the U.S. Bureau of the Census socioeconomic index to a large, diversified patient population . Soc Sci Med . 1968 ; 2 : 283 – 299 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Leslie, Grasset,;Hélène, Jacqmin-Gadda,;Cécile, Proust-Lima,;Karine, Pérès,;Hélène, Amieva,;Jean-François, Dartigues,;Catherine, Helmer,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy118pmid: 29893786

Abstract In line with declining trends in dementia incidence, we compared the cognitive and functional evolution of 2 “generations” of elderly individuals aged 78–88 years, who were included 10 years apart in the French Personnes Agées Quid cohort (n = 612 in 1991–1992 and n = 628 in 2001–2002) and followed-up for 12 years with assessments of cognition and disability. The impact of specific risk factors on this evolution was evaluated. Differences between the generations in baseline levels and decline over time were estimated using a joint model to account for differential attrition. Compared with the first generation, the second generation had higher performances at baseline on 4 cognitive tests (from P < 0.005). Differences in global cognition, verbal fluency, and processing speed, but not in working memory, were mostly explained by improvement in educational level. The second generation also exhibited less cognitive decline in verbal fluency and working memory. Progression of disability was less over the follow-up period for the second generation than for the first. The cognitive state of this elderly population improved, partially due to improvements in educational level. cognitive aging, cohort studies, disability, trends Dementia is a severe, age-related syndrome characterized by progressive cognitive decline leading to loss of autonomy and death. Due to the aging of the general population, the burden from dementia is expected to increase. Current projections predict that by 2050, the number of people living with dementia could increase to 131.5 million (1). Studying secular trends in dementia has thus become of major interest over the past few years. A promising trend toward a decline in the prevalence and incidence of dementia over recent decades has been reported in several studies (2–9). However, these studies all contained methodological weaknesses that may have biased the reported results, such as differential attrition and mortality rates, or changes in the diagnostic criteria for dementia. Moreover, hypotheses explaining a decline in dementia frequency are based on changes in modifiable risk factors across decades (10, 11), such as educational level, vascular factors, or healthier lifestyles. However, in the few studies in which the role of these risk factors in dementia decrease was analyzed, researchers could not fully explain the observed decrease (8, 9). A closer examination of the 2 main components of the disease over time, cognition and disability, could provide a better understanding of the mechanisms involved. Different determinants may affect cognition or function at different periods of life and may differentially affect cognitive domains. Indeed, education may alter cognition early in life, whereas vascular risk factors may influence cognition only during midlife or after 65 years of age. Moreover, a major advantage of studying secular trends in cognition instead of secular trends in dementia risk is that trends in cognition cannot be biased by changes in diagnostic criteria. Secular trends in level and decline of cognitive performances have been studied. If the level of cognitive abilities has been consistently shown to improve over generations, results regarding the rate of cognitive decline are more conflicted. Improvements in the rate of cognitive decline between generations have been reported in some studies (12–14), whereas more profound decline in later generations (15, 16) or similar rates of decline between generations (17, 18) have been reported in others. These inconsistencies may be due to methodological issues that arise when studying secular trends, such as differential representativeness between 2 birth cohorts, differential attrition and mortality rates, and insufficient follow-up duration. For example, death rates have regularly evolved between generations (19). Moreover, because dropout is closely related to cognitive performance, differential dropout rates may bias comparisons. To date, to our knowledge, studies published on cognition have not fully considered these issues. In this study, we used the following approach to analyze secular trends of cognition and disability: the same cohort was studied to limit differences in participation rates, and joint modeling was applied to account for attrition over the follow-up period. We compared 2 generations of elderly people from the French population-based Personnes Agées Quid (PAQUID) Study in the same age group, 10 years apart, who were followed for 12 years. In addition, we evaluated the extent to which trends of cognition and disability were influenced by the evolution between generations attributed to determinants such as education or vascular factors. METHODS Study population This study was based on the PAQUID cohort, a prospective, population-based cohort consisting of a representative sample of 3,777 participants in the departments of Gironde and Dordogne (southwest France), who were randomly chosen from the electoral rolls in 1988–1989. There were 3 inclusion criteria: age at least 65 years by December 31, 1987; living at home at the time of the initial data collection phase; and the provision of informed consent for study participation. Participants were followed for 25 years with follow-up at 3, 5, 8, 10, 13, 15, 17, 20, 22, and 25 years after the baseline evaluation. Of the 5,554 persons selected, 3,777 (68%) agreed to participate in the study. Full details of the PAQUID study have been described elsewhere (20). A standardized questionnaire assessing sociodemographic, medical, cognitive, and functional data was administered at each participant’s home by trained neuropsychologists during face-to-face interviews, at baseline, and at each follow-up session. An ethical review committee approved the PAQUID study. Temporal trends in cognition were studied 10 years apart. To avoid the first-passing effect inherent to cognitive tests (21), we did not include the baseline evaluation. Thus, 2 generations were selected and compared as follows (Figure 1): The first generation (G1) included subjects born between 1903 and 1912 (78–88 years old) and living at home in 1991–1992 (at the 3-year follow-up). Of the 1,286 individuals who were eligible at time 0, 182 died, 350 were lost track of, and 44 had moved to an institution, leaving 710 assessed at the 3-year follow-up. The second generation (G2) included subjects born between 1913 and 1922 (78–88 years old) and living at home in 2001–2002 (13-year follow-up). Of the 1,898 individuals eligible at time 0, 653 died, 474 were lost track of, and 56 had moved to an institution, leaving 715 participants assessed at the 13-year follow-up. Figure 1. View largeDownload slide Repartition of subjects in the first and second generations (G1 and G2, respectively) including percentages of subjects’ participation and death (cumulative) at each follow-up (FU) time (T), Personnes Agées Quid Study, Bordeaux, France, 1988–2013. Figure 1. View largeDownload slide Repartition of subjects in the first and second generations (G1 and G2, respectively) including percentages of subjects’ participation and death (cumulative) at each follow-up (FU) time (T), Personnes Agées Quid Study, Bordeaux, France, 1988–2013. For each of these 2 generations, the 12 subsequent years of follow-up were analyzed. The third-year follow-up for G1 and the 13-year follow-up for G2 will henceforth be referred to as the baseline for G1 and G2, respectively. For each generation, participants with prevalent dementia at baseline (for G1, n = 98, 13.8%; for G2: n = 89, 12.4%) were excluded to investigate the global evolution in an initially dementia-free population. We previously demonstrated that the clinical diagnosis of dementia has changed over time (8); therefore, these prevalent cases were excluded on the basis of an algorithmic diagnosis (Mini-Mental State Examination (MMSE) score <24 and disability for at least 2 of 4 activities of the 4 Instrumental Activities of Daily Living (4-IADL) scale: telephone use, transportation, medication, and finances). The final study population thus consisted of 1,238 subjects: 612 in G1 and 626 in G2. Adjustment factors The demographic factors evaluated included age, sex, educational level (divided into 5 categories: no diploma, primary school, short secondary school, long secondary school, and validated higher education), occupation (7 classes: farm workers and managers, domestic service employees, blue-collar workers, craftsmen and shopkeepers, other employees, intellectual occupations, and housewives), and living status (alone or not). Medication use was recorded at baseline and antihypertensive, antidiabetic, and lipid-lowering drugs were controlled for in the present study and used as proxies for cardiovascular diseases, diabetes, and hypercholesterolemia, respectively. Cognition and function assessment At baseline and at each follow-up, participants underwent a complete cognitive and functional evaluation. Cognition was evaluated using 4 cognitive tests: 1) MMSE (22), assessing global cognitive functioning, with scores ranging from 0 to 30; 2) the 15-second version of the Isaacs Set Test (IST) (23), assessing semantic verbal fluency; 3) the Benton Visual Retention Test (BVRT) (24) measuring visual working memory, with scores ranging from 0 to 15; and 4) the digit symbol substitution test (DSST) (25), evaluating executive function and processing speed; this test was conducted at each follow-up except the 3-year follow-up. Higher cognitive scores indicated better cognitive performances. Functional abilities were evaluated using 4-IADL scale: the ability to use the telephone, transportation, responsibility for medications, and the ability to manage finances. These 4 activities assess cognitive-specific functions and are most highly correlated to cognitive impairment (26) and to predict incident dementia (27). For each activity, 3–5 different levels of disability were assessed. In this study, participants were considered disabled for the activity at the less-severe level. Disability was evaluated as a binary outcome, indicating whether the participant was disabled for more than 1 activity out of the 4. Statistical analyses For each generation, cognitive abilities and disability were analyzed over a 12-year follow-up (from 1991–1992 to 2003–2004 for G1 and from 2001–2002 to 2013–2014 for G2) (Figure 1). χ2 and Student t tests were used to compare the 2 generations in terms of sociodemographic characteristics; MMSE, BVRT, IST, and DSST scores; 4-IADL disability at baseline; and the intake of 3 specific drug categories (i.e., antihypertensive, antidiabetic, and lipid-lowering treatments). Follow-up and attrition data were recorded for each generation (Figure 1). To account for possible differential attrition between the 2 generations, cognitive and functional trajectories were analyzed in 2 separate joint models combining a longitudinal submodel for repeated measures of either cognitive scores or IADL disability and a survival submodel for attrition (28). Analyzing raw cognitive scores can result in biased estimations, mostly due to ceiling and floor effects and curvilinearity (i.e., unequal interval scaling) (29). Indeed, different scores on the same test may not have the same meaning regarding true cognitive function at different points on the scale. To correct for these issues, each longitudinal marker was first optimally transformed using a spline transformation in a separate model (30). The transformed scores were then entered into the linear mixed submodel of the joint modeling approach. The baseline distribution of cognitive scores after transformation are presented in Web Figure 1 (available at https://academic.oup.com/aje). For the binary disability indicator built from the 4-IADLs, a logistic mixed submodel replaced the linear mixed submodel in the joint model. Attrition was defined as either dropout or death, whichever occurred first. The proportional hazards model for the risk of attrition was adjusted for generation, sex, age at baseline, and educational level. The linear mixed model assumed a linear trajectory with time with correlated individual random intercept and slope. Time was defined as the number of years since baseline (the 3-year and 13-year follow-ups for G1 and G2, respectively). A quadratic time trend was tested but not retained in the model, because it was systematically nonsignificant. The interaction between generation and sex was tested and data were analyzed globally, because this interaction was nonsignificant. For each outcome, we thus estimated 3 models. The first was systematically adjusted for generation (G2 vs. G1), age, and sex, and it included an interaction between generation and time (model 1). The simple effect of generation quantified the difference in the baseline scores at the 3-year and 13-year follow-ups for G1 and G2, respectively, whereas the interaction with time quantified the generation impact on the score change over the follow-up period. The models were then additionally adjusted for educational level and occupation (model 2) and for vascular factors (for which antihypertensive, antidiabetic, and lipid-lowering drugs were proxies) and living alone (model 3). Interactions between time and adjustment factors were tested and added to the models when they were significant. It is important to mention that the effect sizes could not be interpreted according to the natural scale of the raw scores, and the normalized scales are not z-scores, although they were standardized as z-scores. The goodness of fit of each model was assessed using residual plots. In the sensitivity analysis, we defined attrition as death only (dropout was not considered). Statistical analyses were performed with SAS statistical software, version 9.3 (SAS Institute, Inc., Cary, North Carolina) and R packages lcmm, version 1.7.5 (31), JM version 1.4–5 (32), and JMbayes, version 0.8–70 (R Foundation for Statistical Computing, Vienna, Austria) (33). Web Appendix 1 provides the R code for analysis replication. RESULTS Study sample description The sex distribution did not differ between the 2 generations (Table 1). At baseline, G2 was slightly younger, had a higher educational level, had more intellectual occupations, and took more antihypertensive and lipid-lowering drugs than G1. The baseline scores on MMSE, BVRT, and DSST (for DSST, mean scores at the 5-year and 15-year follow-ups were assessed because scores were not available at the 3-year follow-up) were significantly higher for G2 than for G1. However, the proportion of disabled subjects at baseline did not differ between the 2 generations. The 12-year mortality rate was also lower in G2 than G1 (66.9% vs. 80.0%, respectively) (Figure 1). Table 1. Baselinea Characteristics of the 2 Generations From the Personnes Agées Quid Study, Bordeaux, France, 1988–2013 Characteristic First Generation (n = 612) Second Generation (n = 626) P Valueb No. % Mean (SD) No. % Mean (SD) Women 361 59.0 368 58.8 0.94 Mean age, years 82.4 (2.4) 81.95 (2.7) 0.0007 Educational level <0.0001  No diploma 192 31.4 114 18.2  Primary school 299 48.9 305 48.7  Short secondary school 65 10.6 97 15.5  Long secondary school 33 5.4 55 8.8  Validated higher education 23 3.7 55 8.8 Occupation <0.0001  Farmworkers and farm managers 92 15.0 75 12.0  Domestic service employees 49 8.0 49 7.8  Blue-collar workers 99 16.2 78 12.5  Craftsmen and shopkeepers 113 18.5 77 12.3  Other employees 144 23.5 191 30.5  Intellectual occupations 56 9.2 97 15.5  Housewives/inactive 59 9.6 59 9.4 Living alone 296 48.4 284 45.4 0.29 Antihypertensive drug use 383 62.6 440 70.3 0.004 Antidiabetics drug use 50 8.2 35 5.6 0.07 Lipid-lowering drug use 70 11.4 154 24.6 <0.0001 Mean baseline MMSE 26.6 (2.4) 27.1 (2.2) <0.0001 Mean baseline IST15 26.9 (5.4) 27.5 (6.1) 0.08 Mean baseline BVRT 10.3 (2.4) 11.24 (2.2) <0.0001 Mean baseline + 2 years DSST 24.0 (9.2) 27.0 (10.0) <0.0001 Baseline 4-IADL disability 89 14.6 83 13.3 0.61 Characteristic First Generation (n = 612) Second Generation (n = 626) P Valueb No. % Mean (SD) No. % Mean (SD) Women 361 59.0 368 58.8 0.94 Mean age, years 82.4 (2.4) 81.95 (2.7) 0.0007 Educational level <0.0001  No diploma 192 31.4 114 18.2  Primary school 299 48.9 305 48.7  Short secondary school 65 10.6 97 15.5  Long secondary school 33 5.4 55 8.8  Validated higher education 23 3.7 55 8.8 Occupation <0.0001  Farmworkers and farm managers 92 15.0 75 12.0  Domestic service employees 49 8.0 49 7.8  Blue-collar workers 99 16.2 78 12.5  Craftsmen and shopkeepers 113 18.5 77 12.3  Other employees 144 23.5 191 30.5  Intellectual occupations 56 9.2 97 15.5  Housewives/inactive 59 9.6 59 9.4 Living alone 296 48.4 284 45.4 0.29 Antihypertensive drug use 383 62.6 440 70.3 0.004 Antidiabetics drug use 50 8.2 35 5.6 0.07 Lipid-lowering drug use 70 11.4 154 24.6 <0.0001 Mean baseline MMSE 26.6 (2.4) 27.1 (2.2) <0.0001 Mean baseline IST15 26.9 (5.4) 27.5 (6.1) 0.08 Mean baseline BVRT 10.3 (2.4) 11.24 (2.2) <0.0001 Mean baseline + 2 years DSST 24.0 (9.2) 27.0 (10.0) <0.0001 Baseline 4-IADL disability 89 14.6 83 13.3 0.61 Abbreviations: 4-IADL, 4 Instrumental Activities of Daily Living; BVRT, Benton Visual Retention Test; IST15, Isaacs Set Test at 15 seconds; MMSE, Mini-Mental State Examination; SD, standard deviation. a Baseline time corresponds to the 3-year follow-up for the first generation and the 13-year follow-up for the second generation in the study. Because DSST scores at the 3-year follow-up were not available, we compared the DSST scores at 5 and 15 years. b Differences in percentages and mean values were determined using a χ2 test and Student t test, respectively. Table 1. Baselinea Characteristics of the 2 Generations From the Personnes Agées Quid Study, Bordeaux, France, 1988–2013 Characteristic First Generation (n = 612) Second Generation (n = 626) P Valueb No. % Mean (SD) No. % Mean (SD) Women 361 59.0 368 58.8 0.94 Mean age, years 82.4 (2.4) 81.95 (2.7) 0.0007 Educational level <0.0001  No diploma 192 31.4 114 18.2  Primary school 299 48.9 305 48.7  Short secondary school 65 10.6 97 15.5  Long secondary school 33 5.4 55 8.8  Validated higher education 23 3.7 55 8.8 Occupation <0.0001  Farmworkers and farm managers 92 15.0 75 12.0  Domestic service employees 49 8.0 49 7.8  Blue-collar workers 99 16.2 78 12.5  Craftsmen and shopkeepers 113 18.5 77 12.3  Other employees 144 23.5 191 30.5  Intellectual occupations 56 9.2 97 15.5  Housewives/inactive 59 9.6 59 9.4 Living alone 296 48.4 284 45.4 0.29 Antihypertensive drug use 383 62.6 440 70.3 0.004 Antidiabetics drug use 50 8.2 35 5.6 0.07 Lipid-lowering drug use 70 11.4 154 24.6 <0.0001 Mean baseline MMSE 26.6 (2.4) 27.1 (2.2) <0.0001 Mean baseline IST15 26.9 (5.4) 27.5 (6.1) 0.08 Mean baseline BVRT 10.3 (2.4) 11.24 (2.2) <0.0001 Mean baseline + 2 years DSST 24.0 (9.2) 27.0 (10.0) <0.0001 Baseline 4-IADL disability 89 14.6 83 13.3 0.61 Characteristic First Generation (n = 612) Second Generation (n = 626) P Valueb No. % Mean (SD) No. % Mean (SD) Women 361 59.0 368 58.8 0.94 Mean age, years 82.4 (2.4) 81.95 (2.7) 0.0007 Educational level <0.0001  No diploma 192 31.4 114 18.2  Primary school 299 48.9 305 48.7  Short secondary school 65 10.6 97 15.5  Long secondary school 33 5.4 55 8.8  Validated higher education 23 3.7 55 8.8 Occupation <0.0001  Farmworkers and farm managers 92 15.0 75 12.0  Domestic service employees 49 8.0 49 7.8  Blue-collar workers 99 16.2 78 12.5  Craftsmen and shopkeepers 113 18.5 77 12.3  Other employees 144 23.5 191 30.5  Intellectual occupations 56 9.2 97 15.5  Housewives/inactive 59 9.6 59 9.4 Living alone 296 48.4 284 45.4 0.29 Antihypertensive drug use 383 62.6 440 70.3 0.004 Antidiabetics drug use 50 8.2 35 5.6 0.07 Lipid-lowering drug use 70 11.4 154 24.6 <0.0001 Mean baseline MMSE 26.6 (2.4) 27.1 (2.2) <0.0001 Mean baseline IST15 26.9 (5.4) 27.5 (6.1) 0.08 Mean baseline BVRT 10.3 (2.4) 11.24 (2.2) <0.0001 Mean baseline + 2 years DSST 24.0 (9.2) 27.0 (10.0) <0.0001 Baseline 4-IADL disability 89 14.6 83 13.3 0.61 Abbreviations: 4-IADL, 4 Instrumental Activities of Daily Living; BVRT, Benton Visual Retention Test; IST15, Isaacs Set Test at 15 seconds; MMSE, Mini-Mental State Examination; SD, standard deviation. a Baseline time corresponds to the 3-year follow-up for the first generation and the 13-year follow-up for the second generation in the study. Because DSST scores at the 3-year follow-up were not available, we compared the DSST scores at 5 and 15 years. b Differences in percentages and mean values were determined using a χ2 test and Student t test, respectively. Cognitive and functional evolution between generations Results of the joint model analysis for each psychometric test and disability are presented in Table 2. Estimates are given in the transformed scales for the cognitive tests. Figure 2 displays the unadjusted predicted mean score trajectory, according to generation, on the transformed cognitive score scale and on the logit scale for disability. Table 2. Parameter Estimates of the Linear Mixed Submodel for Transformed Cognitive Scores and the Logistic Mixed Submodel for Disability From the Joint Model Analysis, Personnes Agées Quid Study, Bordeaux, France, 1988–2013 Cognitive Test and Disability Model 1a Model 2b Model 3c βd OR 95% CIe βd OR 95% CIe βd OR 95% CIe MMSEf  Generation (G2 vs. G1) 0.29 0.13, 0.45 0.08 −0.07, 0.23 0.05 −0.10, 0.21  Timeg −0.14 −0.19, −0.09 −0.14 −0.19, −0.09 −0.14 −0.19, −0.09  G2 × time 0.01 −0.02, 0.04 0.01 −0.02, 0.04 0.01 −0.02, 0.04 IST15h  Generation (G2 vs. G1) 0.27 0.08, 0.46 0.05 −0.13, 0.24 0.02 −0.16, 0.21  Timeg −0.23 −0.28, −0.18 −0.23 −0.28, −0.18 −0.23 −0.28, −0.18  G2 × time 0.03 0.0008, 0.06 0.03 0.0009, 0.06 0.03 0.001, 0.06 BVRTh  Generation (G2 vs. G1) 0.51 0.36, 0.66 0.32 0.18, 0.46 0.32 0.18, 0.46  Timeg −0.07 −0.11, −0.04 −0.07 −0.11, −0.04 −0.07 −0.11, −0.04  G2 × time 0.03 0.001, 0.05 0.03 0.002, 0.05 0.03 0.002, 0.05 DSSTi  Generation (G2 vs. G1) 0.56 0.21, 0.91 0.25 −0.04, 0.57 0.30 −0.01, 0.60  Timeg −0.27 −0.31, −0.22 −0.30 −0.37, −0.23 −0.31 −0.38, −0.24  G2 × time 0.002 −0.04, 0.04 0.006 −0.03, 0.04 0.005 −0.03, 0.04 4-IADL disabilityj  Generation (G2 vs. G1) 0.97 0.76, 1.22 1.14 0.85, 1.54 1.17 0.95, 1.45  Timeg 1.26 1.22, 1.28 1.27 1.25, 1.28 1.23 1.18, 1.27  G2 × time 0.92 0.90, 0.93 0.93 0.92, 0.94 0.93 0.91, 0.95 Cognitive Test and Disability Model 1a Model 2b Model 3c βd OR 95% CIe βd OR 95% CIe βd OR 95% CIe MMSEf  Generation (G2 vs. G1) 0.29 0.13, 0.45 0.08 −0.07, 0.23 0.05 −0.10, 0.21  Timeg −0.14 −0.19, −0.09 −0.14 −0.19, −0.09 −0.14 −0.19, −0.09  G2 × time 0.01 −0.02, 0.04 0.01 −0.02, 0.04 0.01 −0.02, 0.04 IST15h  Generation (G2 vs. G1) 0.27 0.08, 0.46 0.05 −0.13, 0.24 0.02 −0.16, 0.21  Timeg −0.23 −0.28, −0.18 −0.23 −0.28, −0.18 −0.23 −0.28, −0.18  G2 × time 0.03 0.0008, 0.06 0.03 0.0009, 0.06 0.03 0.001, 0.06 BVRTh  Generation (G2 vs. G1) 0.51 0.36, 0.66 0.32 0.18, 0.46 0.32 0.18, 0.46  Timeg −0.07 −0.11, −0.04 −0.07 −0.11, −0.04 −0.07 −0.11, −0.04  G2 × time 0.03 0.001, 0.05 0.03 0.002, 0.05 0.03 0.002, 0.05 DSSTi  Generation (G2 vs. G1) 0.56 0.21, 0.91 0.25 −0.04, 0.57 0.30 −0.01, 0.60  Timeg −0.27 −0.31, −0.22 −0.30 −0.37, −0.23 −0.31 −0.38, −0.24  G2 × time 0.002 −0.04, 0.04 0.006 −0.03, 0.04 0.005 −0.03, 0.04 4-IADL disabilityj  Generation (G2 vs. G1) 0.97 0.76, 1.22 1.14 0.85, 1.54 1.17 0.95, 1.45  Timeg 1.26 1.22, 1.28 1.27 1.25, 1.28 1.23 1.18, 1.27  G2 × time 0.92 0.90, 0.93 0.93 0.92, 0.94 0.93 0.91, 0.95 Abbreviations: 4-IADL, 4 Instrumental Activities of Daily Living; BVRT, Benton Visual Retention Test; CI, confidence interval; DSST, Digit Symbol Substitution Test; G1, first generation; G2, second generation; IST15, Isaacs Set Test at 15 seconds; MMSE, Mini-Mental State Examination; OR, odds ratio. a Adjusted for age and sex. b Adjusted for age, sex, educational level, and occupation. c Adjusted for age, sex, educational level, occupation, living alone, and intake of antihypertensive, antidiabetic, and lipid-lowering drugs. d Parameters are the results of the linear mixed submodels of the joint model analysis. Due to score transformation, parameters cannot be interpreted according to the scores’ natural scales. e For linear mixed regression, 95% confidence intervals are reported; for logistic mixed regression, 95% credible intervals are reported. f Each model was adjusted for interactions between time and age, and between time and sex. g Values are for men 75 years old in the interactions with time. h Model was adjusted for interaction between time and age. i Model 1 was adjusted for interaction between time and age and between time and sex; models 2 and 3 were additionally adjusted for time and education. j Model 1 was adjusted for interactions between time and age and between time and sex; model 2 was additionally adjusted for time and education; and model 3 was additionally adjusted for time and antidiabetic drug treatment. Table 2. Parameter Estimates of the Linear Mixed Submodel for Transformed Cognitive Scores and the Logistic Mixed Submodel for Disability From the Joint Model Analysis, Personnes Agées Quid Study, Bordeaux, France, 1988–2013 Cognitive Test and Disability Model 1a Model 2b Model 3c βd OR 95% CIe βd OR 95% CIe βd OR 95% CIe MMSEf  Generation (G2 vs. G1) 0.29 0.13, 0.45 0.08 −0.07, 0.23 0.05 −0.10, 0.21  Timeg −0.14 −0.19, −0.09 −0.14 −0.19, −0.09 −0.14 −0.19, −0.09  G2 × time 0.01 −0.02, 0.04 0.01 −0.02, 0.04 0.01 −0.02, 0.04 IST15h  Generation (G2 vs. G1) 0.27 0.08, 0.46 0.05 −0.13, 0.24 0.02 −0.16, 0.21  Timeg −0.23 −0.28, −0.18 −0.23 −0.28, −0.18 −0.23 −0.28, −0.18  G2 × time 0.03 0.0008, 0.06 0.03 0.0009, 0.06 0.03 0.001, 0.06 BVRTh  Generation (G2 vs. G1) 0.51 0.36, 0.66 0.32 0.18, 0.46 0.32 0.18, 0.46  Timeg −0.07 −0.11, −0.04 −0.07 −0.11, −0.04 −0.07 −0.11, −0.04  G2 × time 0.03 0.001, 0.05 0.03 0.002, 0.05 0.03 0.002, 0.05 DSSTi  Generation (G2 vs. G1) 0.56 0.21, 0.91 0.25 −0.04, 0.57 0.30 −0.01, 0.60  Timeg −0.27 −0.31, −0.22 −0.30 −0.37, −0.23 −0.31 −0.38, −0.24  G2 × time 0.002 −0.04, 0.04 0.006 −0.03, 0.04 0.005 −0.03, 0.04 4-IADL disabilityj  Generation (G2 vs. G1) 0.97 0.76, 1.22 1.14 0.85, 1.54 1.17 0.95, 1.45  Timeg 1.26 1.22, 1.28 1.27 1.25, 1.28 1.23 1.18, 1.27  G2 × time 0.92 0.90, 0.93 0.93 0.92, 0.94 0.93 0.91, 0.95 Cognitive Test and Disability Model 1a Model 2b Model 3c βd OR 95% CIe βd OR 95% CIe βd OR 95% CIe MMSEf  Generation (G2 vs. G1) 0.29 0.13, 0.45 0.08 −0.07, 0.23 0.05 −0.10, 0.21  Timeg −0.14 −0.19, −0.09 −0.14 −0.19, −0.09 −0.14 −0.19, −0.09  G2 × time 0.01 −0.02, 0.04 0.01 −0.02, 0.04 0.01 −0.02, 0.04 IST15h  Generation (G2 vs. G1) 0.27 0.08, 0.46 0.05 −0.13, 0.24 0.02 −0.16, 0.21  Timeg −0.23 −0.28, −0.18 −0.23 −0.28, −0.18 −0.23 −0.28, −0.18  G2 × time 0.03 0.0008, 0.06 0.03 0.0009, 0.06 0.03 0.001, 0.06 BVRTh  Generation (G2 vs. G1) 0.51 0.36, 0.66 0.32 0.18, 0.46 0.32 0.18, 0.46  Timeg −0.07 −0.11, −0.04 −0.07 −0.11, −0.04 −0.07 −0.11, −0.04  G2 × time 0.03 0.001, 0.05 0.03 0.002, 0.05 0.03 0.002, 0.05 DSSTi  Generation (G2 vs. G1) 0.56 0.21, 0.91 0.25 −0.04, 0.57 0.30 −0.01, 0.60  Timeg −0.27 −0.31, −0.22 −0.30 −0.37, −0.23 −0.31 −0.38, −0.24  G2 × time 0.002 −0.04, 0.04 0.006 −0.03, 0.04 0.005 −0.03, 0.04 4-IADL disabilityj  Generation (G2 vs. G1) 0.97 0.76, 1.22 1.14 0.85, 1.54 1.17 0.95, 1.45  Timeg 1.26 1.22, 1.28 1.27 1.25, 1.28 1.23 1.18, 1.27  G2 × time 0.92 0.90, 0.93 0.93 0.92, 0.94 0.93 0.91, 0.95 Abbreviations: 4-IADL, 4 Instrumental Activities of Daily Living; BVRT, Benton Visual Retention Test; CI, confidence interval; DSST, Digit Symbol Substitution Test; G1, first generation; G2, second generation; IST15, Isaacs Set Test at 15 seconds; MMSE, Mini-Mental State Examination; OR, odds ratio. a Adjusted for age and sex. b Adjusted for age, sex, educational level, and occupation. c Adjusted for age, sex, educational level, occupation, living alone, and intake of antihypertensive, antidiabetic, and lipid-lowering drugs. d Parameters are the results of the linear mixed submodels of the joint model analysis. Due to score transformation, parameters cannot be interpreted according to the scores’ natural scales. e For linear mixed regression, 95% confidence intervals are reported; for logistic mixed regression, 95% credible intervals are reported. f Each model was adjusted for interactions between time and age, and between time and sex. g Values are for men 75 years old in the interactions with time. h Model was adjusted for interaction between time and age. i Model 1 was adjusted for interaction between time and age and between time and sex; models 2 and 3 were additionally adjusted for time and education. j Model 1 was adjusted for interactions between time and age and between time and sex; model 2 was additionally adjusted for time and education; and model 3 was additionally adjusted for time and antidiabetic drug treatment. Figure 2. View largeDownload slide Unadjusted predicted mean trajectories for the A) Mini-Mental State Examination (MMSE), B) Isaacs Set Test (IST), C) Benton Visual Retention Test (BVRT), and D) Digit Symbol Substitution Test (DSST) in their transformed scales and for E) 4 Instrumental Activities of Daily Living (4-IADL) disability on the logit scale, Personnes Agées Quid Study, Bordeaux, France, 1988–2013. Shown are data for the 2 generations with 95% confidence intervals. Figure 2. View largeDownload slide Unadjusted predicted mean trajectories for the A) Mini-Mental State Examination (MMSE), B) Isaacs Set Test (IST), C) Benton Visual Retention Test (BVRT), and D) Digit Symbol Substitution Test (DSST) in their transformed scales and for E) 4 Instrumental Activities of Daily Living (4-IADL) disability on the logit scale, Personnes Agées Quid Study, Bordeaux, France, 1988–2013. Shown are data for the 2 generations with 95% confidence intervals. For baseline cohort effects, model 1 indicated that G2 had higher scores on the 4 cognitive tests than G1. After adjusting for education and occupation (model 2), the impact of generation on the mean transformed scores at baseline was attenuated by 72.4% for MMSE, 81.5% for IST, and 55.4% for DSST, and was no longer significant. Additional adjustment in model 3 did not further modify the improvement in scores at baseline. In contrast, the association between generation and baseline performance remained significant for BVRT, although it was slightly attenuated (by 37.3% in model 2) after additional adjustments for education and occupation (model 2) and vascular factors (model 3). For cognitive decline over time, the trajectories did not differ according to generation for MMSE and DSST (β = 0.01, 95% confidence interval (CI): −0.02, 0.04; and β = 0.002, 95% CI: −0.04, 0.04, respectively), whereas G2 had significantly lower decline rates than G1 on IST and BVRT scores (β = 0.03, 95% CI: 0.0008, 0.06; and β = 0.03, 95% CI: 0.001, 0.05, respectively). For IST and BVRT, the lower declines for G2 were not attenuated when adjusting for education, occupation, or vascular factors. For the 4-IADL binary indicator, the baseline probability of disability did not differ significantly between the generations (odds ratio (OR) = 0.97, 95% CI: 0.76, 1.22). This increase in disability over time was significantly slower for G2 than for G1 (OR = 0.92, 95% CI: 0.90, 0.93, in model 1), and the difference remained significant and was not attenuated after further adjustments in model 2 and 3 (OR = 0.93, 95% CI: 0.92, 0.94; and OR = 0.93, 95% CI: 0.91, 0.95, respectively). Results of the sensitivity analyses, considering only death and assuming that the dropout mechanism was random, did not differ from the main analysis (see Web Table 1). DISCUSSION Our main findings in this study were improved performances in global cognitive functioning, verbal fluency, working memory, and processing speed between 2 generations of subjects aged 78–88 years and evaluated 10 years apart, as well as slower declines in verbal fluency and working memory over the follow-up period. For global cognition, verbal fluency, and processing speed, the improvement at baseline was mostly explained by increases in educational and occupational levels, although not for visual working memory. G2’s slower cognitive decline in verbal fluency and working memory was not explained by education or occupation. Despite our hypothesis, antihypertensive, antidiabetes, and lipid-lowering treatments only slightly explained the relationship between generation and cognition. No improvement was found between generations in functional capacities in activities of daily living at baseline. However, G2 exhibited less pronounced progression toward disability over time, which was not explained by adjustment factors. Supporting the decrease in dementia occurrence, cognitive performance showed a global improvement over the 10 years. Educational level was highly improved between G1 and G2 (31.4% of G1 had no diploma vs. 18.2% of G2) and this explained a large part of the differences in baseline score between the 2 generations, although this was not true for all cognitive domains. Education is beneficial early in life, and this benefit may last until old age. Educational level was highly associated with the mean cognitive score at baseline but was not associated with the decline over time (except for DSST) (data not shown). This finding agrees with that of a review in which it was reported that education was highly associated with cognitive performance but did not moderate age-associated cognitive decline (34). These findings are consistent with the Flynn effect, described as an improvement in intellectual quotient with improvements in education (35, 36). Depending on the cognitive domains implicated in each test, it is reasonable that some tests are more influenced by educational level than others. However, although education influences BVRT scores (37), it only partially explained the impact of generation found in our results. This suggests that other factors in addition to education may contribute to the improvement in cognitive level over time. The improvement in cognitive levels between generations is in agreement with results from previous studies (12–18, 38–41). However, as mentioned, the results concerning cognitive decline are more conflicted. Studying secular trends in cognitive decline is challenging, and in several previous studies, researchers faced methodological limitations. Short follow-up durations were used in some of these studies (16) and/or there were large intervals between cognitive assessments (12, 15, 17). In several studies, differential selection was encountered when comparing generations from different populations (15, 17). An important strength of our study is that the 2 generations were from the same population-based prospective cohort, with up to 6 assessments of cognition and function every 2 to 3 years over 12 years of follow-up for each generation. Moreover, follow-up questionnaires were administered by trained neuropsychologists and managed by the same team over time. However, comparing 2 generations within the same cohort required us to restrict the age range of our study sample to 78–88 years at baseline, and results for younger subjects might be different. Investigating the impact of generation in younger populations may help elucidate the dynamic of cognitive improvement over the life course. However, elderly people experience more cognitive decline, and it is probably easier to examine differences between generations at more advanced ages than in younger populations. Another common limitation of previous studies was their different attrition rates during follow-up, which may have biased the results and were often not taken into account using appropriate statistical methods (12, 14, 15, 17, 18). Indeed, death rates evolved over time, and nondeath-related dropout may have also differed between compared samples. Lower attrition rates in younger generations could lead to an underestimation of the generation effect. A major strength of this work is that it used appropriate statistical models to account for attrition and avoid biases due to differential dropout or death rates between the 2 generations. Moreover, interval scaling issues such as ceiling and floor effects of most cognitive scores were not addressed in most previous studies of cognitive differences between generations. In our work, normality and interval scaling problems were taken into account, which avoided the substantial bias highlighted when studying the decline over time in cognitive scores with an asymmetric distribution (29). Finally, because it is typical for cognitive aging data, practice effect can also influence cognitive trajectories and bias cohort effects in studies in which it is not accounted for (12, 15, 17). Due to the study design of our sample, participants in G2 had undergone more cognitive testing (because they had more follow-up visits). Thus, we cannot fully exclude practice effect, which may have led to higher cognitive scores for G2. However, by deleting the first testing date, we avoided bias due to first-passing effect, which has been shown to be much larger than residual practice effect after the second visit (42, 43). No improvement in disability was found between generations. However, the trajectory of disability over time was better in G2, a result that was not explained by the adjustment factors. In contrast to this finding, a significant improvement in the level of functional abilities between cohorts has been reported in some studies (41, 44–46). In contrast, Jagger et al. (47) found a nonsignificant trend toward an increase in disability between 5-year–interval cohorts, and Steiber (38) reported a decrease in physical health scores (lower physical performance) in subjects aged 50–90 years who were born 6 years apart. These differences between studies could be caused by the use of different methods for assessing functional status. Disability in IADL may have decreased, whereas disability in basic activities of daily living may have increased due to the longer survival of frailer people. In the present study, in which our objective was to evaluate the evolution of disability in relation to cognition, we focused on only the 4 instrumental activities with high cognitive demand. We did not find any improvement between the generations at baseline. However, our analysis of other instrumental activities (i.e., housekeeping, laundry, shopping, and cooking) revealed an improvement in cooking and shopping for women (data not shown). Another possible reason for the lack of improvement in IADL at baseline between the generations in our study could be because the abilities required to perform some activities may have evolved over time, with activities such as driving, using the telephone, or managing a budget requiring greater cognitive abilities than they did previously. In general, members of younger generations have higher cognitive performance; thus, because of lower cognitive decline rates for some cognitive domains, those in younger generations may reach the clinical threshold for dementia later than older generations and experience delayed onset of dementia. These higher cognitive performances in some domains are partly explained by the higher level of education in younger generations. This is in line with results we have reported regarding the evolution of dementia incidence, in which we found that a decreased risk of dementia was partially explained by education (8). However, more complicated processes seem to be involved in this decrease in dementia and improvement in cognition, and factors other than educational level have been implicated. In most previous studies evaluating cohort effects on cognition, education was the determinant most frequently taken into account. In the present study, we analyzed education and occupation, as well as vascular factors. However, without objective measures of vascular factors, drug use was used as a proxy, which could have underestimated the role of the actual vascular factors. Even if they are difficult to obtain, long-term or lifetime records of risk factors would be valuable for understanding improvement in cognition and disability. Although we could not provide evidence for these factors, individual factors such as vascular factors, behavioral habits, and environmental factors may still be implicated. Moreover, progression toward dementia is a long process, beginning several years before the onset of clinical dementia (48); understanding this process would require a lifelong evaluation of cognitive performance. Our analyses indicate that initial cohort differences in cognitive performance are maintained throughout aging and are exacerbated by advancing age for only some cognitive domains. Thus, improvements in basic cognitive states are associated with smaller declines in verbal fluency and working memory in old age, which may explain the possible decrease in the incidence of dementia. However, the real link between cognition and function still needs to be investigated. ACKNOWLEDGMENTS Author affiliations: University of Bordeaux, Inserm, Bordeaux Population Health Research Center, UMR1219, Bordeaux, France (Leslie Grasset, Hélène Jacqmin-Gadda, Cécile Proust-Lima, Karine Pérès, Hélène Amieva, Jean-François Dartigues, Catherine Helmer); and Bordeaux University Hospital, Memory Consultation, Memory Resource and Research Centre, Bordeaux, France (Jean-François Dartigues). The PAQUID study was funded by IPSEN France, Novartis Pharma France, and the Caisse Nationale de Solidarité et d’Autonomie. This work received no specific funding or grants beyond the funding for the PAQUID cohort. The funders had no role in study design; in data collection, analysis, and interpretation; or in writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Conflict of interest: J.-F.D. has received research grants from IPSEN and Roche, outside the submitted work. The other authors report no conflicts. Abbreviations 4-IADL 4 Instrumental Activities of Daily Living BVRT Benton Visual Retention Test CI confidence interval DSST Digit Symbol Substitution Test G1 first generation G2 second generation IST Isaacs Set Test MMSE Mini-Mental State Examination OR odds ratio PAQUID Personnes Agées Quid REFERENCES 1 Alzheimer’s Disease International . World Alzheimer Report: The Global Impact of Dementia. 2015 ; http://www.alz.co.uk/research/WorldAlzheimerReport2015.pdf. Accessed April 22, 2016. 2 Matthews FE , Arthur A , Barnes LE , et al. . A two-decade comparison of prevalence of dementia in individuals aged 65 years and older from three geographical areas of England: results of the Cognitive function and Ageing Study I and II . Lancet . 2013 ; 382 ( 9902 ): 1405 – 1412 . Google Scholar Crossref Search ADS PubMed 3 Qiu C , von Strauss E , Bäckman L , et al. . Twenty-year changes in dementia occurrence suggest decreasing incidence in central Stockholm, Sweden . Neurology . 2013 ; 80 ( 20 ): 1888 – 1894 . Google Scholar Crossref Search ADS PubMed 4 Lobo A , Saz P , Marcos G , et al. . Prevalence of dementia in a southern European population in two different time periods: the ZARADEMP Project . Acta Psychiatr Scand . 2007 ; 116 ( 4 ): 299 – 307 . Google Scholar Crossref Search ADS PubMed 5 Doblhammer G , Fink A , Fritze T . Short-term trends in dementia prevalence in Germany between the years 2007 and 2009 . Alzheimers Dement . 2015 ; 11 ( 3 ): 291 – 299 . Google Scholar Crossref Search ADS PubMed 6 Schrijvers EM , Verhaaren BF , Koudstaal PJ , et al. . Is dementia incidence declining?: Trends in dementia incidence since 1990 in the Rotterdam Study . Neurology . 2012 ; 78 ( 19 ): 1456 – 1463 . Google Scholar Crossref Search ADS PubMed 7 Rocca WA , Petersen RC , Knopman DS , et al. . Trends in the incidence and prevalence of Alzheimer’s disease, dementia, and cognitive impairment in the United States . Alzheimers Dement . 2011 ; 7 ( 1 ): 80 – 93 . Google Scholar Crossref Search ADS PubMed 8 Grasset L , Brayne C , Joly P , et al. . Trends in dementia incidence: evolution over a 10-year period in France . Alzheimers Dement . 2016 ; 12 ( 3 ): 272 – 280 . Google Scholar Crossref Search ADS PubMed 9 Satizabal CL , Beiser AS , Chouraki V , et al. . Incidence of dementia over three decades in the Framingham Heart Study . N Engl J Med . 2016 ; 374 ( 6 ): 523 – 532 . Google Scholar Crossref Search ADS PubMed 10 Jacqmin-Gadda H , Alperovitch A , Montlahuc C , et al. . 20-Year prevalence projections for dementia and impact of preventive policy about risk factors . Eur J Epidemiol . 2013 ; 28 ( 6 ): 493 – 502 . Google Scholar Crossref Search ADS PubMed 11 Norton S , Matthews FE , Barnes DE , et al. . Potential for primary prevention of Alzheimer’s disease: an analysis of population-based data . Lancet Neurol . 2014 ; 13 ( 8 ): 788 – 794 . Google Scholar Crossref Search ADS PubMed 12 Gerstorf D , Ram N , Hoppmann C , et al. . Cohort differences in cognitive aging and terminal decline in the Seattle Longitudinal Study . Dev Psychol . 2011 ; 47 ( 4 ): 1026 – 1041 . Google Scholar Crossref Search ADS PubMed 13 Dodge HH , Zhu J , Hughes TF , et al. . Cohort effects in verbal memory function and practice effects: a population-based study . Int Psychogeriatr . 2017 ; 29 ( 1 ): 137 – 148 . Google Scholar Crossref Search ADS PubMed 14 Dodge HH , Zhu J , Lee CW , et al. . Cohort effects in age-associated cognitive trajectories . J Gerontol A Biol Sci Med Sci . 2014 ; 69 ( 6 ): 687 – 694 . Google Scholar Crossref Search ADS PubMed 15 Karlsson P , Thorvaldsson V , Skoog I , et al. . Birth cohort differences in fluid cognition in old age: comparisons of trends in levels and change trajectories over 30 years in three population-based samples . Psychol Aging . 2015 ; 30 ( 1 ): 83 – 94 . Google Scholar Crossref Search ADS PubMed 16 Brailean A , Huisman M , Prince M , et al. . Cohort differences in cognitive aging in the Longitudinal Aging Study Amsterdam [published online ahead of print September 30, 2016]. J Gerontol B Psychol Sci Soc Sci . 2016 (doi: 10.1093/geronb/gbw129 ). 17 Zelinski EM , Kennison RF . Not your parents’ test scores: cohort reduces psychometric aging effects . Psychol Aging . 2007 ; 22 ( 3 ): 546 – 557 . Google Scholar Crossref Search ADS PubMed 18 Finkel D , Reynolds CA , McArdle JJ , et al. . Cohort differences in trajectories of cognitive aging . J Gerontol B Psychol Sci Soc Sci . 2007 ; 62 ( 5 ): P286 – P294 . Google Scholar Crossref Search ADS PubMed 19 Chong Y , Tejada Vera B , Lu L , et al. . Mortality trends in the United States, 1900–2013. https://www.cdc.gov/nchs/data-visualization/mortality-trends/. Accessed April 22, 2016. 20 Dartigues JF , Gagnon M , Letenneur L , et al. . Principal lifetime occupation and cognitive impairment in a French elderly cohort (Paquid) . Am J Epidemiol . 1992 ; 135 ( 9 ): 981 – 988 . Google Scholar Crossref Search ADS PubMed 21 Jacqmin-Gadda H , Fabrigoule C , Commenges D , et al. . A 5-year longitudinal study of the Mini-Mental State Examination in normal aging . Am J Epidemiol . 1997 ; 145 ( 6 ): 498 – 506 . Google Scholar Crossref Search ADS PubMed 22 Folstein MF , Folstein SE , McHugh PR . “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . 1975 ; 12 ( 3 ): 189 – 198 . Google Scholar Crossref Search ADS PubMed 23 Isaacs B , Kennie AT . The set test as an aid to the detection of dementia in old people . Br J Psychiatry . 1973 ; 123 ( 575 ): 467 – 470 . Google Scholar Crossref Search ADS PubMed 24 Benton A . Manuel Pour L’Application du Test de Retention Visuelle. Applications Cliniques et Expérimentales . Deuxième édition française. Paris, France : Centre de Psychologie Appliquée ; 1965 . 25 Wechsler D . WAIS-R Manual . New York, NY : Psychological Corporation ; 1981 . 26 Barberger-Gateau P , Commenges D , Gagnon M , et al. . Instrumental activities of daily living as a screening tool for cognitive impairment and dementia in elderly community dwellers . J Am Geriatr Soc . 1992 ; 40 ( 11 ): 1129 – 1134 . Google Scholar Crossref Search ADS PubMed 27 Barberger-Gateau P , Dartigues JF , Letenneur L . Four Instrumental Activities of Daily Living Score as a predictor of one-year incident dementia . Age Ageing . 1993 ; 22 ( 6 ): 457 – 463 . Google Scholar Crossref Search ADS PubMed 28 Rizopoulos D . Joint Models for Longitudinal and Time-to-Event Data: With Applications in R . Vol 6 . Abingdon, United Kingdom : Chapman & Hall/CRC Biostatistics Series ; 2012 . 29 Proust-Lima C , Dartigues JF , Jacqmin-Gadda H . Misuse of the linear mixed model when evaluating risk factors of cognitive decline . Am J Epidemiol . 2011 ; 174 ( 9 ): 1077 – 1088 . Google Scholar Crossref Search ADS PubMed 30 Philipps V , Amieva H , Andrieu S , et al. . Normalized Mini-Mental State Examination for assessing cognitive change in population-based brain aging studies . Neuroepidemiology . 2014 ; 43 ( 1 ): 15 – 25 . Google Scholar Crossref Search ADS PubMed 31 Proust-Lima C , Philipps V , Liquet B . Estimation of extended mixed models using latent classes and latent processes: the R package lcmm . J Stat Softw . 2017 ; 78 ( 2 ): 1 – 56 . Google Scholar Crossref Search ADS 32 Rizopoulos DJM . JM: an R package for the joint modelling of londitudinal and time-to-event data . J Stat Softw . 2010 ; 35 ( 9 ): 1 – 33 . Google Scholar Crossref Search ADS PubMed 33 Rizopoulos D . The R package JMbayes for fitting Joint models for longitudinal and Time-to-event data using MCMC . J Stat Softw . 2016 ; 72 ( 7 ): 1 – 44 . Google Scholar Crossref Search ADS 34 Lenehan ME , Summers MJ , Saunders NL , et al. . Relationship between education and age-related cognitive decline: a review of recent research . Psychogeriatrics . 2015 ; 15 ( 2 ): 154 – 162 . Google Scholar Crossref Search ADS PubMed 35 Salthouse TA . Implications of the Flynn effect for age-cognition relations . Intelligence . 2015 ; 48 : 51 – 57 . Google Scholar Crossref Search ADS PubMed 36 Skirbekk V , Stonawski M , Bonsang E , et al. . The Flynn effect and population aging . Intelligence . 2013 ; 41 ( 3 ): 169 – 177 . Google Scholar Crossref Search ADS 37 Le Carret N , Rainville C , Lechevallier N , et al. . Influence of education on the Benton visual retention test performance as mediated by a strategic search component . Brain Cogn . 2003 ; 53 ( 2 ): 408 – 411 . Google Scholar Crossref Search ADS PubMed 38 Steiber N . Population aging at cross-roads: diverging secular trends in average cognitive functioning and physical health in the older population of Germany . PLoS One . 2015 ; 10 ( 8 ): e0136583 . Google Scholar Crossref Search ADS PubMed 39 de Rotrou J , Wu YH , Mabire JB , et al. . Does cognitive function increase over time in the healthy elderly? PLoS One . 2013 ; 8 ( 11 ): e78646 . Google Scholar Crossref Search ADS PubMed 40 Sacuiu S , Gustafson D , Sjögren M , et al. . Secular changes in cognitive predictors of dementia and mortality in 70-year-olds . Neurology . 2010 ; 75 ( 9 ): 779 – 785 . Google Scholar Crossref Search ADS PubMed 41 Christensen K , Thinggaard M , Oksuzyan A , et al. . Physical and cognitive functioning of people older than 90 years: a comparison of two danish cohorts born 10 years apart . Lancet . 2013 ; 382 ( 9903 ): 1507 – 1513 . Google Scholar Crossref Search ADS PubMed 42 Vivot A , Power MC , Glymour MM , et al. . Jump, hop, or skip: modeling practice effects in studies of determinants of cognitive change in older adults . Am J Epidemiol . 2016 ; 183 ( 4 ): 302 – 314 . Google Scholar Crossref Search ADS PubMed 43 Jacqmin-Gadda H , Fabrigoule C , Commenges D , et al. . Longitudinal study of cognitive aging in non-demented elderly subjects [in French]. Rev Epidemiol Sante Publique . 1997 ; 45 ( 5 ): 363 – 372 . Google Scholar PubMed 44 Falk H , Johansson L , Ostling S , et al. . Functional disability and ability 75-year-olds: a comparison of two Swedish cohorts born 30 years apart . Age Ageing . 2014 ; 43 ( 5 ): 636 – 641 . Google Scholar Crossref Search ADS PubMed 45 Heikkinen E , Kauppinen M , Rantanen T , et al. . Cohort differences in health, functioning and physical activity in the young-old finnish population . Aging Clin Exp Res . 2011 ; 23 ( 2 ): 126 – 134 . Google Scholar PubMed 46 Pérès K , Helmer C , Letenneur L , et al. . Ten-year change in disability prevalence and related factors in two generations of French elderly community dwellers: data from the PAQUID study . Aging Clin Exp Res . 2005 ; 17 ( 3 ): 229 – 235 . Google Scholar Crossref Search ADS PubMed 47 Jagger C , Matthews RJ , Matthews FE , et al. . Cohort differences in disease and disability in the young-old: findings from the MRC Cognitive Function and Ageing Study (MRC-CFAS) . BMC Public Health . 2007 ; 7 : 156 . Google Scholar Crossref Search ADS PubMed 48 Amieva H , Mokri H , Le Goff M , et al. . Compensatory mechanisms in higher-educated subjects with Alzheimer’s disease: a study of 20 years of cognitive decline . Brain . 2014 ; 137 ( Pt 4 ): 1167 – 1175 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

B, Andrea, Sarah;C, Messer, Lynne;Miguel, Marino,;Janne, Boone-Heinonen,

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy123pmid: 29893781

Abstract Precarious work is concentrated in the service industry in the United States and is a risk factor for poor mental health. Service occupations in which workers receive tips are potentially more precarious due to unstable schedule and income, and lack of benefits. We tested hypotheses that individuals working in tipped service occupations have greater odds of experiencing poor mental health (as indicated by self-reported depression, sleep problems, and/or greater perceived stress) relative to individuals in untipped service and nonservice occupations, using cross-sectional data from wave IV of the National Longitudinal Study of Adolescent to Adult Health data set (2007–2008; age range, 24–33 years). To improve comparability of occupation types, propensity scores were computed as a function of childhood factors, then used to construct a sample of 2,815 women and 2,586 men. In gender-stratified multivariable regression, women in tipped service had greater odds of reporting a depression diagnosis or symptoms relative to women in nonservice work (odds ratio = 1.61; 95% confidence interval: 1.11, 2.34). Associations of similar magnitude for sleep problems and perceived stress were observed among women but were not statistically significant; all associations were close to the null among men. Additional research is necessary to understand the factors that underlie differences in poor mental health in tipped and untipped service versus nonservice workers. employment, gender, occupational health, occupational stress, precarious work, psychosocial stress, tipped work An individual’s occupation can lead to differential exposure to physical, psychosocial, and environmental factors with the potential to influence their health (1). Individuals in service occupations, especially tipped service, may be particularly vulnerable; however, the potential health effects of these occupations are understudied. Service work is precarious (2) and characterized by lack of control over hours and shift worked (3, 4), insufficient benefits (5–7), and lower wages. Service workers represent 42.5% of US workers earning the federal minimum wage and 78.1% of US workers earning less than minimum wage (8). Tipped work may be particularly precarious service work for several reasons. First, the normalization of tipping in certain service occupations in the United States has led to differential minimum wage standards; workers in tipped occupations can be paid a wage 71% lower than the federal minimum wage (9) with the expectation that highly unpredictable and inequitable tips from customers will make up the difference (7). On average, tipped workers are nearly twice as likely to live in poverty relative to untipped workers (7). Second, tipped workers are disproportionately exposed to last-minute scheduling practices (3, 5) and insufficient provision of benefits (7). Third, workers in tipped and untipped service occupations must frequently express or suppress certain emotions during interactions with customers (10–14) and manage sexualized or hostile customer behavior (15). These aspects of tipped service work have direct consequences, such as physical harm, and indirect consequences for health, such as psychosocial stress, with the latter representing an important determinant of mental health (16, 17). A 2007 report from the National Survey on Drug Use and Health revealed the 12-month prevalence of depression among workers aged 18–64 years was highest among those in personal care and service (10.8%) and food preparation– and serving–related occupations (10.3%) (18). Similarly, prevalence of short sleep duration and of other sleep disturbances is highest within service-industry occupation categories (19) and in financially precarious employment in general (20). However, the potential health implications of tipped work have been minimally assessed and are limited to substance use (21). In light of the dearth of research on the potential impact of working in a tipped service occupation on health, our objective was to test the hypotheses that 1) individuals in service occupations (tipped and untipped) have greater odds of experiencing depression, sleep problems, and/or stress relative to their nonservice counterparts; and 2) individuals in tipped service occupations are particularly vulnerable to these mental health outcomes. METHODS Participants We used data from the National Longitudinal Study of Adolescent to Adult Health (Add Health), a nationally representative cohort of US adolescents followed for more than 14 years into adulthood (22). In Add Health, a stratified, school-based, clustered sampling design was used to ensure data were representative of the US adolescent, school population. Detailed information on the Add Health study design and procedures are described elsewhere (22). Briefly, a core subset of respondents and parents was randomly selected from within school and gender strata to participate in in-home interviews by trained interviewers (23). Of 20,745 baseline respondents, 76% (n = 15,701) completed interviews for wave IV of Add Health, when participants were aged 24–33 years. This study was exempted by the Oregon Health & Science University institutional review board. Variables and measurement Exposure: occupation type (nonservice, tipped service, untipped service) Occupation type was classified from responses to 2 wave IV questions about current or recent occupation. The classification “service” was assigned to participants if their response to the question “When you see the list of categories, please tell me which best describes what you (do/did) at your (current/most recent) job?” was consistent with a service occupation according to the US Bureau of Labor Statistics industry classification system (e.g., “food preparation and serving occupation”) (Web Table 1; available at https://academic.oup.com/aje) (24). The classification “nonservice” was assigned to participants reporting other occupation types. Tipped service was assigned to respondents classified as working in service occupations if their response to the question “Out of these categories, which one best describes this job?” was consistent with a predominantly tipped occupation according to the Economic Policy Institute (e.g., “waiters and waitresses”; Web Table 1) (7). The classification “untipped service” was assigned to respondents in service occupations reporting any other occupation. Outcomes The following variables developed by Add Health (25) were constructed in wave IV to be used as measures of the 3 mental health outcomes: 1) Depression or depressive symptoms (yes/no) were defined as self-reported diagnosis of depression and/or depressive symptoms within the past 7 days reported on the modified version of the Columbia Center for Epidemiological Studies Depression Scale-10; 2) sleep problems (yes/no) included self-reported difficulty falling or staying asleep and/or symptoms of sleep apnea over the past 4 weeks; and 3) perceived stress was a 3-level ordinal variable constructed from tertiles of the Cohen’s Perceived Stress Scale score (i.e., 0–3, 4–6, and 7–16). There are no broadly applicable score cutoffs and others advised within-sample comparisons (26). Statistical analysis Analyses were conducted in Stata/IC, version 14.2 (StataCorp LP, College Station, Texas), incorporating Add Health survey weights and sample design parameters to account for clustered sampling, attrition, and oversampling, thus approximating the target population of US adolescents in grades 7–12 in 1994. The analytic sample was restricted to respondents who participated in waves I–IV and had an Add Health sampling weight for analysis (n = 9,421), reported a current or recent job in wave IV (n = 9,205), and had complete exposure and outcome data (n = 9,140). Participants with missing covariates were included; multiple imputation methods were applied as described later in Methods. Add Health sampling weights incorporate a nonresponse adjustment for nonparticipation in 1 or more wave of in-home interviews (27). The analytic sample thus contained 9,140 respondents (n = 4,996 women and n = 4,144 men) before application of propensity score (PS) methods. Propensity score methods Our analytic approach addressed 2 methodological challenges related to occupation stratification. First, there is gender-based stratification into occupational categories and specific occupations within those categories (28): 56.6% of all service workers (29) and 67% of tipped workers are women (7). Second, additional nonrandom assignment to occupational category resulting from social selection based on sociodemographic characteristics and other predisposing life experiences (30) may affect health. Therefore, we stratified all analyses by gender and used gender-specific PS to address residual structural confounding present in occupation-type assignment. For each gender, a single set of PS was generated to be used for all outcome models using variables for participant sociodemographics, parental characteristics, and childhood adverse experiences, behaviors, and health (Web Appendix 1) selected using our conceptual framework (Web Figure 1). Web Appendix 2 describes the process used to calculate PS, including the application of multiple imputation to address missingness of variables pertinent to PS development. Multinomial logistic regression was used to model occupation type as a function of these variables and predicted probabilities for each occupation type were computed (Web Appendix 2, Web Table 2). PS first served as decision aids for visually guided restriction of the analytic sample to satisfy the positivity assumption and provide support for exchangeability (Web Appendix 3) (31). Multivariable analysis of outcomes Multivariable analyses were conducted with the PS-restricted sample. Logistic regression was used for binary outcomes (depression, sleep problems) and ordinal logistic regression was used for the ordered 3-level categorical outcome (perceived stress), producing odds ratios. Though prevalence of study outcomes was high (>10%), it was not possible to estimate relative risk, because of model complexity with survey design parameters, multiple imputation, and inclusion of an ordinal outcome Within the PS-restricted sample, PS regression adjustment was used in multivariable analyses to achieve models that were parsimonious and adequately adjusted (Web Appendix 4). In addition to PS regression adjustment, variables that remained unbalanced after sample restriction were included in our models for residual confounding adjustment. Sensitivity analyses We conducted 3 sensitivity analyses. First, the prevalence of childhood depression is disproportionately higher among individuals in service occupations and previous depression is a strong predictor of future depression (32). Therefore, to further account for the social selection of individuals with poor mental health into service occupations, we restricted analyses to respondents with no prior reported depression (i.e., those who had wave II and III Center for Epidemiological Studies Depression Scale scores ≤3). Similarly, we restricted assessment of the association between occupation type and sleep problems to respondents who reported never having difficulty falling or staying asleep or having difficulty “just a few times” in childhood (wave II). Measures of perceived stress in childhood were unavailable. Third, to examine the potential contributions of precarious work beyond the effects of underemployment (33), all outcomes were assessed with data restricted to those working full-time (≥35 hours/week). RESULTS Selected characteristics of 4,996 women and 4,144 men who reported a current or recent job during their wave IV interview are presented in Tables 1 and 2, respectively. Participants in the full analytic sample were, on average, 28 years old at wave IV (data not shown). Prior to PS-based sample restriction, women in service occupations tended to experience more adversity in early life, whereas women in nonservice occupations were more advantaged. For instance, parental income and educational attainment were highest for women in nonservice (mean income = $49,400; 36.1% college graduates) and lowest in untipped service occupations (mean income = $39,500; 19.4% college graduates). In contrast, parental incarceration was lowest among women in nonservice (14.0%) and highest among those in untipped service (23.8%) occupations. This trend was not as prominent in men. Among women and men, high educational attainment was most common among individuals in nonservice occupations. At the wave IV interview, women reported higher depression prevalence (across all occupation types: 25.6% in women vs. 13.5% in men). Table 1. Select Characteristicsa of Women Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,751) PS-Restricted Sample (n = 1,990) Full Sample (n = 931) PS-Restricted Sample (n = 614) Full Sample (n = 314) PS-Restricted Sample (n = 211) Sociodemographic Characteristics Race  White 71.7 74.7 66.6 67.0 80.5 81.1  Black 14.1 11.5 21.0 18.3 7.8 8.1  Other 14.2 13.8 12.4 14.7 11.8 10.8 Hispanic ethnicity 12.3 14.2 10.4 13.4 7.6 9.9 Parent’s education (wave I)  Less than high school 8.9 8.6 16.6 11.6 9.6 6.1  High school graduate 26.1 32.6 34.6 37.9 29.7 28.3  Some college or vocational training 28.8 34.1 29.3 30.4 35.0 33.8  College graduate 36.1 24.7 19.4 20.0 25.7 31.7 Parent’s income (in $1,000)b 49.4 (2.4) 43.1 (1.9) 39.5 (2.7) 38.8 (2.3) 42.6 (2.4) 42.7 (2.7) Parental incarceration 14.0 14.1 23.8 18.9 18.2 17.6 Highest level of education  Less than high school 5.1 3.5 12.4 5.5 8.6 7.1  High school graduate 12.3 19.4 16.3 20.4 16.8 17.0  Some college or vocational training 39.7 56.3 57.3 68.1 62.9 67.5  College graduate 42.9 20.8 14.0 6.0 11.7 8.4 Household income (wave IV; in $1,000)b 63.8 (1.3) 61.3 (1.4) 45.8 (1.6) 46.5 (1.9) 50.5 (2.8) 48.5 (3.1) Mental Health Outcomes (Wave IV) Depression 22.8 24.6 31.7 32.1 37.0 37.8 Sleep problems 9.5 11.4 16.9 18.2 13.8 16.6 Cohen perceived stress score tertile  Low (0–3) 36.1 33.3 29.0 30.0 22.8 24.3  Medium (4–6) 36.7 37.2 31.8 32.2 38.1 37.3  High (7–18) 27.1 29.5 39.2 37.8 39.0 38.4 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,751) PS-Restricted Sample (n = 1,990) Full Sample (n = 931) PS-Restricted Sample (n = 614) Full Sample (n = 314) PS-Restricted Sample (n = 211) Sociodemographic Characteristics Race  White 71.7 74.7 66.6 67.0 80.5 81.1  Black 14.1 11.5 21.0 18.3 7.8 8.1  Other 14.2 13.8 12.4 14.7 11.8 10.8 Hispanic ethnicity 12.3 14.2 10.4 13.4 7.6 9.9 Parent’s education (wave I)  Less than high school 8.9 8.6 16.6 11.6 9.6 6.1  High school graduate 26.1 32.6 34.6 37.9 29.7 28.3  Some college or vocational training 28.8 34.1 29.3 30.4 35.0 33.8  College graduate 36.1 24.7 19.4 20.0 25.7 31.7 Parent’s income (in $1,000)b 49.4 (2.4) 43.1 (1.9) 39.5 (2.7) 38.8 (2.3) 42.6 (2.4) 42.7 (2.7) Parental incarceration 14.0 14.1 23.8 18.9 18.2 17.6 Highest level of education  Less than high school 5.1 3.5 12.4 5.5 8.6 7.1  High school graduate 12.3 19.4 16.3 20.4 16.8 17.0  Some college or vocational training 39.7 56.3 57.3 68.1 62.9 67.5  College graduate 42.9 20.8 14.0 6.0 11.7 8.4 Household income (wave IV; in $1,000)b 63.8 (1.3) 61.3 (1.4) 45.8 (1.6) 46.5 (1.9) 50.5 (2.8) 48.5 (3.1) Mental Health Outcomes (Wave IV) Depression 22.8 24.6 31.7 32.1 37.0 37.8 Sleep problems 9.5 11.4 16.9 18.2 13.8 16.6 Cohen perceived stress score tertile  Low (0–3) 36.1 33.3 29.0 30.0 22.8 24.3  Medium (4–6) 36.7 37.2 31.8 32.2 38.1 37.3  High (7–18) 27.1 29.5 39.2 37.8 39.0 38.4 Abbreviation: PS, propensity score. a Percentages, means, and standard errors were calculated by accounting for survey weights, strata, and clusters. b Values are expressed as mean (standard error). Table 1. Select Characteristicsa of Women Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,751) PS-Restricted Sample (n = 1,990) Full Sample (n = 931) PS-Restricted Sample (n = 614) Full Sample (n = 314) PS-Restricted Sample (n = 211) Sociodemographic Characteristics Race  White 71.7 74.7 66.6 67.0 80.5 81.1  Black 14.1 11.5 21.0 18.3 7.8 8.1  Other 14.2 13.8 12.4 14.7 11.8 10.8 Hispanic ethnicity 12.3 14.2 10.4 13.4 7.6 9.9 Parent’s education (wave I)  Less than high school 8.9 8.6 16.6 11.6 9.6 6.1  High school graduate 26.1 32.6 34.6 37.9 29.7 28.3  Some college or vocational training 28.8 34.1 29.3 30.4 35.0 33.8  College graduate 36.1 24.7 19.4 20.0 25.7 31.7 Parent’s income (in $1,000)b 49.4 (2.4) 43.1 (1.9) 39.5 (2.7) 38.8 (2.3) 42.6 (2.4) 42.7 (2.7) Parental incarceration 14.0 14.1 23.8 18.9 18.2 17.6 Highest level of education  Less than high school 5.1 3.5 12.4 5.5 8.6 7.1  High school graduate 12.3 19.4 16.3 20.4 16.8 17.0  Some college or vocational training 39.7 56.3 57.3 68.1 62.9 67.5  College graduate 42.9 20.8 14.0 6.0 11.7 8.4 Household income (wave IV; in $1,000)b 63.8 (1.3) 61.3 (1.4) 45.8 (1.6) 46.5 (1.9) 50.5 (2.8) 48.5 (3.1) Mental Health Outcomes (Wave IV) Depression 22.8 24.6 31.7 32.1 37.0 37.8 Sleep problems 9.5 11.4 16.9 18.2 13.8 16.6 Cohen perceived stress score tertile  Low (0–3) 36.1 33.3 29.0 30.0 22.8 24.3  Medium (4–6) 36.7 37.2 31.8 32.2 38.1 37.3  High (7–18) 27.1 29.5 39.2 37.8 39.0 38.4 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,751) PS-Restricted Sample (n = 1,990) Full Sample (n = 931) PS-Restricted Sample (n = 614) Full Sample (n = 314) PS-Restricted Sample (n = 211) Sociodemographic Characteristics Race  White 71.7 74.7 66.6 67.0 80.5 81.1  Black 14.1 11.5 21.0 18.3 7.8 8.1  Other 14.2 13.8 12.4 14.7 11.8 10.8 Hispanic ethnicity 12.3 14.2 10.4 13.4 7.6 9.9 Parent’s education (wave I)  Less than high school 8.9 8.6 16.6 11.6 9.6 6.1  High school graduate 26.1 32.6 34.6 37.9 29.7 28.3  Some college or vocational training 28.8 34.1 29.3 30.4 35.0 33.8  College graduate 36.1 24.7 19.4 20.0 25.7 31.7 Parent’s income (in $1,000)b 49.4 (2.4) 43.1 (1.9) 39.5 (2.7) 38.8 (2.3) 42.6 (2.4) 42.7 (2.7) Parental incarceration 14.0 14.1 23.8 18.9 18.2 17.6 Highest level of education  Less than high school 5.1 3.5 12.4 5.5 8.6 7.1  High school graduate 12.3 19.4 16.3 20.4 16.8 17.0  Some college or vocational training 39.7 56.3 57.3 68.1 62.9 67.5  College graduate 42.9 20.8 14.0 6.0 11.7 8.4 Household income (wave IV; in $1,000)b 63.8 (1.3) 61.3 (1.4) 45.8 (1.6) 46.5 (1.9) 50.5 (2.8) 48.5 (3.1) Mental Health Outcomes (Wave IV) Depression 22.8 24.6 31.7 32.1 37.0 37.8 Sleep problems 9.5 11.4 16.9 18.2 13.8 16.6 Cohen perceived stress score tertile  Low (0–3) 36.1 33.3 29.0 30.0 22.8 24.3  Medium (4–6) 36.7 37.2 31.8 32.2 38.1 37.3  High (7–18) 27.1 29.5 39.2 37.8 39.0 38.4 Abbreviation: PS, propensity score. a Percentages, means, and standard errors were calculated by accounting for survey weights, strata, and clusters. b Values are expressed as mean (standard error). Table 2. Select Characteristicsa of Men Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,446) PS-Restricted Sample (n = 2,145) Full Sample (n = 586) PS-Restricted Sample (n = 372) Full Sample (n = 112) PS-Restricted Sample (n = 69) Sociodemographic Characteristics Race  White 71.9 75.1 64.8 71.7 66.5 70.7  Black 12.5 11.4 22.7 16.5 15.6 11.3  Other 15.6 13.5 12.5 11.8 17.8 18 Hispanic ethnicity 12.7 11.0 10.7 9.0 7.0 5.3 Parent’s education (wave I)  Less than high school 10.0 8.0 13.8 8.4 8.4 5.0  High school graduate 25.3 26.8 26.4 27.5 18.6 22.3  Some college or vocational training 31.4 32.9 30.5 33.9 36.7 35.6  College graduate 33.7 32.3 29.4 30.2 36.3 37.1 Parent’s income (in $1,000)b 46.5 (2.1) 44.2 (1.7) 40.9 (2.5) 41.8 (2.3) 44.7 (3.5) 45.1 (4.5) Parental incarceration 17.2 17.6 15.1 14.3 17.8 13.2 Highest level of education  Less than high school graduate 10.4 5.0 12.1 9.0 3.3 0.0  High school graduate 19.6 26.0 23.0 23.8 17.1 22.5  Some college or vocational training 38.5 49.5 48.9 55.9 63.7 74.9  College graduate 31.4 19.5 16.1 11.4 16.3 2.5 Household income (wave IV; in $1,000)b 65.8 (1.2) 64.6 (1.3) 55.0 (2.6) 53.8 (2.7) 54.9 (4.1) 60.6 (5.6) Mental Health Outcomes (Wave IV) Depression 13.0 12.4 16.1 14.9 15.7 10.8 Sleep problems 11.4 11.1 11.5 10.3 10.7 14.7 Cohen perceived stress score tertile  Low (0–3) 39.8 38.3 40.4 41.0 29.9 29.5  Medium (4–6) 37.5 38.8 30.1 29.0 40.6 45.3  High (7–18) 22.6 22.9 29.5 30.0 29.5 25.2 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,446) PS-Restricted Sample (n = 2,145) Full Sample (n = 586) PS-Restricted Sample (n = 372) Full Sample (n = 112) PS-Restricted Sample (n = 69) Sociodemographic Characteristics Race  White 71.9 75.1 64.8 71.7 66.5 70.7  Black 12.5 11.4 22.7 16.5 15.6 11.3  Other 15.6 13.5 12.5 11.8 17.8 18 Hispanic ethnicity 12.7 11.0 10.7 9.0 7.0 5.3 Parent’s education (wave I)  Less than high school 10.0 8.0 13.8 8.4 8.4 5.0  High school graduate 25.3 26.8 26.4 27.5 18.6 22.3  Some college or vocational training 31.4 32.9 30.5 33.9 36.7 35.6  College graduate 33.7 32.3 29.4 30.2 36.3 37.1 Parent’s income (in $1,000)b 46.5 (2.1) 44.2 (1.7) 40.9 (2.5) 41.8 (2.3) 44.7 (3.5) 45.1 (4.5) Parental incarceration 17.2 17.6 15.1 14.3 17.8 13.2 Highest level of education  Less than high school graduate 10.4 5.0 12.1 9.0 3.3 0.0  High school graduate 19.6 26.0 23.0 23.8 17.1 22.5  Some college or vocational training 38.5 49.5 48.9 55.9 63.7 74.9  College graduate 31.4 19.5 16.1 11.4 16.3 2.5 Household income (wave IV; in $1,000)b 65.8 (1.2) 64.6 (1.3) 55.0 (2.6) 53.8 (2.7) 54.9 (4.1) 60.6 (5.6) Mental Health Outcomes (Wave IV) Depression 13.0 12.4 16.1 14.9 15.7 10.8 Sleep problems 11.4 11.1 11.5 10.3 10.7 14.7 Cohen perceived stress score tertile  Low (0–3) 39.8 38.3 40.4 41.0 29.9 29.5  Medium (4–6) 37.5 38.8 30.1 29.0 40.6 45.3  High (7–18) 22.6 22.9 29.5 30.0 29.5 25.2 Abbreviation: PS, propensity score. a Percentages, means, and standard errors were calculated by accounting for survey weights, strata, and clusters. b Values are expressed as mean (standard error). Table 2. Select Characteristicsa of Men Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,446) PS-Restricted Sample (n = 2,145) Full Sample (n = 586) PS-Restricted Sample (n = 372) Full Sample (n = 112) PS-Restricted Sample (n = 69) Sociodemographic Characteristics Race  White 71.9 75.1 64.8 71.7 66.5 70.7  Black 12.5 11.4 22.7 16.5 15.6 11.3  Other 15.6 13.5 12.5 11.8 17.8 18 Hispanic ethnicity 12.7 11.0 10.7 9.0 7.0 5.3 Parent’s education (wave I)  Less than high school 10.0 8.0 13.8 8.4 8.4 5.0  High school graduate 25.3 26.8 26.4 27.5 18.6 22.3  Some college or vocational training 31.4 32.9 30.5 33.9 36.7 35.6  College graduate 33.7 32.3 29.4 30.2 36.3 37.1 Parent’s income (in $1,000)b 46.5 (2.1) 44.2 (1.7) 40.9 (2.5) 41.8 (2.3) 44.7 (3.5) 45.1 (4.5) Parental incarceration 17.2 17.6 15.1 14.3 17.8 13.2 Highest level of education  Less than high school graduate 10.4 5.0 12.1 9.0 3.3 0.0  High school graduate 19.6 26.0 23.0 23.8 17.1 22.5  Some college or vocational training 38.5 49.5 48.9 55.9 63.7 74.9  College graduate 31.4 19.5 16.1 11.4 16.3 2.5 Household income (wave IV; in $1,000)b 65.8 (1.2) 64.6 (1.3) 55.0 (2.6) 53.8 (2.7) 54.9 (4.1) 60.6 (5.6) Mental Health Outcomes (Wave IV) Depression 13.0 12.4 16.1 14.9 15.7 10.8 Sleep problems 11.4 11.1 11.5 10.3 10.7 14.7 Cohen perceived stress score tertile  Low (0–3) 39.8 38.3 40.4 41.0 29.9 29.5  Medium (4–6) 37.5 38.8 30.1 29.0 40.6 45.3  High (7–18) 22.6 22.9 29.5 30.0 29.5 25.2 Variable Nonservice, % Untipped Service, % Tipped Service, % Full Sample (n = 3,446) PS-Restricted Sample (n = 2,145) Full Sample (n = 586) PS-Restricted Sample (n = 372) Full Sample (n = 112) PS-Restricted Sample (n = 69) Sociodemographic Characteristics Race  White 71.9 75.1 64.8 71.7 66.5 70.7  Black 12.5 11.4 22.7 16.5 15.6 11.3  Other 15.6 13.5 12.5 11.8 17.8 18 Hispanic ethnicity 12.7 11.0 10.7 9.0 7.0 5.3 Parent’s education (wave I)  Less than high school 10.0 8.0 13.8 8.4 8.4 5.0  High school graduate 25.3 26.8 26.4 27.5 18.6 22.3  Some college or vocational training 31.4 32.9 30.5 33.9 36.7 35.6  College graduate 33.7 32.3 29.4 30.2 36.3 37.1 Parent’s income (in $1,000)b 46.5 (2.1) 44.2 (1.7) 40.9 (2.5) 41.8 (2.3) 44.7 (3.5) 45.1 (4.5) Parental incarceration 17.2 17.6 15.1 14.3 17.8 13.2 Highest level of education  Less than high school graduate 10.4 5.0 12.1 9.0 3.3 0.0  High school graduate 19.6 26.0 23.0 23.8 17.1 22.5  Some college or vocational training 38.5 49.5 48.9 55.9 63.7 74.9  College graduate 31.4 19.5 16.1 11.4 16.3 2.5 Household income (wave IV; in $1,000)b 65.8 (1.2) 64.6 (1.3) 55.0 (2.6) 53.8 (2.7) 54.9 (4.1) 60.6 (5.6) Mental Health Outcomes (Wave IV) Depression 13.0 12.4 16.1 14.9 15.7 10.8 Sleep problems 11.4 11.1 11.5 10.3 10.7 14.7 Cohen perceived stress score tertile  Low (0–3) 39.8 38.3 40.4 41.0 29.9 29.5  Medium (4–6) 37.5 38.8 30.1 29.0 40.6 45.3  High (7–18) 22.6 22.9 29.5 30.0 29.5 25.2 Abbreviation: PS, propensity score. a Percentages, means, and standard errors were calculated by accounting for survey weights, strata, and clusters. b Values are expressed as mean (standard error). PS distributions revealed 659 individuals with PS in regions where not all exposure levels were represented; an additional 3,080 individuals were below the fifth or above the 95th percentile of 1 or more PS distribution (Web Figures 2A–C, 3A–C). The PS-restricted analytic sample was thus reduced to 2,815 women and 2,586 men (Web Appendix 5). Compared with the full sample, women in the PS-restricted sample (Table 1) had parents with lower educational attainment (across all occupation types: 24.2% graduated college in PS-restricted sample vs. 32.1% in full sample) and household incomes (across all occupation types: $42,000 vs. $47,000), and had lower educational attainment themselves (across all occupation types 16.4% graduated college vs. 34.9%). Men in the PS-restricted sample similarly had lower educational attainment (Table 2). The following variables remained unbalanced after analytic sample restriction and were included as covariates in the multivariable models: participant educational attainment (for women and men), race (women only), and parental educational attainment (women only). Occupation characteristics in the PS-restricted sample The top 4 major occupation categories were as follows: sales and related, office and administrative support, food preparation and serving related, and construction and extraction occupations (Web Table 3); 60% of tipped workers were waiters, waitresses, and bartenders (Web Table 4). Job characteristics, such as shift type and access to paid leave, varied by broad occupation type (e.g., nonservice, untipped service, tipped service) (Web Table 5). Multivariable models Women in tipped service work had 61% higher odds of reporting depression diagnoses or symptoms relative to women in nonservice work (95% confidence interval: 1.11, 2.34) (Table 3). The association between untipped service work (vs. nonservice work) and reported depression diagnosis or symptoms was weaker and not significant (odds ratio = 1.25, 95% confidence interval: 0.93, 1.68). Associations for sleep problems and higher perceived stress tertile were not significant but of similar magnitude and direction. Although associations with depression, sleep problems, and perceived stress were not significant, they were all of greater magnitude for women in tipped relative to untipped occupations in women. Men exhibited an association similar to that seen in women for perceived stress, though it was weaker and not statistically significant. Table 3. Mental Health Outcomes Regressed on Employment Category in Women and Men Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Gender and Occupation Type Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI Womena  Nonservice 1.00 1.00 1.00  Untipped service 1.25 0.93, 1.68 1.38 0.94, 2.03 1.13 0.88, 1.44  Tipped service 1.61 1.11, 2.34 1.49 0.98, 2.24 1.32 0.95, 1.84 Menb  Nonservice 1.00 1.00 1.00  Untipped service 1.23 0.81, 1.88 0.86 0.50, 1.49 1.10 0.78, 1.55  Tipped service 0.82 0.29, 2.31 1.26 0.50, 3.22 1.24 0.73, 2.11 Gender and Occupation Type Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI Womena  Nonservice 1.00 1.00 1.00  Untipped service 1.25 0.93, 1.68 1.38 0.94, 2.03 1.13 0.88, 1.44  Tipped service 1.61 1.11, 2.34 1.49 0.98, 2.24 1.32 0.95, 1.84 Menb  Nonservice 1.00 1.00 1.00  Untipped service 1.23 0.81, 1.88 0.86 0.50, 1.49 1.10 0.78, 1.55  Tipped service 0.82 0.29, 2.31 1.26 0.50, 3.22 1.24 0.73, 2.11 Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; OR, odds ratio; PS, propensity score. a The PS-restricted sample (n = 2,815) : Overlapping asymmetrically trimmed propensity distributions adjusted for PS, race, parental educational attainment, and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. b PS-restricted sample (n = 2,586): Overlapping asymmetrically trimmed propensity distributions adjusted for PS and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. Table 3. Mental Health Outcomes Regressed on Employment Category in Women and Men Who Reported a Current or Recent Job During the Wave IV Interview, National Longitudinal Study of Adolescent Health, 1994–2008 Gender and Occupation Type Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI Womena  Nonservice 1.00 1.00 1.00  Untipped service 1.25 0.93, 1.68 1.38 0.94, 2.03 1.13 0.88, 1.44  Tipped service 1.61 1.11, 2.34 1.49 0.98, 2.24 1.32 0.95, 1.84 Menb  Nonservice 1.00 1.00 1.00  Untipped service 1.23 0.81, 1.88 0.86 0.50, 1.49 1.10 0.78, 1.55  Tipped service 0.82 0.29, 2.31 1.26 0.50, 3.22 1.24 0.73, 2.11 Gender and Occupation Type Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI Womena  Nonservice 1.00 1.00 1.00  Untipped service 1.25 0.93, 1.68 1.38 0.94, 2.03 1.13 0.88, 1.44  Tipped service 1.61 1.11, 2.34 1.49 0.98, 2.24 1.32 0.95, 1.84 Menb  Nonservice 1.00 1.00 1.00  Untipped service 1.23 0.81, 1.88 0.86 0.50, 1.49 1.10 0.78, 1.55  Tipped service 0.82 0.29, 2.31 1.26 0.50, 3.22 1.24 0.73, 2.11 Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; OR, odds ratio; PS, propensity score. a The PS-restricted sample (n = 2,815) : Overlapping asymmetrically trimmed propensity distributions adjusted for PS, race, parental educational attainment, and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. b PS-restricted sample (n = 2,586): Overlapping asymmetrically trimmed propensity distributions adjusted for PS and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. Sensitivity analyses Associations were stronger in samples restricted to women with no previous history of depression (for untipped service, odds ratio = 1.60; for tipped service, odds ratio = 2.98) (Table 4). Similarly, when analysis was restricted to women with no previous sleep problems, women in untipped service occupations had 72% higher odds of reporting sleep problems than did women in nonservice occupations. Stronger associations for depression or sleep problems were not observed in men. Associations for women and men were similar but statistically nonsignificant after sample restriction to full-time workers. Estimates obtained using the full analytic sample with logistic regression covariate adjustment were comparable in direction and magnitude (Web Table 6). Table 4. Mental Health Outcomes Regressed on Employment Category in Women and Men Who Reported a Current or Recent Job During the Wave IV Interview: Sensitivity Analyses, National Longitudinal Study of Adolescent Health, 1994–2008 Sensitivity Analysis Subgroup and Occupation Type Womena Menb Depression Sleep Problems Higher Perceived Stress Tertile Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI No previous depressionc  Nonservice 1.00 1.00  Untipped service 1.60 1.04, 2.46 1.00 0.58, 1.71  Tipped service 2.98 1.55, 5.70d 1.33 0.40, 4.44 No previous sleep problemse  Nonservice 1.00 1.00  Untipped service 1.72 1.12, 2.66 0.73 0.39, 1.35  Tipped service 1.42 0.75, 2.71 1.12 0.43, 2.88 Full-time workersf  Nonservice 1.00 1.00 1.00 1.00 1.00 1.00  Untipped service 1.35 0.96, 1.90 1.38 0.84, 2.27 1.03 0.76, 1.39 1.37 0.87, 2.16 0.84 0.47, 1.50 1.12 0.77, 1.63  Tipped service 1.49 0.91, 2.44 1.56 0.91, 2.67 1.12 0.70, 1.80 1.52 0.50, 4.62 1.61 0.51, 5.07 1.41 0.67, 2.93 Sensitivity Analysis Subgroup and Occupation Type Womena Menb Depression Sleep Problems Higher Perceived Stress Tertile Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI No previous depressionc  Nonservice 1.00 1.00  Untipped service 1.60 1.04, 2.46 1.00 0.58, 1.71  Tipped service 2.98 1.55, 5.70d 1.33 0.40, 4.44 No previous sleep problemse  Nonservice 1.00 1.00  Untipped service 1.72 1.12, 2.66 0.73 0.39, 1.35  Tipped service 1.42 0.75, 2.71 1.12 0.43, 2.88 Full-time workersf  Nonservice 1.00 1.00 1.00 1.00 1.00 1.00  Untipped service 1.35 0.96, 1.90 1.38 0.84, 2.27 1.03 0.76, 1.39 1.37 0.87, 2.16 0.84 0.47, 1.50 1.12 0.77, 1.63  Tipped service 1.49 0.91, 2.44 1.56 0.91, 2.67 1.12 0.70, 1.80 1.52 0.50, 4.62 1.61 0.51, 5.07 1.41 0.67, 2.93 Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; OR, odds ratio; PS, propensity score. a Overlapping asymmetrically trimmed propensity distributions adjusted for PS, race, parental educational attainment, and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. b Overlapping asymmetrically trimmed propensity distributions adjusted for PS and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. c The category “no previous depression” was defined as childhood (waves I and II) maximum CES-D score ≤3; n = 1,223 women and n = 1,552 men. d Confidence intervals for tipped service workers differed significantly from those of untipped service workers. e The category “no previous sleep problems” was defined as either never having difficulty falling or staying asleep or having difficulty “just a few times” during childhood (wave II); n = 916 women and n = 997 men. f Full time work was defined as ≥35 hours/week; n = 2,209 women and n = 2,338 men. Table 4. Mental Health Outcomes Regressed on Employment Category in Women and Men Who Reported a Current or Recent Job During the Wave IV Interview: Sensitivity Analyses, National Longitudinal Study of Adolescent Health, 1994–2008 Sensitivity Analysis Subgroup and Occupation Type Womena Menb Depression Sleep Problems Higher Perceived Stress Tertile Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI No previous depressionc  Nonservice 1.00 1.00  Untipped service 1.60 1.04, 2.46 1.00 0.58, 1.71  Tipped service 2.98 1.55, 5.70d 1.33 0.40, 4.44 No previous sleep problemse  Nonservice 1.00 1.00  Untipped service 1.72 1.12, 2.66 0.73 0.39, 1.35  Tipped service 1.42 0.75, 2.71 1.12 0.43, 2.88 Full-time workersf  Nonservice 1.00 1.00 1.00 1.00 1.00 1.00  Untipped service 1.35 0.96, 1.90 1.38 0.84, 2.27 1.03 0.76, 1.39 1.37 0.87, 2.16 0.84 0.47, 1.50 1.12 0.77, 1.63  Tipped service 1.49 0.91, 2.44 1.56 0.91, 2.67 1.12 0.70, 1.80 1.52 0.50, 4.62 1.61 0.51, 5.07 1.41 0.67, 2.93 Sensitivity Analysis Subgroup and Occupation Type Womena Menb Depression Sleep Problems Higher Perceived Stress Tertile Depression Sleep Problems Higher Perceived Stress Tertile OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI No previous depressionc  Nonservice 1.00 1.00  Untipped service 1.60 1.04, 2.46 1.00 0.58, 1.71  Tipped service 2.98 1.55, 5.70d 1.33 0.40, 4.44 No previous sleep problemse  Nonservice 1.00 1.00  Untipped service 1.72 1.12, 2.66 0.73 0.39, 1.35  Tipped service 1.42 0.75, 2.71 1.12 0.43, 2.88 Full-time workersf  Nonservice 1.00 1.00 1.00 1.00 1.00 1.00  Untipped service 1.35 0.96, 1.90 1.38 0.84, 2.27 1.03 0.76, 1.39 1.37 0.87, 2.16 0.84 0.47, 1.50 1.12 0.77, 1.63  Tipped service 1.49 0.91, 2.44 1.56 0.91, 2.67 1.12 0.70, 1.80 1.52 0.50, 4.62 1.61 0.51, 5.07 1.41 0.67, 2.93 Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; OR, odds ratio; PS, propensity score. a Overlapping asymmetrically trimmed propensity distributions adjusted for PS, race, parental educational attainment, and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. b Overlapping asymmetrically trimmed propensity distributions adjusted for PS and participant educational attainment. PS include the following variables: race; whether born in the United States; highest level of education attained; parent’s highest level of education; and wave I household income, childhood maltreatment, incarcerated parent, maximum childhood CES-D score, childhood smoking history, childhood AUDIT-C score, childhood general health, childhood sleep, rolling average body mass index (weight (kg)/weight (m)2), and childhood physical activity. c The category “no previous depression” was defined as childhood (waves I and II) maximum CES-D score ≤3; n = 1,223 women and n = 1,552 men. d Confidence intervals for tipped service workers differed significantly from those of untipped service workers. e The category “no previous sleep problems” was defined as either never having difficulty falling or staying asleep or having difficulty “just a few times” during childhood (wave II); n = 916 women and n = 997 men. f Full time work was defined as ≥35 hours/week; n = 2,209 women and n = 2,338 men. DISCUSSION We investigated the association between occupation type and 3 adverse mental health outcomes within a nationally representative cohort of adolescents followed into adulthood. We observed cross-sectional associations between working in service occupations and poor mental health outcomes in women and men. Although only 1 of the examined associations was statistically significant—women in tipped service work had greater odds of reporting depression than did women in nonservice work—the magnitudes of associations were consistently highest among women in tipped service occupations. In men, associations were weaker and not statistically significant; these findings were confirmed in sensitivity analyses. Our observation that service work was positively associated with depression is consistent with findings of previous analyses in which workers in the service industry were identified as having the highest prevalence of depression and the role of interpersonal conflict and encounters with difficult people were highlighted (18, 34). Observed associations with adverse mental health outcomes may further reflect the precarious nature of service work, which often entails lack of access to health-promoting benefits (5–7), low wages (35), and last-minute scheduling practices (3, 5). Associations with adverse mental health outcomes are also consistent with research in hotel employees, a subset of service workers that includes tipped and untipped workers in which job-related factors like low control, high psychological demands, and atypical work schedules are associated with a higher burden of morbidity (36–38). Although observed associations for all outcomes were only significant for self-reported depression, they were of greater magnitude for tipped relative to untipped service work in women. In analyses restricted to women with no previous history of depression, women in tipped service occupations had greater odds of reporting depressive symptoms or a depression diagnosis than women in untipped service. This finding may reflect characteristics of tipped work that make it more precarious than untipped work, such as more unstable income (7) and greater emotional labor demands (39, 40). In a bivariable examination of job characteristics by occupation type, we found that a smaller proportion of study participants in tipped service had access to paid leave, health insurance, regular shift schedules, or freedom to make important decisions, compared with those in untipped service and nonservice occupations. Higher odds of depression were not observed for men in tipped service. Women in tipped service occupations have more unstable income and earn less than men in tipped service occupations (wave IV household income: $48,500 for women vs. $60,000 for men). Nationwide, women in tipped occupations earn 6% less per hour than men (7). Part of this discrepancy is attributed to further gender-based occupational stratification. For instance, in our PS-restricted sample, women in tipped service occupations were largely restaurant wait staff (44.3% of all women tipped service workers vs. 30.8% of all men tipped service workers). However, even among wait staff, women are less likely than their male counterparts to work in fine dining establishments (41). Moreover, tipping practices are discriminatory. In 1 study, women were found to only have earned equivalent tips to men when their service was rated by the customer as “exceptional,” suggesting that women are being held to higher standards, especially by male customers (42). In another study, researchers found that male customers tipped more favorably if they found the female server attractive and/or she was wearing makeup (43). Discrepancies in the occupation category–depression association may reflect gender-based differential exposure to the discriminatory aspects of tipped service work. Researchers have also observed that the association between various psychosocial work exposures and poor mental health may differ by gender (44). In service occupations in particular, having to manage challenging customers may undermine gender-role authenticity, with detrimental mental health effects (45). As such, women may experience differential vulnerability to the emotional demands inherent to tipped and untipped service work. Gender-based differences in observed associations may also be a product of gender differences in stratification into specific occupation types within the 3 categories we examined. Among the PS-restricted sample (which largely excluded professionals), 68% of the nonservice occupations filled by women were administrative, whereas 64% of the nonservice occupations filled by men were blue-collar physical labor–oriented occupations, with different job types providing a different constellation of physical, psychosocial, and environmental exposures. Our observations are consistent with data from the Department of Labor on gender segregation in the workforce (46) and may partially explain some of the weaker associations observed with the other outcomes in men. Our analysis has limitations. First, because we restricted our analytic sample, generalizability is limited to individuals from lower- to middle-class upbringings without college degrees. However, use of PS enhanced internal validity. Although the unrestricted analysis yielded estimates that are nationally representative for adolescents followed into adulthood, these estimates were computed in a sample that contained individuals who were not exchangeable, due to social stratification, as evidenced by the nonoverlapping propensity distributions observed in Web Figure 2A. We further argue that given the bimodal distribution of the propensity for nonservice among women in service occupations, our sample restriction likely removed affluent, atypical individuals. For instance, women entering service work despite having a high propensity for nonservice work may be entering outlier occupations (e.g., fine dining establishments, high-end salons) and/or be selecting this type of work for the flexibility it affords. Second, our estimates may be biased due to unmeasured confounding, because PS only balance measured variables. We posit that estimates observed from our restricted sample are more conservative than those in unrestricted analysis to account for unmeasured confounding that may make an individual with a high propensity for entering 1 occupation still enter another. Third, exposure and outcomes are subject to misclassification error. Although it is likely individuals were appropriately classified as being in service-industry professions or not, some degree of nondifferential misclassification is expected on determination of whether the occupation was tipped. Notably, specific occupations for those reporting tipped service work were predominantly occupations that are less ambiguous in regard to tipping status in the United States. Here, we expected estimates would be biased toward the null. Also, self-report of study outcomes may introduce differential misclassification. The symptom-recall periods for measures of perceived stress and sleep problems were 1 month, whereas the recall period for depressive symptoms was 1 week. Researchers have observed a systematic bias in recall that is largest for those asked to reflect on a longer time (47). Regardless, the consistency of the magnitude and direction of the observed associations for these 3 mechanistically interconnected outcomes irrespective of differences in recall periods lends credibility to our observed associations. Furthermore, we leveraged the use of a prospective cohort initiated in childhood, which allowed us to minimize recall bias and ensure temporality of the items included in the study. Notably, our sensitivity analysis restricted to individuals with no history of depression yielded larger point estimates among tipped service workers. We posit that because individuals in untipped service work experienced disproportionately more childhood adversity relative to nonservice and tipped service workers and because of the relationship between childhood adversity and childhood mental health (48), this sensitivity analysis addressed further residual confounding related to childhood disadvantage. Fourth, 51% of the original Add Health sample did not complete all 4 waves and of those that did, 34% were missing 1 or more measure pertinent to PS development. We mitigated the potential introduction of selection bias by including sampling weights that accounted for loss to follow-up and multiply imputing in conjunction with PS development to address missing covariate data. Although there is concern that certain variables were not missing randomly, such as responses to childhood maltreatment questions, it is likely that any bias introduced would bias estimates to the null. Fifth, we were unable to perform risk-estimation procedures. Outcomes evaluated in this analysis are all prevalent in this population; however, we expected direction and magnitude of the associations to remain largely consistent, and this was observed for models we were able to evaluate (Web Table 6). Sixth, the incorporation of PS in our analysis was limited in that within our statistical software, we were unable to incorporate standard errors that accounted for multiple imputation. In addition, the 3-category exposure variable coupled with the complex survey design of the Add Health data limited our ability to incorporate PS beyond simple adjustment. Different approaches to the incorporation of PS in analyses can yield different results; however, our use of asymmetric trimming before performing analyses enabled us to procure estimates that are likely more similar to those that would be obtained through other PS methods (49). Finally, our analyses may have been underpowered to detect further significant differences between occupation categories due, in part, to small sample sizes, particularly among tipped service workers (50). We conclude that the heightened precariousness of tipped service work may place individuals in these occupations, especially women, at increased risk of poor mental health. Optimal public policy and employment practices to alleviate the excess risk of depression in tipped service workers will depend on understanding the factors that underlie these differences in health status. ACKNOWLEDGMENTS Author affiliations: OHSU-PSU School of Public Health, Oregon Health & Science University, Portland, Oregon (Sarah B. Andrea, Lynne C. Messer, Miguel Marino, Janne Boone-Heinonen); and Department of Family Medicine, Oregon Health & Science University, Portland, Oregon (Miguel Marino). The project described was supported by the Office of Research in Women’s Health and the National Institute of Child Health and Human Development, Oregon Building Interdisciplinary Research Careers in Women’s Health grant K12HD043488 (J.B.-H.) and National Institute of Digestive Disorders and Nutrition grant K01DK102857 (J.B.-H.). For this research, we used data from the National Longitudinal Study of Adolescent to Adult Health, a program project directed by Dr. Kathleen Mullan Harris and designed by Dr. J. Richard Udry, Dr. Peter S. Bearman, and Dr. Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Dr. Ronald R. Rindfuss and Dr. Barbara Entwisle for assistance in the original design. No direct support was received from grant P01-HD31921 for this analysis. Components of this work were presented by S.B.A. at the American Heart Association EPI Lifestyle Scientific Sessions on March 9, 2017, Portland, Oregon; and the Society for Epidemiologic Research Annual Meeting on June 22, 2017, Seattle, Washington. Conflict of interest: none declared. Abbreviations Add Health National Longitudinal Study of Adolescent to Adult Health PS propensity score. REFERENCES 1 Burgard SA , Lin KY . Bad jobs, bad health? How work and working conditions contribute to health disparities . Am Behav Sci . 2013 ; 57 ( 8 ): 1105 – 1127 . Google Scholar Crossref Search ADS 2 Vosko LF . Introduction. In: Managing the Margins: Gender, Citizenship, and the International Regulation of Precarious Employment . Oxford, United Kingdom : Oxford University Press ; 2010 : 1 – 34 . 3 Boushey H , Ansel B . Working by the hour: the economic consequences of unpredictable scheduling practices. Washington, DC: Washington Center for Equitable Growth; 2016 . http://equitablegrowth.org/research-analysis/working-by-the-hour-the-economic-consequences-of-unpredictable-scheduling-practices/. Accessed March 7, 2017. 4 Scott EK , King MC , Reddy R . The Impact on Oregonians of the Rise of Irregular Scheduling. 2017 . http://sociology.uoregon.edu/wp-content/uploads/sites/10/2017/02/Unpredict-Sched-Final.pdf. Accessed November 1, 2017. 5 McVeigh K . US employers slashing worker hours to avoid Obamacare insurance mandate. The Guardian. 2013 ; https://www.theguardian.com/world/2013/sep/30/us-employers-slash-hours-avoid-obamacare. Accessed April 2, 2017. 6 Perry-Jenkins M , Smith JZ , Wadsworth LP , et al. . Workplace policies and mental health among working-class, new parents . Community Work Fam . 2017 ; 20 ( 2 ): 226 – 249 . Google Scholar Crossref Search ADS PubMed 7 Allegretto S , Cooper D . Twenty-Three Years and Still Waiting for Change: Why It’s Time to Give Tipped Workers the Regular Minimum Wage. Economic Policy Institute. 2014 . http://www.epi.org/publication/waiting-for-change-tipped-minimum-wage/. Accessed January 30, 2017. 8 US Bureau of Labor statistics . Characteristics of Minimum Wage Workers, 2016. 2017 . https://www.bls.gov/opub/reports/minimum-wage/2016/home.htm. Accessed July 18, 2017. 9 US Department of Labor, Wage and Hour Division . Minimum wages for tipped employees. 2017 ; https://www.dol.gov/whd/state/tipped.htm. Accessed March 7, 2017. 10 Adelmann PK . Emotional labor as a potential source of job stress. In: Sauter SL , Murphy LR , eds. Organizational Risk Factors for Job Stress . Washington, DC : American Psychological Association ; 1995 : 371 – 381 . 11 Leidner R . Working on people. In: Fast Food, Fast Talk: Service Work and the Routinization of Everyday Life . Berkeley, CA : University of California Press ; 1993 : 1 – 23 . 12 Sharma U , Black P . Look good, feel better: beauty therapy as emotional labour . Sociology . 2001 ; 35 ( 4 ): 913 – 931 . Google Scholar Crossref Search ADS 13 Andrews BK , Karcz S , Rosenberg B . Hooked on a feeling: emotional labor as an occupational hazard of the post-industrial age . New Solut . 2008 ; 18 ( 2 ): 245 – 255 . Google Scholar Crossref Search ADS PubMed 14 de Castro AB , Agnew J , Fitzgerald ST . Emotional labor: relevant theory for occupational health practice in post-industrial America . AAOHN J . 2004 ; 52 ( 3 ): 109 – 115 . Google Scholar PubMed 15 Matulewicz K . Law and the construction of institutionalized sexual harassment in restaurants . Can J Law Soc . 2015 ; 30 ( 3 ): 401 – 419 . Google Scholar Crossref Search ADS 16 Fisk GM , Neville LB . Effects of customer entitlement on service workers’ physical and psychological well-being: a study of waitstaff employees . J Occup Health Psychol . 2011 ; 16 ( 4 ): 391 – 405 . Google Scholar Crossref Search ADS PubMed 17 Chirico F . The forgotten realm of the new and emerging psychosocial risk factors . J Occup Health . 2017 ; 59 ( 5 ): 433 – 435 . Google Scholar Crossref Search ADS PubMed 18 National Technical Reports Library . Depression among adults employed full-time, by occupational category. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2007 . https://ntrl.ntis.gov/NTRL/dashboard/searchResults/titleDetail/PB2009115316.xhtml. Accessed August 1, 2017. 19 Shockey TM , Wheaton AG . Short sleep duration by occupation group - 29 states, 2013–2014 . MMWR Morb Mortal Wkly Rep . 2017 ; 66 ( 8 ): 207 – 213 . Google Scholar Crossref Search ADS PubMed 20 Aronsson G , Dallner M , Lindh T , et al. . Flexible pay but fixed expenses: personal financial strain among on-call employees . Int J Health Serv . 2005 ; 35 ( 3 ): 499 – 528 . Google Scholar Crossref Search ADS PubMed 21 Zhu J , Tews MJ , Stafford K , et al. . Alcohol and illicit substance use in the food service industry: assessing self-selection and job-related risk factors . J Hosp Tour Res . 2011 ; 35 ( 1 ): 45 – 63 . Google Scholar Crossref Search ADS 22 National Longitudinal Study of Adolescent to Adult Health . Add Health. http://www.cpc.unc.edu/projects/addhealth/. Accessed January 19, 2017. 23 National Longitudinal Study of Adolescent to Adult Health . Study design. http://www.cpc.unc.edu/projects/addhealth/design. Accessed February 1, 2017. 24 US Bureau of Labor Statistics . Overview of BLS statistics by industry. https://www.bls.gov/bls/industry.htm. Accessed February 2, 2017. 25 National Longitudinal Study of Adolescent to Adult Health . Restricted-use contractual data. http://www.cpc.unc.edu/projects/addhealth/documentation/restricteduse. Accessed January 11, 2017. 26 Dr. Cohen’s Scales. http://www.psy.cmu.edu/~scohen/scales.html. Updated February 19, 2015. Accessed January 24, 2017. 27 Chen P . Appropriate analysis in add health correcting for design effects & selecting weights. Presented at the 2014 Add Health Users Conference, June 26 and 27, 2014 , Bethesda, MD. http://www.cpc.unc.edu/projects/addhealth/events/2014-add-health-users-conference-appropriate-analysis-ping-chen. Accessed July 1, 2017. 28 Cartwright B , Edwards R , Wang Q . Job and industry gender segregation: NAICS categories and EEO-1 job groups . Mon Labor Rev 2011 ; 134 ( 11 ): 37 – 50 . 29 US Bureau of Labor Statistics . Labor force statistics from the current population survey. Household data annual averages: 11. Employed persons by detailed occupation, sex, race, and Hispanic or Latino ethnicity. https://www.bls.gov/cps/cpsaat11.htm. Accessed July 17, 2017. 30 Oakes JM , Andrade KE . The measurement of socioeconomic status. In: Oakes JM , Kaufman JS , eds. Methods in Social Epidemiology . San Francisco, CA : John Wiley & Sons ; 2006 : 23 – 42 . 31 Stürmer T , Rothman KJ , Avorn J , et al. . Treatment effects in the presence of unmeasured confounding: dealing with observations in the tails of the propensity score distribution—a simulation study . Am J Epidemiol . 2010 ; 172 ( 7 ): 843 – 854 . Google Scholar Crossref Search ADS PubMed 32 Lewinsohn PM , Zeiss AM , Duncan EM . Probability of relapse after recovery from an episode of depression . J Abnorm Psychol . 1989 ; 98 ( 2 ): 107 – 116 . Google Scholar Crossref Search ADS PubMed 33 Dooley D , Prause J , Ham-Rowbottom KA . Underemployment and depression: longitudinal relationships . J Health Soc Behav . 2000 ; 41 ( 4 ): 421 – 436 . Google Scholar Crossref Search ADS PubMed 34 Wulsin L , Alterman T , Timothy Bushnell P , et al. . Prevalence rates for depression by industry: a claims database analysis . Soc Psychiatry Psychiatr Epidemiol . 2014 ; 49 ( 11 ): 1805 – 1821 . Google Scholar Crossref Search ADS PubMed 35 US Bureau of Labor Statistics . Occupational employment statistics: May 2016 national occupational employment and wage estimates, United States. https://www.bls.gov/oes/2016/may/oes_nat.htm. Accessed July 18, 2017. 36 Hannerz H , Tüchsen F , Kristensen TS . Hospitalizations among employees in the Danish hotel and restaurant industry . Eur J Public Health . 2002 ; 12 ( 3 ): 192 – 197 . Google Scholar Crossref Search ADS PubMed 37 Chiang FFT , Birtch TA , Kwan HK . The moderating roles of job control and work-life balance practices on employee stress in the hotel and catering industry . Int J Hosp Manag . 2010 ; 29 ( 1 ): 25 – 32 . Google Scholar Crossref Search ADS 38 McNamara M , Bohle P , Quinlan M . Precarious employment, working hours, work-life conflict and health in hotel work . Appl Ergon . 2011 ; 42 ( 2 ): 225 – 232 . Google Scholar Crossref Search ADS PubMed 39 Azar OH . The social norm of tipping: a review . J Appl Soc Psychol . 2007 ; 37 ( 2 ): 380 – 402 . Google Scholar Crossref Search ADS 40 Jung HS , Yoon HH . Antecedents and consequences of employees’ job stress in a foodservice industry: focused on emotional labor and turnover intent . Int J Hosp Manag . 2014 ; 38 : 84 – 88 . Google Scholar Crossref Search ADS 41 Johnson S . Tipped workers hope for hike in sub-minimum wage. WeNews. February 5, 2012 . http://womensenews.org/2012/02/tipped-workers-hope-hike-in-sub-minimum-wage/. Accessed July 20, 2017. 42 Parrett M . Customer discrimination in restaurants: dining frequency matters . J Labor Res . 2011 ; 32 ( 2 ): 87 – 112 . Google Scholar Crossref Search ADS 43 Guéguen N , Jacob C . Enhanced female attractiveness with use of cosmetics and male tipping behavior in restaurants . J Cosmet Sci . 2011 ; 62 ( 3 ): 283 – 290 . Google Scholar PubMed 44 Wieclaw J , Agerbo E , Mortensen PB , et al. . Psychosocial working conditions and the risk of depression and anxiety disorders in the Danish workforce . BMC Public Health . 2008 ; 8 : 280 . Google Scholar Crossref Search ADS PubMed 45 Johnson HA , Spector PE . Service with a smile: do emotional intelligence, gender, and autonomy moderate the emotional labor process? J Occup Health Psychol . 2007 ; 12 ( 4 ): 319 – 333 . Google Scholar Crossref Search ADS PubMed 46 US Department of Labor, Women’s Bureau . Occupations by gender shares of employment. https://www.dol.gov/wb/stats/occ_gender_share_em_1020_txt.htm. Accessed October 20, 2017. 47 Tasman A , Kay J , Lieberman JA , et al. . eds. Social psychology: theory, research, and mental health implications. In: Psychiatry . West Sussex, United Kingdom : John Wiley & Sons ; 2015 : 452 – 462 . 48 Fitzsimons E , Goodman A , Kelly E , et al. . Poverty dynamics and parental mental health: determinants of childhood mental health in the UK . Soc Sci Med . 2017 ; 175 : 43 – 51 . Google Scholar Crossref Search ADS PubMed 49 Kurth T , Walker AM , Glynn RJ , et al. . Results of multivariable logistic regression, propensity matching, propensity adjustment, and propensity-based weighting under conditions of nonuniform effect . Am J Epidemiol . 2006 ; 163 ( 3 ): 262 – 270 . Google Scholar Crossref Search ADS PubMed 50 Stang A , Poole C , Kuss O . The ongoing tyranny of statistical significance testing in biomedical research . Eur J Epidemiol . 2010 ; 25 ( 4 ): 225 – 230 . Google Scholar Crossref Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Yiqiang, Zhan,;Xing-Rong, Liu,;A, Reynolds, Chandra;L, Pedersen, Nancy;Sara, Hägg,;S, Clements, Mark

2018 American Journal of Epidemiology

doi: 10.1093/aje/kwy128pmid: 29961868

Abstract Although previous studies examining leukocyte telomere length (LTL) and all-cause mortality controlled for several confounders, the observed association could be biased due to unmeasured confounders, including familial factors. We aimed to examine the association of LTL with all-cause mortality in a Swedish twin sample while adjusting for familial factors and allowing for time-dependent effects. A total of 366 participants (174 twin pairs and 18 individuals) were recruited from the Swedish Twin Registry. LTL was assessed using the Southern blot method. All-cause mortality data were obtained through linkage with the Swedish Population Registry, updated through November 15, 2017. To control for familial factors within twin pairs, we applied a between-within shared frailty model based on generalized survival models. Overall, 115 (31.4%) participants were men and 251 (68.6%) were women. The average age of the study participants when blood was drawn was 79.1 years, and follow-up duration ranged from 10 days to 25.7 years (mean = 10.2 years). During the follow-up period, 341 (93.2%) participants died. Shorter LTL was associated with higher mortality rates when controlling for familial factors in the between-within shared frailty model. We found significant time-dependent effects of LTL on all-cause mortality, where the mortality rate ratios were attenuated with increasing age. between-within model, generalized survival model, mortality, shared frailty, telomere Telomeres—repeat sequences of nucleotides at the ends of chromosomes—maintain the structural integrity and stability of chromosomes (1). Each time a cell divides, the telomeres become shorter, as some nucleotides at the ends of the telomeres cannot be replicated. When telomeres reach a critically short length, cells are unable to replenish or divide and enter the senescence stage (2). Cellular senescence has been implicated in the aging process and in aging-related diseases (3). Previous studies have found that short leukocyte telomere length (LTL) is associated with a number of aging-related traits, including cardiovascular and metabolic diseases (4, 5), cognitive dec