Associations between maternal lifestyle factors and neonatal body composition in the Screening for Pregnancy Endpoints (Cork) cohort study

Associations between maternal lifestyle factors and neonatal body composition in the Screening... Abstract Background Neonatal body composition likely mediates fetal influences on life long chronic disease risk. A better understanding of how maternal lifestyle is related to newborn body composition could thus inform intervention efforts. Methods Using Cork participant data (n = 1754) from the Screening for Pregnancy Endpoints (SCOPE) cohort study [ECM5(10)05/02/08], we estimated how pre-pregnancy body size, gestational weight gain, exercise, alcohol, smoking and diet were related to neonatal fat and fat-free mass, as well as length and gestational age at birth, using quantile regression. Maternal factors were measured by a trained research midwife at 15 gestational weeks, in addition to a 3rd trimester weight measurement used to calculate weight gain. Infant body composition was measured using air-displacement plethysmography. Results Healthy (versus excess) gestational weight gain was associated with lower median fat-free mass [−112 g, 95% confidence interval (CI): −47 to −176) and fat mass (−33 g, 95% CI: −1 to −65) in the offspring; and a 103 g decrease in the 95th centile of fat mass (95% CI: −33 to −174). Maternal normal weight status (versus obesity) was associated with lower median fat mass (−48 g, 95% CI: −12 to −84). At the highest centiles, fat mass was lower among infants of women who engaged in frequent moderate-intensity exercise early in the pregnancy (−92 g at the 95th centile, 95% CI: −168 to −16). Lastly, women who never smoked tended to have longer babies with more fat mass and fat-free mass. No other lifestyle factors were strongly related to infant body composition. Conclusions These results suggest that supporting healthy maternal lifestyles could reduce the risk of excess fat accumulation in the offspring, without adversely affecting fat-free mass development, length or gestational age. Developmental origins, body composition, lifestyle, birth cohort, quantile regression Key Messages Maternal body size and gestational weight gain were most strongly associated with newborn body composition. Associations between maternal lifestyle and newborn body composition were often stronger in the tails of the outcome’s distribution. Quantile regression provided useful insights that would not have been apparent with linear or logistic regression. Introduction The environment experienced by a fetus during development influences its risk for cardiovascular disease, diabetes and obesity across the life course.1–4 Healthy maternal lifestyles before and during pregnancy may affect fetal development and thus influence these outcomes,5,6 but our understanding is limited since many birth cohort studies capable of linking maternal factors to cardiometabolic risks in the offspring are not yet into middle age.7,8 However, newborn body composition is a useful reflection of the fetal environment in this context,9,10 and likely mediates developmental effects on long-term cardiometabolic outcomes.7,11,12 Thus a better understanding of how maternal lifestyle is related to newborn body composition could provide important insights into the developmental origins of adult health and disease. Some of the challenges in measuring fat and fat-free mass in infants have recently lessened,13 facilitating relatively large population-based studies of infant body composition.14–18 Another challenge is how to best model the impact of one or more predictors on a single continuous anthropometric measure where we expect distinct aetiological mechanisms at opposite ends of the measure’s distribution (e.g. the causes of low birthweight versus those of macrosomia). This is typically approximated by categorizing the otherwise continuous variable and investigating it using binary or multinomial logistic regression. This approach is useful to the degree that the categorisation is clinically relevant, but it ignores the continuous nature of the variable and inevitably discards useful information about the rank order of individuals in the sample. To avoid this, we used quantile regression19 to estimate associations between maternal lifestyle factors and offspring neonatal fat mass and fat-free mass in one of the largest relevant studies to date. Quantile regression is analogous to multiple linear regression, but instead of modelling the mean of the dependent variable conditional on the predictor(s), it can be used to similarly model any centile. Importantly, quantile regression allows you to estimate how the tails of a variable’s distribution (such as the 5th or 95th centile) vary across levels of a predictor, which can occur even when there is no obvious shift in the mean or median. This allowed us to test the hypothesis that infants born to mothers exhibiting healthier lifestyle factors would be less likely to have very high or low levels of fat and fat-free mass. Methods Ethical standards The authors assert that all procedures contributing to this work comply with the International Ethical Guidelines for Epidemiological Studies (CIOMS/WHO) and with the Helsinki Declaration of 1975, as revised in 2008, and has been approved by the Research Ethics Committee of the Cork Teaching Hospitals provided for the SCOPE-Ireland study [ref: ECM5(10)05/02/08]. Study and sample Data are from the Cork site of the Screening for Pregnancy Endpoints (SCOPE) pregnancy cohort study (ACTRN12607000551493), and its follow-up birth cohort study, Babies after SCOPE: Evaluating the Longitudinal Impact on Neurological and Nutritional Endpoints (BASELINE; ClinicalTrials.gov NCT 01498965; www.birthcohorts.net). The primary aim of SCOPE was to identify clinical factors and biomarkers that were predictive of pre-eclampsia, small for gestational age babies and spontaneous preterm birth. Based on this aim, the study included healthy, nulliparous women with singleton pregnancies, and the exclusion criteria were: known major fetal anomalies; pre-pregnancy essential hypertension; moderate to severe hypertension at booking; pre-existing diabetes; renal disease; systemic lupus erythematosus; antiphospholipid syndrome; HIV positivity; major uterine anomaly; cervical suture; knife cone biopsy; ruptured membranes at recruitment; three or more miscarriages; three or more terminations; long-term steroid use; and treatment with low-dose aspirin, calcium (>1000 g/24 h) or Vitamin E (≥400 iu), low-molecular-weight heparin, fish oil or antioxidants. All women who participated in the SCOPE study were informed about the birth cohort, and if consent was obtained infants were registered to the Cork BASELINE birth cohort. Participants were recruited from Cork University Maternity Hospital between February 2008 and February 2011, at 15 ± 1 weeks of gestation. During this period, 2579 nulliparous women were invited to participate and 1774 (69%) consented to do so. Study participants were interviewed and examined by a trained research midwife at recruitment and again at 20 ± 1 weeks of gestation. Demographic, socioeconomic and medical data were collected, as well as information on current diet, physical activity and other lifestyle factors. Of these 1774 women, 1537 went on to have an infant enrolled into the BASELINE birth cohort (87%). Infant measures were taken by a trained research midwife within 72 h of birth. All data were managed using an internet accessible database with a clear audit trail and automated quality assurance procedures (MedSciNet AB, Stockholm, Sweden). Additional details on the study’s methods have been previously reported.20,21 Measurements and variable definitions Infant measures Infant measures used in this analysis were body composition, length, weight, gestational age at delivery and sex. Per study protocol, the aim was to measure body composition within 48 h. First, newborn body density was calculated using weight (measured with an electronic scale to the nearest gram), divided by newborn body volume (estimated by air-displacement plethysmography with a PEA POD® Infant Body Composition System, COSMED USA, Concord, CA), giving body density. Based on a two-compartment model of body composition (fat and fat-free mass) and body density values from Fomon,22 percent body fat was calculated, and was in turn used to estimate fat and fat-free mass in grams, which were the primary outcomes of interest. Gestational age was based on expected date of delivery, which was estimated from respondent recall of last menstrual period (LMP). If the respondent was uncertain about their LMP, or it differed substantially from a 16- or 20-week scan (by ≥7 or ≥10 days, respectively), the earliest available scan date was used. Length was measured with a neonatometer to the nearest millimetre. Maternal lifestyle factors The maternal lifestyle factors we considered were those that most closely aligned with current pregnancy recommendations regarding nutrition and lifestyle in Ireland.23 Pre-pregnancy weight (kg) was estimated as weight measured at recruitment, less 1.25 kg [the assumed average amount of weight gain in the first 15 weeks of pregnancy based on the 2009 Institute of Medicine (IOM) guidelines].24 Pre-pregnancy body mass index (BMI) was calculated from pre-pregnancy weight (kg) divided by measured height squared (m2). Pre-pregnancy weight status was subsequently categorised based on World Health Organization guidelines as underweight (BMI < 18.5 kg/m2), normal weight (BMI 18.5 to 25 kg/m2), overweight (BMI ≥ 25 to 30 kg/m2) and obese (BMI ≥ 30 kg/m2). In addition to the weight measurement at recruitment, weight was also typically measured multiple times across the pregnancy during routine care. To account for these differences, the weight gain rate (kg/week) was calculated as the difference between the last of these measures (whenever it occurred, usually late in the third trimester) and the initial weight measured at recruitment, divided by the number of weeks between those measures. Excessive gestational weight gain (GWG) was defined according to Institute of Medicine (IOM) guidelines as a weight gain rate exceeding 0.5 kg/week (wk) in underweight and normal weight women, 0.33 kg/wk in overweight women, and 0.27 kg/wk in obese women; and inadequate GWG was defined as a weight gain rate below 0.35 kg/wk in underweight and normal weight women, 0.23 kg/wk in overweight women, and 0.17 kg/wk in obese women.24,25 Women were asked at recruitment (15 ± 1 weeks of gestation) how many times each week they engaged in exercise that that did not result in heavier breathing, which was the study’s definition of moderate-intensity exercise. Their responses were categorized as Never, Some (1 to 3 times a week) and Often (4 or more times a week). Respondents also reported daily hours of television viewing in the past month, a commonly used marker of sedentary activity26 particularly in women,27 which was categorised as <2 h, 2 to 4 h and 5 or more h. Based on participant-reported consumption at 15 ± 1 and/or 20 ± 1 weeks of gestation, alcohol use was categorised as Never, Quit before Pregnancy, Quit during Pregnancy and Still Drinking; and smoking was categorised as Never Smoked, Quit during Pregnancy and Still Smoking. Women were asked about pre-pregnancy folic-acid supplementation, and their responses were dichotomized as those meeting the recommended 400 µg versus those who did not (Yes versus No). The questionnaire administered at recruitment asked women to report the frequency with which they consumed several food items in the first 15 weeks of pregnancy. Their responses were used to determine whether they were meeting the recommended five servings of fruit and veg per day (Yes versus No) and at least 1 serving of oily fish per week (Yes versus No). Covariates Available covariates likely to influence both maternal lifestyle and birth outcomes were selected based on the expert opinion of the study authors, who were also careful to not select covariates that were probable consequences of maternal lifestyle. All selected covariates were assessed at 15 ± 1 weeks’ gestation, and included: maternal age (years); the mother’s reported weight at birth (g); her gravidity (1 versus > 1); her ethnicity (White versus Non-white); whether she had a partner or not (Yes versus No); whether she had any third level education (Yes versus No); whether she used Public versus Private maternity care; her socioeconomic index (SEI), based on the New Zealand SEI28 (with higher values reflecting higher social status); her risk of depression based on the Edinburgh Postnatal Depression Score29,30 (Unlikely to experience depression versus At risk of depression in the next year versus Likely depressed); and her score on the Perceived Stress Scale31 (with higher scores reflecting greater stress). Statistical methods Categorical variables were described by the count and proportion in each category. Continuous variables were described by: their mean and standard deviation; their median and the interquartile range; and their full range.32 Relationships between maternal lifestyle factors and infant outcomes were estimated using quantile regression.19 We first estimated the crude association of each lifestyle factor with each outcome, at every centile from the 2nd to 98th. We then estimated a similar set of fully adjusted models with each outcome regressed on all the lifestyle factors and covariates. The latter were included to account for possible confounding. Based on previous research,33 we also tested for an interaction between maternal pre-pregnancy weight status and IOM-classified healthy GWG. All continuous covariates were centred at their means. Quantile regression coefficients were estimated using a modified version of the Barrodale and Roberts algorithm,34 and standard errors (SE) for coefficients were calculated using the kernel-based method suggested by Powell.35 Missing data were handled using multiple imputation. Thirty imputed datasets were created, after a burn-in of 30 replications, using predicted mean matching.36 The imputation model included all variables included in this analysis, and allowed for non-linear relationships using restricted cubic splines with five knots. We took the’ transform then impute’ approach recommended by von Hippel37 to impute variables derived from other variables with missing values. All models were estimated using each imputed dataset, and parameter estimates were combined using Rubin’s rules.38 Differences in proportions or means across sub-groups with and without missing data were tested using, respectively, Pearson’s chi-square test or Welsh’s t-test with unequal variances. Distributions of imputed values were examined visually. All analyses were conducted using the R Project for Statistical Computing39 (version 3.1.2). Quantile regression models were estimated using the quantreg package40 (version 5.11). Multiple imputation was implemented with the Hmisc package41 (version 3.14–6). All plots were produced using the ggplot2 package42 (version 1.0.0). Results Of the 1774 recruited mothers, three experienced a fetal loss before 20 weeks’ gestation; five pregnancies resulted in stillbirths; and 12 infants were born before 32 weeks’ completed gestation; these were excluded from the final analytical sample of 1754 infants. Variable distributions are described in Tables 1 and 2. The infants’ anthropometrics and gestational ages were consistent with established norms.16 Two-thirds of women had a healthy pre-pregnancy weight (BMI < 25 kg/m2), but only 16% experienced a healthy level of GWG. Whereas 16% of women said they were still consuming alcohol at recruitment, 80% of these women reported drinking one or fewer units per week. Based on the results from the fully adjusted quantile regressions (Tables 3–6), the conditional median for fat-free mass was 112 g less (95% CI −176 to −47) in infants born to women who experienced healthy GWG, compared with those who experienced excessive GWG. This reduction was less extreme at the lower centiles of fat-free mass (Figure 1a and Table 3). Healthy GWG was also associated with a 33 g reduction (95% CI −65 to −1) in median fat mass, and a 103 g reduction (95% CI −174 to −33) at the 95th centile of fat mass (Figure 1b and Table 4). Birth lengths were roughly 0.6 cm less at all centiles in infants born to women who experienced healthy GWG, though 95% CIs at several centiles included the null hypothesis of no difference (Figure 1c). Table 1. Characteristics of 1754 sample mother-infant pairs enrolled in SCOPE-Ireland, 2008 to 2011 Variable  Missing Values  Proportion (n)  Mean (SD)  Median[IQR]  Range  Infant characteristics            Sex  0           Male    0.51(892)         Female    0.49(862)        Birth weight (g)  0    3462 (507.5)  3460 [3150 to 3778]  1200 to 5130  Fat mass (g)  512    378 (172.9)  351 [253 to 481]  36 to 1099  Fat free mass (g)  513    2955 (346.8)  2965 [2730 to 3182]  1848 to 3960  Percent fat mass  514    11.1 (4.1)  10.9 [8.2 to 13.8]  1.3 to 30.1  Length (cm)  59    50.2 (2.4)  50.2 [49 to 51.8]  37.5 to 57  Gestational age (wks)  0    40 (1.5)  40.3 [39.3 to 41]  32 to 42.6  Maternal characteristics  Age  0    29.9 (4.5)  30 [28 to 33]  17 to 45  Height  0    164.6 (5.9)  165 [161 to 168]  147 to 185  Birth weight  60    3360.8 (532.9)  3374 [3062 to 3657]  624 to 6000  Gravidity  0           1    0.85(1483)         2+    0.15 (271)        Ethnicity  0           White    0.98(1712)         Nonwhite    0.02(42)        Has partner  0           Single    0.11(186)         Partner    0.89(1568)        3rd level education  0           No    0.11(195)         Yes    0.89(1559)        SEI†  0    42.7 (16)  45 [29 to 51]  18 to 89  Maternity care  0           Public    0.75(1318)         Private    0.25(436)        Depressed  0           Unlikely    0.41(711)         At risk    0.35(622)         Likely    0.24(421)        Stress score ††  0    13.7 (6.6)  13 [9 to 18]  0 to 35  Variable  Missing Values  Proportion (n)  Mean (SD)  Median[IQR]  Range  Infant characteristics            Sex  0           Male    0.51(892)         Female    0.49(862)        Birth weight (g)  0    3462 (507.5)  3460 [3150 to 3778]  1200 to 5130  Fat mass (g)  512    378 (172.9)  351 [253 to 481]  36 to 1099  Fat free mass (g)  513    2955 (346.8)  2965 [2730 to 3182]  1848 to 3960  Percent fat mass  514    11.1 (4.1)  10.9 [8.2 to 13.8]  1.3 to 30.1  Length (cm)  59    50.2 (2.4)  50.2 [49 to 51.8]  37.5 to 57  Gestational age (wks)  0    40 (1.5)  40.3 [39.3 to 41]  32 to 42.6  Maternal characteristics  Age  0    29.9 (4.5)  30 [28 to 33]  17 to 45  Height  0    164.6 (5.9)  165 [161 to 168]  147 to 185  Birth weight  60    3360.8 (532.9)  3374 [3062 to 3657]  624 to 6000  Gravidity  0           1    0.85(1483)         2+    0.15 (271)        Ethnicity  0           White    0.98(1712)         Nonwhite    0.02(42)        Has partner  0           Single    0.11(186)         Partner    0.89(1568)        3rd level education  0           No    0.11(195)         Yes    0.89(1559)        SEI†  0    42.7 (16)  45 [29 to 51]  18 to 89  Maternity care  0           Public    0.75(1318)         Private    0.25(436)        Depressed  0           Unlikely    0.41(711)         At risk    0.35(622)         Likely    0.24(421)        Stress score ††  0    13.7 (6.6)  13 [9 to 18]  0 to 35  IQR, interquartile range; SD, standard deviation; SEI, Socioeconomic index. †Based on the New Zealand socioeconomic index, with higher values reflecting greater social status. ††Out of a maximum score of 40, with higher scores reflecting higher levels of stress. Table 2. Maternal lifestyle factors in 1754 sample mothers enrolled in SCOPE-Ireland, 2008 to 2011 Variable  Missing values  Proportion (n)  Prepregnancy body size  0     Obese (BMI ≥ 30 kg/m2)    0.11 (190)   Overweight (BMI 25 to 30 kg/m2)    0.24 (419)   Normal weight (BMI < 25 kg/m2)    0.65 (1145)  IOM defined gestational weight gain level  525     Excessive    0.79 (977)   Healthy    0.16 (199)   Inadequate    0.04 (53)  Frequency of moderate intensity exercise  0     None    0.25 (441)   Some    0.55 (965)   Often    0.2 (348)  Amount of daily TV viewing  0     ≥5 h    0.09 (158)   2–4 h    0.55 (958)   <2 h    0.36 (638)  Alcohol use  0     Still drinks    0.16 (288)   Quit during pregnancy    0.65 (1133)   Quit prepregnancy    0.09 (166)   Never drank    0.1 (167)  Any smoking  0     Still smokes    0.1 (174)   Quit during pregnancy    0.18 (307)   Never smoked†    0.73 (1273)  Takes folate  0     No    0.32 (560)   Yes    0.68 (1194)  Eats ≥ 5 servings fruit and veg per day  0     No    0.86 (1508)   Yes    0.14 (246)  Eats ≥ 1 serving oily fish per week  0     No    0.69 (1205)   Yes    0.31 (549)  Variable  Missing values  Proportion (n)  Prepregnancy body size  0     Obese (BMI ≥ 30 kg/m2)    0.11 (190)   Overweight (BMI 25 to 30 kg/m2)    0.24 (419)   Normal weight (BMI < 25 kg/m2)    0.65 (1145)  IOM defined gestational weight gain level  525     Excessive    0.79 (977)   Healthy    0.16 (199)   Inadequate    0.04 (53)  Frequency of moderate intensity exercise  0     None    0.25 (441)   Some    0.55 (965)   Often    0.2 (348)  Amount of daily TV viewing  0     ≥5 h    0.09 (158)   2–4 h    0.55 (958)   <2 h    0.36 (638)  Alcohol use  0     Still drinks    0.16 (288)   Quit during pregnancy    0.65 (1133)   Quit prepregnancy    0.09 (166)   Never drank    0.1 (167)  Any smoking  0     Still smokes    0.1 (174)   Quit during pregnancy    0.18 (307)   Never smoked†    0.73 (1273)  Takes folate  0     No    0.32 (560)   Yes    0.68 (1194)  Eats ≥ 5 servings fruit and veg per day  0     No    0.86 (1508)   Yes    0.14 (246)  Eats ≥ 1 serving oily fish per week  0     No    0.69 (1205)   Yes    0.31 (549)  BMI, body mass index; IOM, Institute of Medicine. †Six women who reported quitting prior to pregnancy were classified as Never Smoked. Table 3. Quantile regression results from the fully adjusted model for fat-free mass (g), n = 1754   Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  2444.4  (2153.9 to 2734.9)  3011.4  (2864.3 to 3158.5)  3587.9  (3387.9 to 3787.9)  Healthy GWG  −38.5  (−173.1 to 96.1)  −111.7  (−176.2 to −47.2)  −109.8  (−186.7 to −32.9)  Inadequate GWG  34.7  (−181.2 to 250.6)  −57  (−165.6 to 51.6)  −83.2  (−240.3 to 73.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −166  (−295.8 to −36.2)  −35  (−100.2 to 30.2)  −121.6  (−249.8 to 6.6)  Overweight (BMI 25 to 30 kg/m2)  −123.3  (−271.5 to 24.9)  −8.4  (−81.1 to 64.3)  −90.6  (−223.5 to 42.3)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −1.7  (−110.5 to 107.1)  34.4  (−20.6 to 89.4)  −6  (−76.3 to 64.3)  Some moderate-intensity exercise  75.7  (−58.6 to 210)  38.1  (−15.1 to 91.3)  15.3  (−59.3 to 89.9)  Frequent moderate-intensity exercise  16.1  (−127.8 to 160)  −5.8  (−74.5 to 62.9)  −12.5  (−104 to 79)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  21.2  (−147 to 189.4)  −24.2  (−103.8 to 55.4)  −75.9  (−176.9 to 25.1)  < 2 hours of television  12.5  (−161 to 186)  −28.2  (−114.4 to 58)  −65.6  (−175.7 to 44.5)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  29.7  (−108.7 to 168.1)  4.1  (−56.5 to 64.7)  5.7  (−88.8 to 100.2)  Quit drinking prepregnancy  26.2  (−169 to 221.4)  −44.1  (−130.2 to 42)  −4.6  (−152.5 to 143.3)  Never drank  38.3  (−145 to 221.6)  36.2  (−63.8 to 136.2)  −51.7  (−156.8 to 53.4)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  165.5  (−14.8 to 345.8)  47.7  (−45.6 to 141)  60.1  (−57.2 to 177.4)  Never smoked  130.2  (−39.6 to 300)  81  (−5.3 to 167.3)  78.5  (−22.5 to 179.5)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  7  (−117.7 to 131.7)  10.8  (−55.2 to 76.8)  54.6  (−30.6 to 139.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −52.6  (−164.7 to 59.5)  18.4  (−28.8 to 65.6)  69.8  (−2.7 to 142.3)    Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  2444.4  (2153.9 to 2734.9)  3011.4  (2864.3 to 3158.5)  3587.9  (3387.9 to 3787.9)  Healthy GWG  −38.5  (−173.1 to 96.1)  −111.7  (−176.2 to −47.2)  −109.8  (−186.7 to −32.9)  Inadequate GWG  34.7  (−181.2 to 250.6)  −57  (−165.6 to 51.6)  −83.2  (−240.3 to 73.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −166  (−295.8 to −36.2)  −35  (−100.2 to 30.2)  −121.6  (−249.8 to 6.6)  Overweight (BMI 25 to 30 kg/m2)  −123.3  (−271.5 to 24.9)  −8.4  (−81.1 to 64.3)  −90.6  (−223.5 to 42.3)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −1.7  (−110.5 to 107.1)  34.4  (−20.6 to 89.4)  −6  (−76.3 to 64.3)  Some moderate-intensity exercise  75.7  (−58.6 to 210)  38.1  (−15.1 to 91.3)  15.3  (−59.3 to 89.9)  Frequent moderate-intensity exercise  16.1  (−127.8 to 160)  −5.8  (−74.5 to 62.9)  −12.5  (−104 to 79)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  21.2  (−147 to 189.4)  −24.2  (−103.8 to 55.4)  −75.9  (−176.9 to 25.1)  < 2 hours of television  12.5  (−161 to 186)  −28.2  (−114.4 to 58)  −65.6  (−175.7 to 44.5)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  29.7  (−108.7 to 168.1)  4.1  (−56.5 to 64.7)  5.7  (−88.8 to 100.2)  Quit drinking prepregnancy  26.2  (−169 to 221.4)  −44.1  (−130.2 to 42)  −4.6  (−152.5 to 143.3)  Never drank  38.3  (−145 to 221.6)  36.2  (−63.8 to 136.2)  −51.7  (−156.8 to 53.4)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  165.5  (−14.8 to 345.8)  47.7  (−45.6 to 141)  60.1  (−57.2 to 177.4)  Never smoked  130.2  (−39.6 to 300)  81  (−5.3 to 167.3)  78.5  (−22.5 to 179.5)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  7  (−117.7 to 131.7)  10.8  (−55.2 to 76.8)  54.6  (−30.6 to 139.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −52.6  (−164.7 to 59.5)  18.4  (−28.8 to 65.6)  69.8  (−2.7 to 142.3)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 4. Quantile regression results from the fully adjusted model for fat mass (g), n = 1754   Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  121.7  (33.7 to 209.7)  371.5  (298.2 to 444.8)  845  (670.3 to 1019.7)  Healthy GWG  −4.5  (−41.2 to 32.2)  −33.4  (−65.5 to −1.3)  −103.2  (−173.8 to −32.6)  Inadequate GWG  27  (−33.7 to 87.7)  −17.1  (−66.6 to 32.4)  −30.5  (−134.4 to 73.4)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −42.2  (−86 to 1.6)  −47.7  (−83.7 to −11.7)  −69.1  (−149.9 to 11.7)  Overweight (BMI 25 to 30 kg/m2)  −36.1  (−87.4 to 15.2)  −25.6  (−64.5 to 13.3)  −39.9  (−128.5 to 48.7)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −8.4  (−38.8 to 22)  8.6  (−18.1 to 35.3)  −16.6  (−74.8 to 41.6)  Some moderate-intensity exercise  9.8  (−23.1 to 42.7)  4.8  (−21.7 to 31.3)  −36.7  (−111.3 to 37.9)  Frequent moderate-intensity exercise  −4.1  (−46.4 to 38.2)  −0.6  (−33.2 to 32)  −91.9  (−168 to −15.8)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  −4.3  (−48.3 to 39.7)  12.6  (−27.9 to 53.1)  51.5  (−26.7 to 129.7)  < 2 hours of television  −25  (−67.8 to 17.8)  −6.8  (−50.9 to 37.3)  57.6  (−31.4 to 146.6)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  2.2  (−38.3 to 42.7)  −26.8  (−58.6 to 5)  −32.6  (−100 to 34.8)  Quit drinking prepregnancy  −3.3  (−69.1 to 62.5)  −22.4  (−67.4 to 22.6)  −88.9  (−191.3 to 13.5)  Never drank  18.2  (−40.1 to 76.5)  −14.3  (−60.1 to 31.5)  −122.3  (−204.1 to −40.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  17.1  (−33.6 to 67.8)  1.9  (−43 to 46.8)  −17.1  (−116.3 to 82.1)  Never smoked  28.8  (−17 to 74.6)  29.1  (−11.7 to 69.9)  29.6  (−54.8 to 114)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  5.4  (−34.1 to 44.9)  6.9  (−26.1 to 39.9)  31  (−38.4 to 100.4)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.7  (−30.6 to 27.2)  −4  (−27.3 to 19.3)  −21.7  (−69.5 to 26.1)    Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  121.7  (33.7 to 209.7)  371.5  (298.2 to 444.8)  845  (670.3 to 1019.7)  Healthy GWG  −4.5  (−41.2 to 32.2)  −33.4  (−65.5 to −1.3)  −103.2  (−173.8 to −32.6)  Inadequate GWG  27  (−33.7 to 87.7)  −17.1  (−66.6 to 32.4)  −30.5  (−134.4 to 73.4)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −42.2  (−86 to 1.6)  −47.7  (−83.7 to −11.7)  −69.1  (−149.9 to 11.7)  Overweight (BMI 25 to 30 kg/m2)  −36.1  (−87.4 to 15.2)  −25.6  (−64.5 to 13.3)  −39.9  (−128.5 to 48.7)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −8.4  (−38.8 to 22)  8.6  (−18.1 to 35.3)  −16.6  (−74.8 to 41.6)  Some moderate-intensity exercise  9.8  (−23.1 to 42.7)  4.8  (−21.7 to 31.3)  −36.7  (−111.3 to 37.9)  Frequent moderate-intensity exercise  −4.1  (−46.4 to 38.2)  −0.6  (−33.2 to 32)  −91.9  (−168 to −15.8)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  −4.3  (−48.3 to 39.7)  12.6  (−27.9 to 53.1)  51.5  (−26.7 to 129.7)  < 2 hours of television  −25  (−67.8 to 17.8)  −6.8  (−50.9 to 37.3)  57.6  (−31.4 to 146.6)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  2.2  (−38.3 to 42.7)  −26.8  (−58.6 to 5)  −32.6  (−100 to 34.8)  Quit drinking prepregnancy  −3.3  (−69.1 to 62.5)  −22.4  (−67.4 to 22.6)  −88.9  (−191.3 to 13.5)  Never drank  18.2  (−40.1 to 76.5)  −14.3  (−60.1 to 31.5)  −122.3  (−204.1 to −40.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  17.1  (−33.6 to 67.8)  1.9  (−43 to 46.8)  −17.1  (−116.3 to 82.1)  Never smoked  28.8  (−17 to 74.6)  29.1  (−11.7 to 69.9)  29.6  (−54.8 to 114)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  5.4  (−34.1 to 44.9)  6.9  (−26.1 to 39.9)  31  (−38.4 to 100.4)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.7  (−30.6 to 27.2)  −4  (−27.3 to 19.3)  −21.7  (−69.5 to 26.1)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 5. Quantile regression results from the fully adjusted model for birth length (cm), n = 1754   Centile     5th   50th   95th   Variable  β (cm)  95% CI  β (cm)  95% CI  β (cm)  95% CI  Intercept  (42.8 to 47)  50.5  (49.5 to 51.5)  53.7  (52.4 to 55)    Healthy GWG  −0.5  (−1.5 to 0.5)  −0.4  (−0.8 to 0)  −0.8  (−1.3 to −0.3)  Inadequate GWG  0.1  (−2.8 to 3)  −0.3  (−1 to 0.4)  −0.1  (−1.1 to 0.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −1.2  (−1.9 to −0.5)  −0.4  (−0.9 to 0.1)  −0.5  (−1.1 to 0.1)  Overweight (BMI 25 to 30 kg/m2)  −0.8  (−1.7 to 0.1)  −0.3  (−0.8 to 0.2)  −0.4  (−1 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  0.4  (−0.5 to 1.3)  0  (−0.3 to 0.3)  0.2  (−0.2 to 0.6)  Some moderate-intensity exercise  0.7  (−0.1 to 1.5)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  Frequent moderate-intensity exercise  −0.2  (−1.2 to 0.8)  −0.3  (−0.7 to 0.1)  −0.2  (−0.8 to 0.4)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  2.2  (0.3 to 4.1)  0  (−0.6 to 0.6)  −0.6  (−1.3 to 0.1)  < 2 hours of television  1.6  (−0.3 to 3.5)  0  (−0.6 to 0.6)  −0.5  (−1.2 to 0.2)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  0  (−0.9 to 0.9)  0.1  (−0.3 to 0.5)  0.2  (−0.2 to 0.6)  Quit drinking prepregnancy  −0.2  (−1.3 to 0.9)  −0.3  (−0.8 to 0.2)  −0.3  (−1.1 to 0.5)  Never drank  −0.9  (−2.9 to 1.1)  0  (−0.5 to 0.5)  −0.3  (−0.9 to 0.3)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  1.1  (0.2 to 2)  0.5  (0 to 1)  0.5  (−0.2 to 1.2)  Never smoked  0.6  (−0.3 to 1.5)  0.6  (0.1 to 1.1)  0.7  (0 to 1.4)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0.2  (−0.6 to 1)  0.2  (−0.2 to 0.6)  0.3  (−0.2 to 0.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.1  (−1.8 to −0.4)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)    Centile     5th   50th   95th   Variable  β (cm)  95% CI  β (cm)  95% CI  β (cm)  95% CI  Intercept  (42.8 to 47)  50.5  (49.5 to 51.5)  53.7  (52.4 to 55)    Healthy GWG  −0.5  (−1.5 to 0.5)  −0.4  (−0.8 to 0)  −0.8  (−1.3 to −0.3)  Inadequate GWG  0.1  (−2.8 to 3)  −0.3  (−1 to 0.4)  −0.1  (−1.1 to 0.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −1.2  (−1.9 to −0.5)  −0.4  (−0.9 to 0.1)  −0.5  (−1.1 to 0.1)  Overweight (BMI 25 to 30 kg/m2)  −0.8  (−1.7 to 0.1)  −0.3  (−0.8 to 0.2)  −0.4  (−1 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  0.4  (−0.5 to 1.3)  0  (−0.3 to 0.3)  0.2  (−0.2 to 0.6)  Some moderate-intensity exercise  0.7  (−0.1 to 1.5)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  Frequent moderate-intensity exercise  −0.2  (−1.2 to 0.8)  −0.3  (−0.7 to 0.1)  −0.2  (−0.8 to 0.4)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  2.2  (0.3 to 4.1)  0  (−0.6 to 0.6)  −0.6  (−1.3 to 0.1)  < 2 hours of television  1.6  (−0.3 to 3.5)  0  (−0.6 to 0.6)  −0.5  (−1.2 to 0.2)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  0  (−0.9 to 0.9)  0.1  (−0.3 to 0.5)  0.2  (−0.2 to 0.6)  Quit drinking prepregnancy  −0.2  (−1.3 to 0.9)  −0.3  (−0.8 to 0.2)  −0.3  (−1.1 to 0.5)  Never drank  −0.9  (−2.9 to 1.1)  0  (−0.5 to 0.5)  −0.3  (−0.9 to 0.3)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  1.1  (0.2 to 2)  0.5  (0 to 1)  0.5  (−0.2 to 1.2)  Never smoked  0.6  (−0.3 to 1.5)  0.6  (0.1 to 1.1)  0.7  (0 to 1.4)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0.2  (−0.6 to 1)  0.2  (−0.2 to 0.6)  0.3  (−0.2 to 0.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.1  (−1.8 to −0.4)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 6. Quantile regression results from the fully adjusted model for gestational age (weeks), n = 1754   Centile     5th   50th   95th   Variable  β (weeks)  95% CI  β (weeks)  95% CI  β (weeks)  95% CI  Intercept  36  (34.2 to 37.8)  40.7  (40.2 to 41.2)  41.6  (41.3 to 41.9)  Healthy GWG  0.1  (−1 to 1.2)  0  (−0.2 to 0.2)  0  (−0.2 to 0.2)  Inadequate GWG  0.8  (−0.4 to 2)  −0.1  (−0.7 to 0.5)  0  (−0.3 to 0.3)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  0  (−0.9 to 0.9)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Overweight (BMI 25 to 30 kg/m2)  0.2  (−0.8 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −0.2  (−0.9 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  Some moderate-intensity exercise  0.8  (0 to 1.6)  0.2  (0 to 0.4)  0  (−0.1 to 0.1)  Frequent moderate-intensity exercise  −0.1  (−1.5 to 1.3)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  0.5  (−0.6 to 1.6)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  < 2 hours of television  0.3  (−0.8 to 1.4)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  −0.4  (−1.1 to 0.3)  −0.2  (−0.4 to 0)  0.1  (0 to 0.2)  Quit drinking prepregnancy  −0.8  (−2.5 to 0.9)  −0.4  (−0.7 to −0.1)  0  (−0.2 to 0.2)  Never drank  0  (−0.9 to 0.9)  −0.1  (−0.5 to 0.3)  0.2  (−0.1 to 0.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  0.5  (−0.5 to 1.5)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Never smoked  0.3  (−0.6 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0  (−0.7 to 0.7)  −0.1  (−0.3 to 0.1)  0  (−0.2 to 0.2)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −0.1  (−0.7 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)    Centile     5th   50th   95th   Variable  β (weeks)  95% CI  β (weeks)  95% CI  β (weeks)  95% CI  Intercept  36  (34.2 to 37.8)  40.7  (40.2 to 41.2)  41.6  (41.3 to 41.9)  Healthy GWG  0.1  (−1 to 1.2)  0  (−0.2 to 0.2)  0  (−0.2 to 0.2)  Inadequate GWG  0.8  (−0.4 to 2)  −0.1  (−0.7 to 0.5)  0  (−0.3 to 0.3)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  0  (−0.9 to 0.9)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Overweight (BMI 25 to 30 kg/m2)  0.2  (−0.8 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −0.2  (−0.9 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  Some moderate-intensity exercise  0.8  (0 to 1.6)  0.2  (0 to 0.4)  0  (−0.1 to 0.1)  Frequent moderate-intensity exercise  −0.1  (−1.5 to 1.3)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  0.5  (−0.6 to 1.6)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  < 2 hours of television  0.3  (−0.8 to 1.4)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  −0.4  (−1.1 to 0.3)  −0.2  (−0.4 to 0)  0.1  (0 to 0.2)  Quit drinking prepregnancy  −0.8  (−2.5 to 0.9)  −0.4  (−0.7 to −0.1)  0  (−0.2 to 0.2)  Never drank  0  (−0.9 to 0.9)  −0.1  (−0.5 to 0.3)  0.2  (−0.1 to 0.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  0.5  (−0.5 to 1.5)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Never smoked  0.3  (−0.6 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0  (−0.7 to 0.7)  −0.1  (−0.3 to 0.1)  0  (−0.2 to 0.2)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −0.1  (−0.7 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Figure 1 View largeDownload slide Healthy gestational weight gain. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for healthy (versus excessive) gestational weight gain at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Panel a shows that almost the entire distribution of fat-free mass is shifted to the left (towards smaller values) among infants born to women with healthy gestational weight gain (versus not), whereas Panel b shows that the right tail of the distribution of fat mass is being pulled in, with little change in the left tail of the distribution. Figure 1 View largeDownload slide Healthy gestational weight gain. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for healthy (versus excessive) gestational weight gain at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Panel a shows that almost the entire distribution of fat-free mass is shifted to the left (towards smaller values) among infants born to women with healthy gestational weight gain (versus not), whereas Panel b shows that the right tail of the distribution of fat mass is being pulled in, with little change in the left tail of the distribution. Pre-pregnancy normal weight status, compared with women classified as obese, was associated with a 48 g reduction (95% CI −84 to −12) in median fat mass, though 95% CIs at several centiles included the null hypothesis of no difference (Figure 2b and Table 4). Pre-pregnancy weight status was otherwise unrelated to outcomes. Frequent bouts of moderate-intensity exercise were associated with a reduction in the upper tail of the fat mass distribution (Figure 3a). For example, the 95th centile of fat mass in infants born to women who exercised frequently was 92 g less (95% CI −168 to −16) than in infants born to women who reported never exercising. The upper centiles of fat mass were also reduced in infants born to women who reported never drinking. For example, never drinking was associated with a 122 g reduction (95% CI −204 to −40) in the 95th centile of fat mass (Figure 4 and Table 4). Babies born to women who never smoked had greater fat-free mass and length, at all centiles (Figure 5). Figure 2 View largeDownload slide Maternal pre-pregnancy normal weight. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for maternal pre-pregnancy normal weight (versus obese) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 2 View largeDownload slide Maternal pre-pregnancy normal weight. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for maternal pre-pregnancy normal weight (versus obese) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 3 View largeDownload slide Frequent moderate-intensity exercise. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for frequent (versus not) moderate-intensity exercise at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 3 View largeDownload slide Frequent moderate-intensity exercise. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for frequent (versus not) moderate-intensity exercise at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 4 View largeDownload slide Never drank alcohol. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never drank alcohol (versus still drinking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 4 View largeDownload slide Never drank alcohol. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never drank alcohol (versus still drinking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 5 View largeDownload slide Never smoked. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never smoked (versus still smoking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 5 View largeDownload slide Never smoked. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never smoked (versus still smoking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. No other lifestyle factors were strongly related to infant outcomes in the fully adjusted quantile regression models (see Supplementary material, available at IJE online). Inclusion of a product interaction term for pre-pregnancy weight status and healthy GWG did not appreciably improve fit in any of the quantile regression models we estimated (likelihood ratio test P > 0.10 in all cases). All reported models thus excluded this term. A description of missing data is provided in the Supplementary material, available at IJE online. Discussion Using data from a large cohort of pregnant women, we used quantile regression to link maternal lifestyle factors to the distributions of fat and fat-free mass in their newborn offspring. Consistent with previous research (described below), maternal obesity and excess GWG were most strongly associated with newborn body composition. Both are established risk factors for macrosomia,25,43 which supports the idea that the associated long-term cardiometabolic risks result from an excess of substrate available to the fetus, particularly glucose, and subsequent fetal hyperinsulinaemia.44–46 Whereas newborn body composition is likely a more sensitive reflection of these fetal influences than total mass,9,10 relatively few studies have looked at the independent effects of GWG and pre-pregnancy body size on infant fat and fat-free mass. We found that maternal obesity (versus normal weight) was associated with an increase in newborn fat mass, though 95% CIs did not exclude the null hypothesis of zero difference at some centiles. This result is broadly consistent with Hull et al.47 (n = 306), Sewell et al.9,(n = 221), Carlsen et al.48 (n = 311), Au et al.49 (n = 599), Starling et al.50 (n = 826) and Friss et al.51 (n = 207), who found that maternal overweight and/or obesity were associated with increased fat mass and/or percentage fat mass. We also found that IOM-defined excessive GWG was associated with increases in both fat and fat-free mass. This finding was consistent with Au et al.49 who observed that each kg of weight gained during pregnancy was associated with increased percentage fat mass and birthweight, and Carlsen et al.48 and Starling et al.50 who found a similar association with fat and fat-free mass. Crozier et al.52 (n = 564) also found that IOM-healthy GWG was associated with reduced fat mass at birth, though they did not control for maternal pre-pregnancy body size. Friis et al.51 was the only relevant study not to observe an association between GWG and infant fat mass. Sewell et al.9 and Hull et al.47 also found that the apparent association between GWG and infant body composition was modified by maternal weight status, though the nature of this interaction differed between studies. The former observed that weight gain during pregnancy was associated with fat-free mass in the infants born to normal weight women, and with percentage fat mass in infants born to overweight and obese women; the latter found that the increase in percentage fat mass associated with excessive GWG was pronounced in overweight mothers. Our analysis, as well as that by Starling et al.,50 found no evidence of such interactions. Relative to the previous research just described, the unique contribution of our analysis comes from our use of quantile regression and its ability to estimate the effect of predictors on the tails of an outcome’s distribution. This approach complements a previous paper using this sample, which used linear regression to investigate lifestyle predictors of neonatal percentage body fat.53 For example, we found that excessive GWG was clearly associated with higher newborn fat-free mass, and some evidence that it was associated with a small increase in length. Importantly, these associations were consistent across the respective centiles of fat-free mass and length. This is expected, since GWG reflects changes in multiple tissues of the mother, placenta and developing fetus.24 This is of course true for fat mass as well, except that the association between GWG and fat mass was more pronounced at the upper end of the fat mass distribution. Further, the observed change in fat mass at the upper end of its distribution is larger than those seen for fat-free mass and length, relative to their respective means and variances. This could be reflecting that newborn fat mass is more modifiable than infant fat-free mass and length, which are under stronger genetic control.49,51,54,55 Consequently, we suggest that the nature of the GWG-fat mass relationship, which would have perhaps been missed using methods other than quantile regression, is reflecting a pathogenic effect of unhealthy GWG on fetal fat mass accumulation; whereas the reduction in fat-free mass associated with healthy GWG is just reflecting the functional relationship between the two variables. Although maternal smoking is a long-recognized determinant of infant length and total mass, very few studies have investigated the associations of maternal smoking with newborn fat and fat-free mass. Lindsay et al.56 (n = 129) and Spady et al.57 (n = 78) found that maternal smoking was associated with decreased fat-free mass and length, but not fat mass. A larger (n = 916), more recent study also found that neonates exposed to smoking throughout the pregnancy had lower fat-free mass.58 Our results are consistent with these. Quantile regression also allowed us to observe that the highest centiles of fat mass were lower among infants born to women who never drank compared with women who were still drinking during pregnancy. It is important to note that the reported amount of alcohol being drunk by the latter group was low. Further, there was no appreciable difference in infant outcomes between women who were still drinking at recruitment and those who reported quitting before pregnancy. It thus seems likely that the observed association between never drinking and outcomes is at least partly explained by other factors experienced by this relatively small group of women (10% of sample). Similarly, quantile regression revealed that the highest centiles of fat mass were also lower among infants born to women who engaged in frequent moderate-intensity exercise early in the pregnancy. This conflicts with a recent finding59 in a sample of 826 mother-neonate pairs where physical activity in early pregnancy was not associated with fat mass, fat-free mass or birthweight. However, this study did not look at the tails of the fat mass distribution, and a difference like the one we observed could have been obscured in a comparison of mean fat mass values. While we caution against over-interpreting this result, it is encouraging to think that frequent, moderate-intensity activity might help reduce the prevalence of babies born with a very high amount of fat mass. Further, we found no evidence that exercise was associated with reduced fetal growth (reflected in birth length) or gestational age, which might help reassure everyone that moderate-intensity exercise during pregnancy is in fact safe. Strengths This analysis uses data from one of the largest studies of newborn body composition60 measured with a reference method.61 Further, the population-based nature of the study allowed us to estimate the independent associations of a variety of healthy and unhealthy behaviours in a more representative sample than is often possible. Our use of quantile regression yielded insights that would not have been apparent with multiple linear regression. Binary or multinomial logistic regression is another common alternative for looking at the ends of a distribution, but when the outcome is inherently continuous in nature (e.g. mass), one must first impose one or more cut-offs. The logistic regression model treats any set of two values falling on different sides of a cut-off as similarly different, regardless of the distance between two values on the real number line. Thus quantile regression allows us to investigate the influence of a predictor on the tails of an outcome’s distribution while still respecting the underlying continuous nature of that variable. The potential value of quantile regression for epidemiologists has been highlighted by Beyerlein,62 and quantile regression has been previously applied to anthropometric studies across the life course.63 The ideal expression of infant body composition has been previously debated.64 It is common to focus on percentage fat mass,65 but this has limitations.66 Ratio measures, generally, have poor statistical properties.67 Additionally, because fat mass appears in both the numerator and the denominator of percentage fat mass, increases in absolute fat mass become less obvious as percentage fat mass increases. Further, focusing on fat mass can lead us to ignore the fat-free mass,65 which is often just as important to consider.68,69 Finally, though the supposed value of calculating percentage fat mass is to arrive at a measure that is normalized for overall body size, percentage fat mass is typically correlated with height, which was true in our sample (results not shown). Contrary to common practice, we did not restrict the sample to term infants, nor include gestational age in our models of body composition or birth length. Because gestational age does not precede any of the maternal lifestyle factors we have investigated, it cannot confound associations between lifestyle factors and birth size. It could, however, mediate the influence of lifestyle on birth size, and thus adjusting for it could lead to collider bias when gestational age and birth size share other unmeasured causes. This issue has been previously described by Wilcox, Weinberg and Basso.70 Similarly, we also did not adjust models of fat and fat-free mass for length, since it is likely that these will share unmeasured causes that again could confound associations between lifestyle factors and infant body composition once length is adjusted for. Limitations The main weakness of this research is that it is a secondary analysis of data collected with an observational study design. There is thus considerable potential for meaningful confounding and selection bias. These were hopefully minimized by thoughtful selection of covariates based on theoretical grounds. It is worth noting that models were not modified to better fit the data or in light of any preliminary analyses. There were missing data for both outcomes and predictors in the sample. We used multiple imputation to estimate parameters under an assumption of missing at random (MAR), conditional on other observed covariates, which is more defensible than the assumption that data were missing completely at random (MCAR).71 Our imputation models included the outcomes, which is recommended when there are missing values in the predictors.72 However, we should note that the majority of missing values for body composition were because the PEA POD was not yet available on site, and thus those values are likely MCAR. Thus the main value of multiple imputation in this analysis is an increased sample size and a consequent increase in power to detect effects than had we instead used the complete case analysis, or an impute-then-delete approach.73 Most of the data on maternal lifestyle factors were based on self-report and are likely measured with considerable error. This was particularly true for the dietary data, which were based on questions about the consumption of nine selected food items rather than more established methods for dietary assessment, and so our ability to predict infant body composition from the dietary data was likely quite poor from the outset. Similarly, the survey question about moderate-intensity exercise, though consistent with the Centres for Disease Control (CDC) and World Health Organization (WHO) definitions, is quite broad. Importantly, useful information on gestational diabetes mellitus (GDM) was lacking on over two-thirds of the sample and was not considered in this analysis. However, whereas a GDM diagnosis can impact on subsequent lifestyle factors later in the pregnancy, it could not have affected the early/pre-pregnancy factors we have considered here, and thus could not have confounded the associations we report. Conclusion Despite its clear role in the developmental origins of health and disease, little is known about maternal influences on infant body composition. Taken as a whole, this analysis suggests that supporting healthy maternal lifestyles could reduce the risk of excess fat accumulation in the offspring, without increasing the risk of low body fat or adversely affecting fat-free mass development, length, or gestational age. We suggest that future evaluations of maternal lifestyle interventions include more direct infant body composition measures, and use quantile regression to analyse the subsequent data. The use of quantile regression led to insights that would not have been apparent in the data if using more commonly applied linear or logistic regression models. Supplementary Data Supplementary data are available at IJE online. Funding This work was supported by the Irish Centre for Fetal and Neonatal Translational Research (INFANT) and was funded in part by Science Foundation Ireland (grant number 12/RC/2272). SCOPE Ireland was funded by the Health Research Board, Ireland (CSA 2007/2). The BASELINE cohort was supported by the National Children’s Research Centre, Dublin, Ireland. Analysis of the data was supported by a Health Research Board Interdisciplinary Capacity Enhancement award (ICE/2012/12). Acknowledgements We thank the pregnant women who participated in the SCOPE study and the mothers who allowed their newborn infants participate in the BASELINE study. Conflicts of interest: None. References 1 Adair LS, Dahly DL. Developmental determinants of blood pressure in adults. Annu Rev Nutr  2005; 25: 407– 34. Google Scholar CrossRef Search ADS PubMed  2 Gillman MW. Developmental origins of health and disease. N Engl J Med  2005; 353: 1848– 50. Google Scholar CrossRef Search ADS PubMed  3 Godfrey KM, Lillycrop KA, Burdge GC, Gluckman PD, Hanson MA. Epigenetic mechanisms and the mismatch concept of the developmental origins of health and disease. Pediatr Res  2007; 61: 5 R– 10 R. Google Scholar CrossRef Search ADS   4 Barker DJP. Developmental origins of adult health and disease. J Epidemiol Community Health  2004; 58: 114– 5. Google Scholar CrossRef Search ADS PubMed  5 Hanson MA, Gluckman PD, Ma RCW, Matzen P, Biesma RG. Early life opportunities for prevention of diabetes in low and middle income countries. BMC Public Health  2012; 12: 1025. Google Scholar CrossRef Search ADS PubMed  6 Horton R. Maternal and child undernutrition: an urgent opportunity. Lancet  2008; 371: 179. Google Scholar CrossRef Search ADS PubMed  7 Wells JCK. Historical cohort studies and the early origins of disease hypothesis: making sense of the evidence. Proc Nutr Soc  2009; 68: 179– 88. Google Scholar CrossRef Search ADS PubMed  8 Vrijheid M, Casas M, Bergström A, et al.   European birth cohorts for environmental health research. Environ Health Perspect  2012; 120: 29– 37. Google Scholar CrossRef Search ADS PubMed  9 Sewell MF, Huston-Presley L, Super DM, Catalano P. Increased neonatal fat mass, not lean body mass, is associated with maternal obesity. Am J Obstet Gynecol  2006; 195: 1100– 3. Google Scholar CrossRef Search ADS PubMed  10 Catalano PM, Tyzbir ED, Allen SR, McBean JH, McAuliffe TL. Evaluation of fetal growth by estimation of neonatal body composition. Obstet Gynecol  1992; 79: 46– 50. Google Scholar PubMed  11 Wells JCK. Body composition in infants: evidence for developmental programming and techniques for measurement. Rev Endocr Metab Disord  2012; 13: 93– 101. Google Scholar CrossRef Search ADS PubMed  12 Hull HR, Dinger MK, Knehans AW, Thompson DM, Fields DA. Impact of maternal body mass index on neonate birthweight and body composition. Am J Obstet Gynecol  2008; 198: 416.e1– 6. Google Scholar CrossRef Search ADS   13 IAEA. Body Composition Assessment from Birth to Two Years of Age . Vienna: International Atomic Energy Agency, 2013. 14 Crozier SR, Inskip HM, Godfrey KM, et al.   Weight gain in pregnancy and childhood body composition: findings from the Southampton Women’s Survey. Am J Clin Nutr  2010; 91: 1745– 51. Google Scholar CrossRef Search ADS PubMed  15 Holzhauer S, Hokken Koelega ACS, Ridder M de, et al.   Effect of birthweight and postnatal weight gain on body composition in early infancy: The Generation R Study. Early Hum Dev  2009; 85: 285– 90. Google Scholar CrossRef Search ADS PubMed  16 Hawkes CP, Hourihane JO, Kenny LC, et al.   Gender- and gestational age-specific body fat percentage at birth. Pediatrics  2011; 128: e645– 51. Google Scholar PubMed  17 Andersen GS, Girma T, Wells JCK, et al.   Body composition from birth to 6 months of age in Ethiopian infants: reference data obtained by air-displacement plethysmography. Am J Clin Nutr  2013; 98: 885– 94. Google Scholar CrossRef Search ADS PubMed  18 Lee W, Riggs T, Koo WWK, et al.   Relationship of newborn adiposity to fetal growth outcome based on birthweight or the modified neonatal growth assessment score. J Matern Fetal Neonatal Med  2012; 25: 1933– 40. Google Scholar CrossRef Search ADS PubMed  19 Koenker R, Hallock KF. Quantile Regression. J Econ Perspect  2001; 15: 143– 56. Google Scholar CrossRef Search ADS   20 O’Donovan SM, Murray DM, Hourihane JO, Kenny LC, Irvine AD, Kiely M. Cohort Profile: The Cork BASELINE Birth Cohort Study: Babies after SCOPE: evaluating the longitudinal impact on neurological and nutritional endpoints. Int J Epidemiol  2015; 44: 764– 75. Google Scholar CrossRef Search ADS PubMed  21 North RA, McCowan LME, Dekker GA, et al.   Clinical risk prediction for pre-eclampsia in nulliparous women: development of model in international prospective cohort. BMJ  2011; 342: d1875. Google Scholar CrossRef Search ADS PubMed  22 Fomon SJ, Haschke F, Ziegler EE, Nelson SE. Body composition of reference children from birth to age 10 years. Am J Clin Nutr  1982; 35: 1169– 75. Google Scholar CrossRef Search ADS PubMed  23 Health Services Executive Ireland. Nutrition for Pregnancy . Clinical Practice Guideline No. 27 V 1.1. Limerick, Ireland: HSEI, 2016. 24 Rasmussen KM, Yaktine AL (eds); Institute of Medicine (US) and National Research Council (US) Committee to Reexamine IOM Pregnancy Weight Guidelines. Weight Gain During Pregnancy: Reexamining the Guidelines . Washington, DC: National Academies Press (US), 2009. 25 Chung JGY, Taylor RS, Thompson JMD, et al.   Gestational weight gain and adverse pregnancy outcomes in a nulliparous cohort. Eur J Obstet Gynecol Reprod Biol  2013; 167: 149– 53. Google Scholar CrossRef Search ADS PubMed  26 Clark BK, Sugiyama T, Healy GN, et al.   Validity and reliability of measures of television viewing time and other non-occupational sedentary behaviour of adults: a review. Obes Rev  2009; 10: 7– 16. Google Scholar CrossRef Search ADS PubMed  27 Sugiyama T, Healy GN, Dunstan DW, Salmon J, Owen N. Is television viewing time a marker of a broader pattern of sedentary behavior? Ann Behav Med  2008; 35: 245– 50. Google Scholar CrossRef Search ADS PubMed  28 Galbraith C, Jenkin G, Davis P, Coope P. New Zealand Socio-economic Index 1996 Users’ Guide . Wellington: Statistics New Zealand, 1996. 29 Cox JL, Holden JM, Sagovsky R. Detection of postnatal depression: development of the 10-item Edinburgh Postnatal Depression scale. Br J Psychiatry  1987; 150: 782– 86. Google Scholar CrossRef Search ADS PubMed  30 Adouard F, Glangeaud-Freudenthal NMC, Golse B. Validation of the Edinburgh postnatal depression scale (EPDS) in a sample of women with high-risk pregnancies in France. Arch Womens Ment Health  2005; 8: 89– 95. Google Scholar CrossRef Search ADS PubMed  31 Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav  1983; 24: 385– 96. Google Scholar CrossRef Search ADS PubMed  32 Tukey JW. Exploratory Data Analysis . Reading, MA: Addison-Wesley, 1977. 33 Nohr EA, Vaeth M, Baker JL, et al.   Combined associations of prepregnancy body mass index and gestational weight gain with the outcome of pregnancy. Am J Clin Nutr  2008; 87: 1750– 59. Google Scholar CrossRef Search ADS PubMed  34 Koenker RW, D’Orey V. Algorithm AS 229: computing regression quantiles. J R Stat Soc  1987; 36: 383– 93. 35 Powell J. Estimation of monotonic regression models under quantile restrictions. In: Powell J, Tauchen G (eds.). Nonparametric and Semiparametric Methods in Econometrics . Cambridge, UK: Cambridge University Press, 1989. 36 van Buuren S. Flexible Imputation of Missing Data . Boca Raton FL: Chapman & Hall/CRC, 2012. Google Scholar CrossRef Search ADS   37 Von Hippel PT. How to impute interactions, squares, and other transformed variables. Sociol Methodol  2009; 39: 265– 91. Google Scholar CrossRef Search ADS   38 Rubin DB. Multiple Imputation for Nonresponse in Surveys . New York, NY: Wiley, 1987. Google Scholar CrossRef Search ADS   39 R Development Core Team. R: A Language and Environment for Statistical Computing . Vienna: R Foundation for Statistical Computing, 2004. 40 Koenker R. quantreg: Quantile Regression . R package version 5.11. Vienna: R Foundation for Statistical Computing, 2014. 41 Harrell FE. Hmisc: Harrell Miscellaneous . R package version 3.14–6. Vienna: R Foundation for Statistical Computing, 2014. 42 Wickham H. ggplot2: Elegant Graphics for Data Analysis . New York, NY: Springer, 2009. 43 Catalano PM. Management of obesity in pregnancy. Obstet Gynecol  2007; 109: 419– 33. Google Scholar CrossRef Search ADS PubMed  44 Catalano PM, Hauguel-De Mouzon S. Is it time to revisit the Pedersen hypothesis in the face of the obesity epidemic? Am J Obstet Gynecol  2011; 204: 479– 87. Google Scholar CrossRef Search ADS PubMed  45 Pedersen J. Diabetes and pregnancy: blood sugar of newborn infants. Ugeskr Laeger  1952; 114: 685. Google Scholar PubMed  46 Mitanchez D, Burguet A, Simeoni U. Infants born to mothers with gestational diabetes mellitus: mild neonatal effects, a long-term threat to global health. J Pediatr  2014; 164: 445– 50. Google Scholar CrossRef Search ADS PubMed  47 Hull HR, Thornton JC, Ji Y, et al.   Higher infant body fat with excessive gestational weight gain in overweight women. Am J Obstet Gynecol  2011; 205: 211– 12. Google Scholar CrossRef Search ADS PubMed  48 Carlsen E, Renault K, Nørgaard K, et al.   Newborn regional body composition is influenced by maternal obesity, gestational weight gain and the birthweight standard score. Acta Paediatr  2014; 103: 939– 45. Google Scholar CrossRef Search ADS PubMed  49 Au CP, Raynes-Greenow CH, Turner RM, Carberry AE, Jeffery H. Fetal and maternal factors associated with neonatal adiposity as measured by air displacement plethysmography: a large cross-sectional study. Early Hum Dev  2013; 89: 839– 43. Google Scholar CrossRef Search ADS PubMed  50 Starling AP, Brinton JT, Glueck DH, et al.   Associations of maternal BMI and gestational weight gain with neonatal adiposity in the Healthy Start study. Am J Clin Nutr  2015; 101: 302– 09. Google Scholar CrossRef Search ADS PubMed  51 Friis CM, Qvigstad E, Paasche Roland MC, et al.   Newborn body fat: associations with maternal metabolic state and placental size. PLoS One  2013; 8: e57467. Google Scholar CrossRef Search ADS PubMed  52 Crozier SR, Inskip HM, Godfrey KM, et al.   Weight gain in pregnancy and childhood body composition: findings from the Southampton Women’s Survey. Am J Clin Nutr  2010; 91: 1745– 51. Google Scholar CrossRef Search ADS PubMed  53 McCarthy FP, Khashan AS, Murray D, et al.   Parental physical and lifestyle factors and their association with newborn body composition. BJOG  2016; 123: 1824– 29. Google Scholar CrossRef Search ADS PubMed  54 Sparks JW. Human intrauterine growth and nutrient accretion. Semin Perinatol  1984; 8: 74– 93. Google Scholar PubMed  55 Catalano PM, Thomas AJ, Huston-Presley L, Amini SB. Increased fetal adiposity: a very sensitive marker of abnormal in utero development. Am J Obstet Gynecol  2003; 189: 1698– 704. Google Scholar CrossRef Search ADS PubMed  56 Lindsay CA, Thomas AJ, Catalano PM. The effect of smoking tobacco on neonatal body composition. Am J Obstet Gynecol  1997; 177: 1124– 28. Google Scholar CrossRef Search ADS PubMed  57 Spady DW, Atrens MA, Szymanski WA. Effects of mother’s smoking on their infants’ body composition as determined by total body potassium. Pediatr Res  1986; 20: 716– 19. Google Scholar CrossRef Search ADS PubMed  58 Harrod CS, Reynolds RM, Chasan-Taber L, et al.   Quantity and timing of maternal prenatal smoking on neonatal body composition: the Healthy Start Study. J Pediatr  2014; 165: 707– 12. Google Scholar CrossRef Search ADS PubMed  59 Harrod CS, Chasan-Taber L, Reynolds RM, et al.   Physical activity in pregnancy and neonatal body composition: the Healthy Start study. Obstet Gynecol  2014; 124: 257– 64. Google Scholar CrossRef Search ADS PubMed  60 Hawkes CP, Hourihane JO, Kenny LC, et al.   Gender- and gestational age-specific body fat percentage at birth. Pediatrics  2011; 128: e645– 51. Google Scholar PubMed  61 Ellis KJ, Yao M, Shypailo RJ, et al.   Body composition assessment in infancy: air-displacement plethysmography compared with a reference 4-compartment model. Am J Clin Nutr  2007; 85: 90– 95. Google Scholar CrossRef Search ADS PubMed  62 Beyerlein A. Quantile regression-opportunities and challenges from a user’s perspective. Am J Epidemiol  2014; 180: 330– 31. Google Scholar CrossRef Search ADS PubMed  63 Wei Y, Terry MB. Re: “Quantile Regression - Opportunities and Challenges From a User’s Perspective.” Am J Epidemiol  2014; 181: 152– 53. Google Scholar CrossRef Search ADS PubMed  64 Wells JC. A critique of the expression of paediatric body composition data. Arch Dis Child  2001; 85: 67– 72. Google Scholar CrossRef Search ADS PubMed  65 Wells JCK. Toward body composition reference data for infants, children, and adolescents. Adv Nutr  2014; 5: 320 S– 29 S. Google Scholar CrossRef Search ADS   66 Wells JCK, Cole TJ. the ALSPAC study team. Adjustment of fat-free mass and fat mass for height in children aged 8 y. Int J Obes Relat Metab Disord  2002; 26: 947– 52. Google Scholar CrossRef Search ADS PubMed  67 Kronmal RA. Spurious correlation and the fallacy of the ratio standard revisited. J R Stat Soc A  1993; 156: 379– 92. Google Scholar CrossRef Search ADS   68 Wells JCK. Historical cohort studies and the early origins of disease hypothesis: making sense of the evidence. Proc Nutr Soc  2009; 68: 179– 88. Google Scholar CrossRef Search ADS PubMed  69 Wells JCK. Body composition in infants: evidence for developmental programming and techniques for measurement. Rev Endocr Metab Disord  2012; 13: 93– 101. Google Scholar CrossRef Search ADS PubMed  70 Wilcox AJ, Weinberg CR, Basso O. On the pitfalls of adjusting for gestational age at birth. Am J Epidemiol  2011; 174: 1062– 68. Google Scholar CrossRef Search ADS PubMed  71 Little RJA, Rubin DB. Statistical Analysis with Missing Data . Hoboken, NJ: Wiley, 2002. Google Scholar CrossRef Search ADS   72 Moons KGM, Donders RART, Stijnen T, Harrell FE. Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol  2006; 59: 1092– 101. Google Scholar CrossRef Search ADS PubMed  73 von Hippel PT. 4. Regression with missing Ys: an improved strategy for analyzing multiply imputed data. Sociol Methodol  2007; 37: 83– 117. Google Scholar CrossRef Search ADS   © The Author 2017; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Epidemiology Oxford University Press

Loading next page...
 
/lp/ou_press/associations-between-maternal-lifestyle-factors-and-neonatal-body-DtPfkHOv8i
Publisher
Oxford University Press
Copyright
© The Author 2017; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association
ISSN
0300-5771
eISSN
1464-3685
D.O.I.
10.1093/ije/dyx221
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Neonatal body composition likely mediates fetal influences on life long chronic disease risk. A better understanding of how maternal lifestyle is related to newborn body composition could thus inform intervention efforts. Methods Using Cork participant data (n = 1754) from the Screening for Pregnancy Endpoints (SCOPE) cohort study [ECM5(10)05/02/08], we estimated how pre-pregnancy body size, gestational weight gain, exercise, alcohol, smoking and diet were related to neonatal fat and fat-free mass, as well as length and gestational age at birth, using quantile regression. Maternal factors were measured by a trained research midwife at 15 gestational weeks, in addition to a 3rd trimester weight measurement used to calculate weight gain. Infant body composition was measured using air-displacement plethysmography. Results Healthy (versus excess) gestational weight gain was associated with lower median fat-free mass [−112 g, 95% confidence interval (CI): −47 to −176) and fat mass (−33 g, 95% CI: −1 to −65) in the offspring; and a 103 g decrease in the 95th centile of fat mass (95% CI: −33 to −174). Maternal normal weight status (versus obesity) was associated with lower median fat mass (−48 g, 95% CI: −12 to −84). At the highest centiles, fat mass was lower among infants of women who engaged in frequent moderate-intensity exercise early in the pregnancy (−92 g at the 95th centile, 95% CI: −168 to −16). Lastly, women who never smoked tended to have longer babies with more fat mass and fat-free mass. No other lifestyle factors were strongly related to infant body composition. Conclusions These results suggest that supporting healthy maternal lifestyles could reduce the risk of excess fat accumulation in the offspring, without adversely affecting fat-free mass development, length or gestational age. Developmental origins, body composition, lifestyle, birth cohort, quantile regression Key Messages Maternal body size and gestational weight gain were most strongly associated with newborn body composition. Associations between maternal lifestyle and newborn body composition were often stronger in the tails of the outcome’s distribution. Quantile regression provided useful insights that would not have been apparent with linear or logistic regression. Introduction The environment experienced by a fetus during development influences its risk for cardiovascular disease, diabetes and obesity across the life course.1–4 Healthy maternal lifestyles before and during pregnancy may affect fetal development and thus influence these outcomes,5,6 but our understanding is limited since many birth cohort studies capable of linking maternal factors to cardiometabolic risks in the offspring are not yet into middle age.7,8 However, newborn body composition is a useful reflection of the fetal environment in this context,9,10 and likely mediates developmental effects on long-term cardiometabolic outcomes.7,11,12 Thus a better understanding of how maternal lifestyle is related to newborn body composition could provide important insights into the developmental origins of adult health and disease. Some of the challenges in measuring fat and fat-free mass in infants have recently lessened,13 facilitating relatively large population-based studies of infant body composition.14–18 Another challenge is how to best model the impact of one or more predictors on a single continuous anthropometric measure where we expect distinct aetiological mechanisms at opposite ends of the measure’s distribution (e.g. the causes of low birthweight versus those of macrosomia). This is typically approximated by categorizing the otherwise continuous variable and investigating it using binary or multinomial logistic regression. This approach is useful to the degree that the categorisation is clinically relevant, but it ignores the continuous nature of the variable and inevitably discards useful information about the rank order of individuals in the sample. To avoid this, we used quantile regression19 to estimate associations between maternal lifestyle factors and offspring neonatal fat mass and fat-free mass in one of the largest relevant studies to date. Quantile regression is analogous to multiple linear regression, but instead of modelling the mean of the dependent variable conditional on the predictor(s), it can be used to similarly model any centile. Importantly, quantile regression allows you to estimate how the tails of a variable’s distribution (such as the 5th or 95th centile) vary across levels of a predictor, which can occur even when there is no obvious shift in the mean or median. This allowed us to test the hypothesis that infants born to mothers exhibiting healthier lifestyle factors would be less likely to have very high or low levels of fat and fat-free mass. Methods Ethical standards The authors assert that all procedures contributing to this work comply with the International Ethical Guidelines for Epidemiological Studies (CIOMS/WHO) and with the Helsinki Declaration of 1975, as revised in 2008, and has been approved by the Research Ethics Committee of the Cork Teaching Hospitals provided for the SCOPE-Ireland study [ref: ECM5(10)05/02/08]. Study and sample Data are from the Cork site of the Screening for Pregnancy Endpoints (SCOPE) pregnancy cohort study (ACTRN12607000551493), and its follow-up birth cohort study, Babies after SCOPE: Evaluating the Longitudinal Impact on Neurological and Nutritional Endpoints (BASELINE; ClinicalTrials.gov NCT 01498965; www.birthcohorts.net). The primary aim of SCOPE was to identify clinical factors and biomarkers that were predictive of pre-eclampsia, small for gestational age babies and spontaneous preterm birth. Based on this aim, the study included healthy, nulliparous women with singleton pregnancies, and the exclusion criteria were: known major fetal anomalies; pre-pregnancy essential hypertension; moderate to severe hypertension at booking; pre-existing diabetes; renal disease; systemic lupus erythematosus; antiphospholipid syndrome; HIV positivity; major uterine anomaly; cervical suture; knife cone biopsy; ruptured membranes at recruitment; three or more miscarriages; three or more terminations; long-term steroid use; and treatment with low-dose aspirin, calcium (>1000 g/24 h) or Vitamin E (≥400 iu), low-molecular-weight heparin, fish oil or antioxidants. All women who participated in the SCOPE study were informed about the birth cohort, and if consent was obtained infants were registered to the Cork BASELINE birth cohort. Participants were recruited from Cork University Maternity Hospital between February 2008 and February 2011, at 15 ± 1 weeks of gestation. During this period, 2579 nulliparous women were invited to participate and 1774 (69%) consented to do so. Study participants were interviewed and examined by a trained research midwife at recruitment and again at 20 ± 1 weeks of gestation. Demographic, socioeconomic and medical data were collected, as well as information on current diet, physical activity and other lifestyle factors. Of these 1774 women, 1537 went on to have an infant enrolled into the BASELINE birth cohort (87%). Infant measures were taken by a trained research midwife within 72 h of birth. All data were managed using an internet accessible database with a clear audit trail and automated quality assurance procedures (MedSciNet AB, Stockholm, Sweden). Additional details on the study’s methods have been previously reported.20,21 Measurements and variable definitions Infant measures Infant measures used in this analysis were body composition, length, weight, gestational age at delivery and sex. Per study protocol, the aim was to measure body composition within 48 h. First, newborn body density was calculated using weight (measured with an electronic scale to the nearest gram), divided by newborn body volume (estimated by air-displacement plethysmography with a PEA POD® Infant Body Composition System, COSMED USA, Concord, CA), giving body density. Based on a two-compartment model of body composition (fat and fat-free mass) and body density values from Fomon,22 percent body fat was calculated, and was in turn used to estimate fat and fat-free mass in grams, which were the primary outcomes of interest. Gestational age was based on expected date of delivery, which was estimated from respondent recall of last menstrual period (LMP). If the respondent was uncertain about their LMP, or it differed substantially from a 16- or 20-week scan (by ≥7 or ≥10 days, respectively), the earliest available scan date was used. Length was measured with a neonatometer to the nearest millimetre. Maternal lifestyle factors The maternal lifestyle factors we considered were those that most closely aligned with current pregnancy recommendations regarding nutrition and lifestyle in Ireland.23 Pre-pregnancy weight (kg) was estimated as weight measured at recruitment, less 1.25 kg [the assumed average amount of weight gain in the first 15 weeks of pregnancy based on the 2009 Institute of Medicine (IOM) guidelines].24 Pre-pregnancy body mass index (BMI) was calculated from pre-pregnancy weight (kg) divided by measured height squared (m2). Pre-pregnancy weight status was subsequently categorised based on World Health Organization guidelines as underweight (BMI < 18.5 kg/m2), normal weight (BMI 18.5 to 25 kg/m2), overweight (BMI ≥ 25 to 30 kg/m2) and obese (BMI ≥ 30 kg/m2). In addition to the weight measurement at recruitment, weight was also typically measured multiple times across the pregnancy during routine care. To account for these differences, the weight gain rate (kg/week) was calculated as the difference between the last of these measures (whenever it occurred, usually late in the third trimester) and the initial weight measured at recruitment, divided by the number of weeks between those measures. Excessive gestational weight gain (GWG) was defined according to Institute of Medicine (IOM) guidelines as a weight gain rate exceeding 0.5 kg/week (wk) in underweight and normal weight women, 0.33 kg/wk in overweight women, and 0.27 kg/wk in obese women; and inadequate GWG was defined as a weight gain rate below 0.35 kg/wk in underweight and normal weight women, 0.23 kg/wk in overweight women, and 0.17 kg/wk in obese women.24,25 Women were asked at recruitment (15 ± 1 weeks of gestation) how many times each week they engaged in exercise that that did not result in heavier breathing, which was the study’s definition of moderate-intensity exercise. Their responses were categorized as Never, Some (1 to 3 times a week) and Often (4 or more times a week). Respondents also reported daily hours of television viewing in the past month, a commonly used marker of sedentary activity26 particularly in women,27 which was categorised as <2 h, 2 to 4 h and 5 or more h. Based on participant-reported consumption at 15 ± 1 and/or 20 ± 1 weeks of gestation, alcohol use was categorised as Never, Quit before Pregnancy, Quit during Pregnancy and Still Drinking; and smoking was categorised as Never Smoked, Quit during Pregnancy and Still Smoking. Women were asked about pre-pregnancy folic-acid supplementation, and their responses were dichotomized as those meeting the recommended 400 µg versus those who did not (Yes versus No). The questionnaire administered at recruitment asked women to report the frequency with which they consumed several food items in the first 15 weeks of pregnancy. Their responses were used to determine whether they were meeting the recommended five servings of fruit and veg per day (Yes versus No) and at least 1 serving of oily fish per week (Yes versus No). Covariates Available covariates likely to influence both maternal lifestyle and birth outcomes were selected based on the expert opinion of the study authors, who were also careful to not select covariates that were probable consequences of maternal lifestyle. All selected covariates were assessed at 15 ± 1 weeks’ gestation, and included: maternal age (years); the mother’s reported weight at birth (g); her gravidity (1 versus > 1); her ethnicity (White versus Non-white); whether she had a partner or not (Yes versus No); whether she had any third level education (Yes versus No); whether she used Public versus Private maternity care; her socioeconomic index (SEI), based on the New Zealand SEI28 (with higher values reflecting higher social status); her risk of depression based on the Edinburgh Postnatal Depression Score29,30 (Unlikely to experience depression versus At risk of depression in the next year versus Likely depressed); and her score on the Perceived Stress Scale31 (with higher scores reflecting greater stress). Statistical methods Categorical variables were described by the count and proportion in each category. Continuous variables were described by: their mean and standard deviation; their median and the interquartile range; and their full range.32 Relationships between maternal lifestyle factors and infant outcomes were estimated using quantile regression.19 We first estimated the crude association of each lifestyle factor with each outcome, at every centile from the 2nd to 98th. We then estimated a similar set of fully adjusted models with each outcome regressed on all the lifestyle factors and covariates. The latter were included to account for possible confounding. Based on previous research,33 we also tested for an interaction between maternal pre-pregnancy weight status and IOM-classified healthy GWG. All continuous covariates were centred at their means. Quantile regression coefficients were estimated using a modified version of the Barrodale and Roberts algorithm,34 and standard errors (SE) for coefficients were calculated using the kernel-based method suggested by Powell.35 Missing data were handled using multiple imputation. Thirty imputed datasets were created, after a burn-in of 30 replications, using predicted mean matching.36 The imputation model included all variables included in this analysis, and allowed for non-linear relationships using restricted cubic splines with five knots. We took the’ transform then impute’ approach recommended by von Hippel37 to impute variables derived from other variables with missing values. All models were estimated using each imputed dataset, and parameter estimates were combined using Rubin’s rules.38 Differences in proportions or means across sub-groups with and without missing data were tested using, respectively, Pearson’s chi-square test or Welsh’s t-test with unequal variances. Distributions of imputed values were examined visually. All analyses were conducted using the R Project for Statistical Computing39 (version 3.1.2). Quantile regression models were estimated using the quantreg package40 (version 5.11). Multiple imputation was implemented with the Hmisc package41 (version 3.14–6). All plots were produced using the ggplot2 package42 (version 1.0.0). Results Of the 1774 recruited mothers, three experienced a fetal loss before 20 weeks’ gestation; five pregnancies resulted in stillbirths; and 12 infants were born before 32 weeks’ completed gestation; these were excluded from the final analytical sample of 1754 infants. Variable distributions are described in Tables 1 and 2. The infants’ anthropometrics and gestational ages were consistent with established norms.16 Two-thirds of women had a healthy pre-pregnancy weight (BMI < 25 kg/m2), but only 16% experienced a healthy level of GWG. Whereas 16% of women said they were still consuming alcohol at recruitment, 80% of these women reported drinking one or fewer units per week. Based on the results from the fully adjusted quantile regressions (Tables 3–6), the conditional median for fat-free mass was 112 g less (95% CI −176 to −47) in infants born to women who experienced healthy GWG, compared with those who experienced excessive GWG. This reduction was less extreme at the lower centiles of fat-free mass (Figure 1a and Table 3). Healthy GWG was also associated with a 33 g reduction (95% CI −65 to −1) in median fat mass, and a 103 g reduction (95% CI −174 to −33) at the 95th centile of fat mass (Figure 1b and Table 4). Birth lengths were roughly 0.6 cm less at all centiles in infants born to women who experienced healthy GWG, though 95% CIs at several centiles included the null hypothesis of no difference (Figure 1c). Table 1. Characteristics of 1754 sample mother-infant pairs enrolled in SCOPE-Ireland, 2008 to 2011 Variable  Missing Values  Proportion (n)  Mean (SD)  Median[IQR]  Range  Infant characteristics            Sex  0           Male    0.51(892)         Female    0.49(862)        Birth weight (g)  0    3462 (507.5)  3460 [3150 to 3778]  1200 to 5130  Fat mass (g)  512    378 (172.9)  351 [253 to 481]  36 to 1099  Fat free mass (g)  513    2955 (346.8)  2965 [2730 to 3182]  1848 to 3960  Percent fat mass  514    11.1 (4.1)  10.9 [8.2 to 13.8]  1.3 to 30.1  Length (cm)  59    50.2 (2.4)  50.2 [49 to 51.8]  37.5 to 57  Gestational age (wks)  0    40 (1.5)  40.3 [39.3 to 41]  32 to 42.6  Maternal characteristics  Age  0    29.9 (4.5)  30 [28 to 33]  17 to 45  Height  0    164.6 (5.9)  165 [161 to 168]  147 to 185  Birth weight  60    3360.8 (532.9)  3374 [3062 to 3657]  624 to 6000  Gravidity  0           1    0.85(1483)         2+    0.15 (271)        Ethnicity  0           White    0.98(1712)         Nonwhite    0.02(42)        Has partner  0           Single    0.11(186)         Partner    0.89(1568)        3rd level education  0           No    0.11(195)         Yes    0.89(1559)        SEI†  0    42.7 (16)  45 [29 to 51]  18 to 89  Maternity care  0           Public    0.75(1318)         Private    0.25(436)        Depressed  0           Unlikely    0.41(711)         At risk    0.35(622)         Likely    0.24(421)        Stress score ††  0    13.7 (6.6)  13 [9 to 18]  0 to 35  Variable  Missing Values  Proportion (n)  Mean (SD)  Median[IQR]  Range  Infant characteristics            Sex  0           Male    0.51(892)         Female    0.49(862)        Birth weight (g)  0    3462 (507.5)  3460 [3150 to 3778]  1200 to 5130  Fat mass (g)  512    378 (172.9)  351 [253 to 481]  36 to 1099  Fat free mass (g)  513    2955 (346.8)  2965 [2730 to 3182]  1848 to 3960  Percent fat mass  514    11.1 (4.1)  10.9 [8.2 to 13.8]  1.3 to 30.1  Length (cm)  59    50.2 (2.4)  50.2 [49 to 51.8]  37.5 to 57  Gestational age (wks)  0    40 (1.5)  40.3 [39.3 to 41]  32 to 42.6  Maternal characteristics  Age  0    29.9 (4.5)  30 [28 to 33]  17 to 45  Height  0    164.6 (5.9)  165 [161 to 168]  147 to 185  Birth weight  60    3360.8 (532.9)  3374 [3062 to 3657]  624 to 6000  Gravidity  0           1    0.85(1483)         2+    0.15 (271)        Ethnicity  0           White    0.98(1712)         Nonwhite    0.02(42)        Has partner  0           Single    0.11(186)         Partner    0.89(1568)        3rd level education  0           No    0.11(195)         Yes    0.89(1559)        SEI†  0    42.7 (16)  45 [29 to 51]  18 to 89  Maternity care  0           Public    0.75(1318)         Private    0.25(436)        Depressed  0           Unlikely    0.41(711)         At risk    0.35(622)         Likely    0.24(421)        Stress score ††  0    13.7 (6.6)  13 [9 to 18]  0 to 35  IQR, interquartile range; SD, standard deviation; SEI, Socioeconomic index. †Based on the New Zealand socioeconomic index, with higher values reflecting greater social status. ††Out of a maximum score of 40, with higher scores reflecting higher levels of stress. Table 2. Maternal lifestyle factors in 1754 sample mothers enrolled in SCOPE-Ireland, 2008 to 2011 Variable  Missing values  Proportion (n)  Prepregnancy body size  0     Obese (BMI ≥ 30 kg/m2)    0.11 (190)   Overweight (BMI 25 to 30 kg/m2)    0.24 (419)   Normal weight (BMI < 25 kg/m2)    0.65 (1145)  IOM defined gestational weight gain level  525     Excessive    0.79 (977)   Healthy    0.16 (199)   Inadequate    0.04 (53)  Frequency of moderate intensity exercise  0     None    0.25 (441)   Some    0.55 (965)   Often    0.2 (348)  Amount of daily TV viewing  0     ≥5 h    0.09 (158)   2–4 h    0.55 (958)   <2 h    0.36 (638)  Alcohol use  0     Still drinks    0.16 (288)   Quit during pregnancy    0.65 (1133)   Quit prepregnancy    0.09 (166)   Never drank    0.1 (167)  Any smoking  0     Still smokes    0.1 (174)   Quit during pregnancy    0.18 (307)   Never smoked†    0.73 (1273)  Takes folate  0     No    0.32 (560)   Yes    0.68 (1194)  Eats ≥ 5 servings fruit and veg per day  0     No    0.86 (1508)   Yes    0.14 (246)  Eats ≥ 1 serving oily fish per week  0     No    0.69 (1205)   Yes    0.31 (549)  Variable  Missing values  Proportion (n)  Prepregnancy body size  0     Obese (BMI ≥ 30 kg/m2)    0.11 (190)   Overweight (BMI 25 to 30 kg/m2)    0.24 (419)   Normal weight (BMI < 25 kg/m2)    0.65 (1145)  IOM defined gestational weight gain level  525     Excessive    0.79 (977)   Healthy    0.16 (199)   Inadequate    0.04 (53)  Frequency of moderate intensity exercise  0     None    0.25 (441)   Some    0.55 (965)   Often    0.2 (348)  Amount of daily TV viewing  0     ≥5 h    0.09 (158)   2–4 h    0.55 (958)   <2 h    0.36 (638)  Alcohol use  0     Still drinks    0.16 (288)   Quit during pregnancy    0.65 (1133)   Quit prepregnancy    0.09 (166)   Never drank    0.1 (167)  Any smoking  0     Still smokes    0.1 (174)   Quit during pregnancy    0.18 (307)   Never smoked†    0.73 (1273)  Takes folate  0     No    0.32 (560)   Yes    0.68 (1194)  Eats ≥ 5 servings fruit and veg per day  0     No    0.86 (1508)   Yes    0.14 (246)  Eats ≥ 1 serving oily fish per week  0     No    0.69 (1205)   Yes    0.31 (549)  BMI, body mass index; IOM, Institute of Medicine. †Six women who reported quitting prior to pregnancy were classified as Never Smoked. Table 3. Quantile regression results from the fully adjusted model for fat-free mass (g), n = 1754   Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  2444.4  (2153.9 to 2734.9)  3011.4  (2864.3 to 3158.5)  3587.9  (3387.9 to 3787.9)  Healthy GWG  −38.5  (−173.1 to 96.1)  −111.7  (−176.2 to −47.2)  −109.8  (−186.7 to −32.9)  Inadequate GWG  34.7  (−181.2 to 250.6)  −57  (−165.6 to 51.6)  −83.2  (−240.3 to 73.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −166  (−295.8 to −36.2)  −35  (−100.2 to 30.2)  −121.6  (−249.8 to 6.6)  Overweight (BMI 25 to 30 kg/m2)  −123.3  (−271.5 to 24.9)  −8.4  (−81.1 to 64.3)  −90.6  (−223.5 to 42.3)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −1.7  (−110.5 to 107.1)  34.4  (−20.6 to 89.4)  −6  (−76.3 to 64.3)  Some moderate-intensity exercise  75.7  (−58.6 to 210)  38.1  (−15.1 to 91.3)  15.3  (−59.3 to 89.9)  Frequent moderate-intensity exercise  16.1  (−127.8 to 160)  −5.8  (−74.5 to 62.9)  −12.5  (−104 to 79)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  21.2  (−147 to 189.4)  −24.2  (−103.8 to 55.4)  −75.9  (−176.9 to 25.1)  < 2 hours of television  12.5  (−161 to 186)  −28.2  (−114.4 to 58)  −65.6  (−175.7 to 44.5)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  29.7  (−108.7 to 168.1)  4.1  (−56.5 to 64.7)  5.7  (−88.8 to 100.2)  Quit drinking prepregnancy  26.2  (−169 to 221.4)  −44.1  (−130.2 to 42)  −4.6  (−152.5 to 143.3)  Never drank  38.3  (−145 to 221.6)  36.2  (−63.8 to 136.2)  −51.7  (−156.8 to 53.4)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  165.5  (−14.8 to 345.8)  47.7  (−45.6 to 141)  60.1  (−57.2 to 177.4)  Never smoked  130.2  (−39.6 to 300)  81  (−5.3 to 167.3)  78.5  (−22.5 to 179.5)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  7  (−117.7 to 131.7)  10.8  (−55.2 to 76.8)  54.6  (−30.6 to 139.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −52.6  (−164.7 to 59.5)  18.4  (−28.8 to 65.6)  69.8  (−2.7 to 142.3)    Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  2444.4  (2153.9 to 2734.9)  3011.4  (2864.3 to 3158.5)  3587.9  (3387.9 to 3787.9)  Healthy GWG  −38.5  (−173.1 to 96.1)  −111.7  (−176.2 to −47.2)  −109.8  (−186.7 to −32.9)  Inadequate GWG  34.7  (−181.2 to 250.6)  −57  (−165.6 to 51.6)  −83.2  (−240.3 to 73.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −166  (−295.8 to −36.2)  −35  (−100.2 to 30.2)  −121.6  (−249.8 to 6.6)  Overweight (BMI 25 to 30 kg/m2)  −123.3  (−271.5 to 24.9)  −8.4  (−81.1 to 64.3)  −90.6  (−223.5 to 42.3)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −1.7  (−110.5 to 107.1)  34.4  (−20.6 to 89.4)  −6  (−76.3 to 64.3)  Some moderate-intensity exercise  75.7  (−58.6 to 210)  38.1  (−15.1 to 91.3)  15.3  (−59.3 to 89.9)  Frequent moderate-intensity exercise  16.1  (−127.8 to 160)  −5.8  (−74.5 to 62.9)  −12.5  (−104 to 79)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  21.2  (−147 to 189.4)  −24.2  (−103.8 to 55.4)  −75.9  (−176.9 to 25.1)  < 2 hours of television  12.5  (−161 to 186)  −28.2  (−114.4 to 58)  −65.6  (−175.7 to 44.5)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  29.7  (−108.7 to 168.1)  4.1  (−56.5 to 64.7)  5.7  (−88.8 to 100.2)  Quit drinking prepregnancy  26.2  (−169 to 221.4)  −44.1  (−130.2 to 42)  −4.6  (−152.5 to 143.3)  Never drank  38.3  (−145 to 221.6)  36.2  (−63.8 to 136.2)  −51.7  (−156.8 to 53.4)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  165.5  (−14.8 to 345.8)  47.7  (−45.6 to 141)  60.1  (−57.2 to 177.4)  Never smoked  130.2  (−39.6 to 300)  81  (−5.3 to 167.3)  78.5  (−22.5 to 179.5)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  7  (−117.7 to 131.7)  10.8  (−55.2 to 76.8)  54.6  (−30.6 to 139.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −52.6  (−164.7 to 59.5)  18.4  (−28.8 to 65.6)  69.8  (−2.7 to 142.3)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 4. Quantile regression results from the fully adjusted model for fat mass (g), n = 1754   Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  121.7  (33.7 to 209.7)  371.5  (298.2 to 444.8)  845  (670.3 to 1019.7)  Healthy GWG  −4.5  (−41.2 to 32.2)  −33.4  (−65.5 to −1.3)  −103.2  (−173.8 to −32.6)  Inadequate GWG  27  (−33.7 to 87.7)  −17.1  (−66.6 to 32.4)  −30.5  (−134.4 to 73.4)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −42.2  (−86 to 1.6)  −47.7  (−83.7 to −11.7)  −69.1  (−149.9 to 11.7)  Overweight (BMI 25 to 30 kg/m2)  −36.1  (−87.4 to 15.2)  −25.6  (−64.5 to 13.3)  −39.9  (−128.5 to 48.7)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −8.4  (−38.8 to 22)  8.6  (−18.1 to 35.3)  −16.6  (−74.8 to 41.6)  Some moderate-intensity exercise  9.8  (−23.1 to 42.7)  4.8  (−21.7 to 31.3)  −36.7  (−111.3 to 37.9)  Frequent moderate-intensity exercise  −4.1  (−46.4 to 38.2)  −0.6  (−33.2 to 32)  −91.9  (−168 to −15.8)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  −4.3  (−48.3 to 39.7)  12.6  (−27.9 to 53.1)  51.5  (−26.7 to 129.7)  < 2 hours of television  −25  (−67.8 to 17.8)  −6.8  (−50.9 to 37.3)  57.6  (−31.4 to 146.6)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  2.2  (−38.3 to 42.7)  −26.8  (−58.6 to 5)  −32.6  (−100 to 34.8)  Quit drinking prepregnancy  −3.3  (−69.1 to 62.5)  −22.4  (−67.4 to 22.6)  −88.9  (−191.3 to 13.5)  Never drank  18.2  (−40.1 to 76.5)  −14.3  (−60.1 to 31.5)  −122.3  (−204.1 to −40.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  17.1  (−33.6 to 67.8)  1.9  (−43 to 46.8)  −17.1  (−116.3 to 82.1)  Never smoked  28.8  (−17 to 74.6)  29.1  (−11.7 to 69.9)  29.6  (−54.8 to 114)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  5.4  (−34.1 to 44.9)  6.9  (−26.1 to 39.9)  31  (−38.4 to 100.4)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.7  (−30.6 to 27.2)  −4  (−27.3 to 19.3)  −21.7  (−69.5 to 26.1)    Centile     5th   50th   95th   Variable  β (g)  95% CI  β (g)  95% CI  β (g)  95% CI  Intercept  121.7  (33.7 to 209.7)  371.5  (298.2 to 444.8)  845  (670.3 to 1019.7)  Healthy GWG  −4.5  (−41.2 to 32.2)  −33.4  (−65.5 to −1.3)  −103.2  (−173.8 to −32.6)  Inadequate GWG  27  (−33.7 to 87.7)  −17.1  (−66.6 to 32.4)  −30.5  (−134.4 to 73.4)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −42.2  (−86 to 1.6)  −47.7  (−83.7 to −11.7)  −69.1  (−149.9 to 11.7)  Overweight (BMI 25 to 30 kg/m2)  −36.1  (−87.4 to 15.2)  −25.6  (−64.5 to 13.3)  −39.9  (−128.5 to 48.7)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −8.4  (−38.8 to 22)  8.6  (−18.1 to 35.3)  −16.6  (−74.8 to 41.6)  Some moderate-intensity exercise  9.8  (−23.1 to 42.7)  4.8  (−21.7 to 31.3)  −36.7  (−111.3 to 37.9)  Frequent moderate-intensity exercise  −4.1  (−46.4 to 38.2)  −0.6  (−33.2 to 32)  −91.9  (−168 to −15.8)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  −4.3  (−48.3 to 39.7)  12.6  (−27.9 to 53.1)  51.5  (−26.7 to 129.7)  < 2 hours of television  −25  (−67.8 to 17.8)  −6.8  (−50.9 to 37.3)  57.6  (−31.4 to 146.6)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  2.2  (−38.3 to 42.7)  −26.8  (−58.6 to 5)  −32.6  (−100 to 34.8)  Quit drinking prepregnancy  −3.3  (−69.1 to 62.5)  −22.4  (−67.4 to 22.6)  −88.9  (−191.3 to 13.5)  Never drank  18.2  (−40.1 to 76.5)  −14.3  (−60.1 to 31.5)  −122.3  (−204.1 to −40.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  17.1  (−33.6 to 67.8)  1.9  (−43 to 46.8)  −17.1  (−116.3 to 82.1)  Never smoked  28.8  (−17 to 74.6)  29.1  (−11.7 to 69.9)  29.6  (−54.8 to 114)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  5.4  (−34.1 to 44.9)  6.9  (−26.1 to 39.9)  31  (−38.4 to 100.4)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.7  (−30.6 to 27.2)  −4  (−27.3 to 19.3)  −21.7  (−69.5 to 26.1)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 5. Quantile regression results from the fully adjusted model for birth length (cm), n = 1754   Centile     5th   50th   95th   Variable  β (cm)  95% CI  β (cm)  95% CI  β (cm)  95% CI  Intercept  (42.8 to 47)  50.5  (49.5 to 51.5)  53.7  (52.4 to 55)    Healthy GWG  −0.5  (−1.5 to 0.5)  −0.4  (−0.8 to 0)  −0.8  (−1.3 to −0.3)  Inadequate GWG  0.1  (−2.8 to 3)  −0.3  (−1 to 0.4)  −0.1  (−1.1 to 0.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −1.2  (−1.9 to −0.5)  −0.4  (−0.9 to 0.1)  −0.5  (−1.1 to 0.1)  Overweight (BMI 25 to 30 kg/m2)  −0.8  (−1.7 to 0.1)  −0.3  (−0.8 to 0.2)  −0.4  (−1 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  0.4  (−0.5 to 1.3)  0  (−0.3 to 0.3)  0.2  (−0.2 to 0.6)  Some moderate-intensity exercise  0.7  (−0.1 to 1.5)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  Frequent moderate-intensity exercise  −0.2  (−1.2 to 0.8)  −0.3  (−0.7 to 0.1)  −0.2  (−0.8 to 0.4)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  2.2  (0.3 to 4.1)  0  (−0.6 to 0.6)  −0.6  (−1.3 to 0.1)  < 2 hours of television  1.6  (−0.3 to 3.5)  0  (−0.6 to 0.6)  −0.5  (−1.2 to 0.2)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  0  (−0.9 to 0.9)  0.1  (−0.3 to 0.5)  0.2  (−0.2 to 0.6)  Quit drinking prepregnancy  −0.2  (−1.3 to 0.9)  −0.3  (−0.8 to 0.2)  −0.3  (−1.1 to 0.5)  Never drank  −0.9  (−2.9 to 1.1)  0  (−0.5 to 0.5)  −0.3  (−0.9 to 0.3)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  1.1  (0.2 to 2)  0.5  (0 to 1)  0.5  (−0.2 to 1.2)  Never smoked  0.6  (−0.3 to 1.5)  0.6  (0.1 to 1.1)  0.7  (0 to 1.4)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0.2  (−0.6 to 1)  0.2  (−0.2 to 0.6)  0.3  (−0.2 to 0.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.1  (−1.8 to −0.4)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)    Centile     5th   50th   95th   Variable  β (cm)  95% CI  β (cm)  95% CI  β (cm)  95% CI  Intercept  (42.8 to 47)  50.5  (49.5 to 51.5)  53.7  (52.4 to 55)    Healthy GWG  −0.5  (−1.5 to 0.5)  −0.4  (−0.8 to 0)  −0.8  (−1.3 to −0.3)  Inadequate GWG  0.1  (−2.8 to 3)  −0.3  (−1 to 0.4)  −0.1  (−1.1 to 0.9)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  −1.2  (−1.9 to −0.5)  −0.4  (−0.9 to 0.1)  −0.5  (−1.1 to 0.1)  Overweight (BMI 25 to 30 kg/m2)  −0.8  (−1.7 to 0.1)  −0.3  (−0.8 to 0.2)  −0.4  (−1 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  0.4  (−0.5 to 1.3)  0  (−0.3 to 0.3)  0.2  (−0.2 to 0.6)  Some moderate-intensity exercise  0.7  (−0.1 to 1.5)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  Frequent moderate-intensity exercise  −0.2  (−1.2 to 0.8)  −0.3  (−0.7 to 0.1)  −0.2  (−0.8 to 0.4)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  2.2  (0.3 to 4.1)  0  (−0.6 to 0.6)  −0.6  (−1.3 to 0.1)  < 2 hours of television  1.6  (−0.3 to 3.5)  0  (−0.6 to 0.6)  −0.5  (−1.2 to 0.2)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  0  (−0.9 to 0.9)  0.1  (−0.3 to 0.5)  0.2  (−0.2 to 0.6)  Quit drinking prepregnancy  −0.2  (−1.3 to 0.9)  −0.3  (−0.8 to 0.2)  −0.3  (−1.1 to 0.5)  Never drank  −0.9  (−2.9 to 1.1)  0  (−0.5 to 0.5)  −0.3  (−0.9 to 0.3)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  1.1  (0.2 to 2)  0.5  (0 to 1)  0.5  (−0.2 to 1.2)  Never smoked  0.6  (−0.3 to 1.5)  0.6  (0.1 to 1.1)  0.7  (0 to 1.4)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0.2  (−0.6 to 1)  0.2  (−0.2 to 0.6)  0.3  (−0.2 to 0.8)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −1.1  (−1.8 to −0.4)  0.1  (−0.2 to 0.4)  0  (−0.4 to 0.4)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Table 6. Quantile regression results from the fully adjusted model for gestational age (weeks), n = 1754   Centile     5th   50th   95th   Variable  β (weeks)  95% CI  β (weeks)  95% CI  β (weeks)  95% CI  Intercept  36  (34.2 to 37.8)  40.7  (40.2 to 41.2)  41.6  (41.3 to 41.9)  Healthy GWG  0.1  (−1 to 1.2)  0  (−0.2 to 0.2)  0  (−0.2 to 0.2)  Inadequate GWG  0.8  (−0.4 to 2)  −0.1  (−0.7 to 0.5)  0  (−0.3 to 0.3)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  0  (−0.9 to 0.9)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Overweight (BMI 25 to 30 kg/m2)  0.2  (−0.8 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −0.2  (−0.9 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  Some moderate-intensity exercise  0.8  (0 to 1.6)  0.2  (0 to 0.4)  0  (−0.1 to 0.1)  Frequent moderate-intensity exercise  −0.1  (−1.5 to 1.3)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  0.5  (−0.6 to 1.6)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  < 2 hours of television  0.3  (−0.8 to 1.4)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  −0.4  (−1.1 to 0.3)  −0.2  (−0.4 to 0)  0.1  (0 to 0.2)  Quit drinking prepregnancy  −0.8  (−2.5 to 0.9)  −0.4  (−0.7 to −0.1)  0  (−0.2 to 0.2)  Never drank  0  (−0.9 to 0.9)  −0.1  (−0.5 to 0.3)  0.2  (−0.1 to 0.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  0.5  (−0.5 to 1.5)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Never smoked  0.3  (−0.6 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0  (−0.7 to 0.7)  −0.1  (−0.3 to 0.1)  0  (−0.2 to 0.2)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −0.1  (−0.7 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)    Centile     5th   50th   95th   Variable  β (weeks)  95% CI  β (weeks)  95% CI  β (weeks)  95% CI  Intercept  36  (34.2 to 37.8)  40.7  (40.2 to 41.2)  41.6  (41.3 to 41.9)  Healthy GWG  0.1  (−1 to 1.2)  0  (−0.2 to 0.2)  0  (−0.2 to 0.2)  Inadequate GWG  0.8  (−0.4 to 2)  −0.1  (−0.7 to 0.5)  0  (−0.3 to 0.3)  Excessive GWG  ref  –  ref  –  ref  –  Normal weight (BMI < 25 kg/m2)  0  (−0.9 to 0.9)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Overweight (BMI 25 to 30 kg/m2)  0.2  (−0.8 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Obese (BMI ≥ 30 kg/m2)  ref  –  ref  –  ref  –  Takes folate (≥400 mg) (vs. not)  −0.2  (−0.9 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  Some moderate-intensity exercise  0.8  (0 to 1.6)  0.2  (0 to 0.4)  0  (−0.1 to 0.1)  Frequent moderate-intensity exercise  −0.1  (−1.5 to 1.3)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  No moderate-intensity exercise  ref  –  ref  –  ref  –  2 to 4 hours of television  0.5  (−0.6 to 1.6)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  < 2 hours of television  0.3  (−0.8 to 1.4)  0  (−0.3 to 0.3)  0.1  (−0.1 to 0.3)  4+ hours of television  ref  –  ref  –  ref  –  Quit drinking during pregnancy  −0.4  (−1.1 to 0.3)  −0.2  (−0.4 to 0)  0.1  (0 to 0.2)  Quit drinking prepregnancy  −0.8  (−2.5 to 0.9)  −0.4  (−0.7 to −0.1)  0  (−0.2 to 0.2)  Never drank  0  (−0.9 to 0.9)  −0.1  (−0.5 to 0.3)  0.2  (−0.1 to 0.5)  Still drinks  ref  –  ref  –  ref  –  Quit smoking during pregnancy  0.5  (−0.5 to 1.5)  0  (−0.3 to 0.3)  0  (−0.2 to 0.2)  Never smoked  0.3  (−0.6 to 1.2)  0.1  (−0.2 to 0.4)  0  (−0.2 to 0.2)  Still smokes  ref  –  ref  –  ref  –  Eats 5 fruit/veg a day (vs. not)  0  (−0.7 to 0.7)  −0.1  (−0.3 to 0.1)  0  (−0.2 to 0.2)  Eats ≥ 1 serving of oily fish weekly (vs. not)  −0.1  (−0.7 to 0.5)  0  (−0.2 to 0.2)  0  (−0.1 to 0.1)  BMI, body mass index; CI, confidence interval; GWG, gestational weight gain. Models further adjusted for infant sex, maternal age, maternal height, gravidity, ethnicity, whether the mother has a partner, maternal education, socioeconomic index, private/public maternity care, risk of depression, and stress score. Figure 1 View largeDownload slide Healthy gestational weight gain. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for healthy (versus excessive) gestational weight gain at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Panel a shows that almost the entire distribution of fat-free mass is shifted to the left (towards smaller values) among infants born to women with healthy gestational weight gain (versus not), whereas Panel b shows that the right tail of the distribution of fat mass is being pulled in, with little change in the left tail of the distribution. Figure 1 View largeDownload slide Healthy gestational weight gain. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for healthy (versus excessive) gestational weight gain at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Panel a shows that almost the entire distribution of fat-free mass is shifted to the left (towards smaller values) among infants born to women with healthy gestational weight gain (versus not), whereas Panel b shows that the right tail of the distribution of fat mass is being pulled in, with little change in the left tail of the distribution. Pre-pregnancy normal weight status, compared with women classified as obese, was associated with a 48 g reduction (95% CI −84 to −12) in median fat mass, though 95% CIs at several centiles included the null hypothesis of no difference (Figure 2b and Table 4). Pre-pregnancy weight status was otherwise unrelated to outcomes. Frequent bouts of moderate-intensity exercise were associated with a reduction in the upper tail of the fat mass distribution (Figure 3a). For example, the 95th centile of fat mass in infants born to women who exercised frequently was 92 g less (95% CI −168 to −16) than in infants born to women who reported never exercising. The upper centiles of fat mass were also reduced in infants born to women who reported never drinking. For example, never drinking was associated with a 122 g reduction (95% CI −204 to −40) in the 95th centile of fat mass (Figure 4 and Table 4). Babies born to women who never smoked had greater fat-free mass and length, at all centiles (Figure 5). Figure 2 View largeDownload slide Maternal pre-pregnancy normal weight. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for maternal pre-pregnancy normal weight (versus obese) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 2 View largeDownload slide Maternal pre-pregnancy normal weight. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for maternal pre-pregnancy normal weight (versus obese) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 3 View largeDownload slide Frequent moderate-intensity exercise. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for frequent (versus not) moderate-intensity exercise at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 3 View largeDownload slide Frequent moderate-intensity exercise. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for frequent (versus not) moderate-intensity exercise at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 4 View largeDownload slide Never drank alcohol. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never drank alcohol (versus still drinking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 4 View largeDownload slide Never drank alcohol. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never drank alcohol (versus still drinking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 5 View largeDownload slide Never smoked. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never smoked (versus still smoking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. Figure 5 View largeDownload slide Never smoked. In each of the four plots above, the enclosed white space depicts the fully adjusted regression coefficients and 95% confidence intervals (values on the y-axis) for never smoked (versus still smoking) at each centile (x-axis) of the dependent variable. The dashed lines similarly reflect the crude estimates and 95% CIs. No other lifestyle factors were strongly related to infant outcomes in the fully adjusted quantile regression models (see Supplementary material, available at IJE online). Inclusion of a product interaction term for pre-pregnancy weight status and healthy GWG did not appreciably improve fit in any of the quantile regression models we estimated (likelihood ratio test P > 0.10 in all cases). All reported models thus excluded this term. A description of missing data is provided in the Supplementary material, available at IJE online. Discussion Using data from a large cohort of pregnant women, we used quantile regression to link maternal lifestyle factors to the distributions of fat and fat-free mass in their newborn offspring. Consistent with previous research (described below), maternal obesity and excess GWG were most strongly associated with newborn body composition. Both are established risk factors for macrosomia,25,43 which supports the idea that the associated long-term cardiometabolic risks result from an excess of substrate available to the fetus, particularly glucose, and subsequent fetal hyperinsulinaemia.44–46 Whereas newborn body composition is likely a more sensitive reflection of these fetal influences than total mass,9,10 relatively few studies have looked at the independent effects of GWG and pre-pregnancy body size on infant fat and fat-free mass. We found that maternal obesity (versus normal weight) was associated with an increase in newborn fat mass, though 95% CIs did not exclude the null hypothesis of zero difference at some centiles. This result is broadly consistent with Hull et al.47 (n = 306), Sewell et al.9,(n = 221), Carlsen et al.48 (n = 311), Au et al.49 (n = 599), Starling et al.50 (n = 826) and Friss et al.51 (n = 207), who found that maternal overweight and/or obesity were associated with increased fat mass and/or percentage fat mass. We also found that IOM-defined excessive GWG was associated with increases in both fat and fat-free mass. This finding was consistent with Au et al.49 who observed that each kg of weight gained during pregnancy was associated with increased percentage fat mass and birthweight, and Carlsen et al.48 and Starling et al.50 who found a similar association with fat and fat-free mass. Crozier et al.52 (n = 564) also found that IOM-healthy GWG was associated with reduced fat mass at birth, though they did not control for maternal pre-pregnancy body size. Friis et al.51 was the only relevant study not to observe an association between GWG and infant fat mass. Sewell et al.9 and Hull et al.47 also found that the apparent association between GWG and infant body composition was modified by maternal weight status, though the nature of this interaction differed between studies. The former observed that weight gain during pregnancy was associated with fat-free mass in the infants born to normal weight women, and with percentage fat mass in infants born to overweight and obese women; the latter found that the increase in percentage fat mass associated with excessive GWG was pronounced in overweight mothers. Our analysis, as well as that by Starling et al.,50 found no evidence of such interactions. Relative to the previous research just described, the unique contribution of our analysis comes from our use of quantile regression and its ability to estimate the effect of predictors on the tails of an outcome’s distribution. This approach complements a previous paper using this sample, which used linear regression to investigate lifestyle predictors of neonatal percentage body fat.53 For example, we found that excessive GWG was clearly associated with higher newborn fat-free mass, and some evidence that it was associated with a small increase in length. Importantly, these associations were consistent across the respective centiles of fat-free mass and length. This is expected, since GWG reflects changes in multiple tissues of the mother, placenta and developing fetus.24 This is of course true for fat mass as well, except that the association between GWG and fat mass was more pronounced at the upper end of the fat mass distribution. Further, the observed change in fat mass at the upper end of its distribution is larger than those seen for fat-free mass and length, relative to their respective means and variances. This could be reflecting that newborn fat mass is more modifiable than infant fat-free mass and length, which are under stronger genetic control.49,51,54,55 Consequently, we suggest that the nature of the GWG-fat mass relationship, which would have perhaps been missed using methods other than quantile regression, is reflecting a pathogenic effect of unhealthy GWG on fetal fat mass accumulation; whereas the reduction in fat-free mass associated with healthy GWG is just reflecting the functional relationship between the two variables. Although maternal smoking is a long-recognized determinant of infant length and total mass, very few studies have investigated the associations of maternal smoking with newborn fat and fat-free mass. Lindsay et al.56 (n = 129) and Spady et al.57 (n = 78) found that maternal smoking was associated with decreased fat-free mass and length, but not fat mass. A larger (n = 916), more recent study also found that neonates exposed to smoking throughout the pregnancy had lower fat-free mass.58 Our results are consistent with these. Quantile regression also allowed us to observe that the highest centiles of fat mass were lower among infants born to women who never drank compared with women who were still drinking during pregnancy. It is important to note that the reported amount of alcohol being drunk by the latter group was low. Further, there was no appreciable difference in infant outcomes between women who were still drinking at recruitment and those who reported quitting before pregnancy. It thus seems likely that the observed association between never drinking and outcomes is at least partly explained by other factors experienced by this relatively small group of women (10% of sample). Similarly, quantile regression revealed that the highest centiles of fat mass were also lower among infants born to women who engaged in frequent moderate-intensity exercise early in the pregnancy. This conflicts with a recent finding59 in a sample of 826 mother-neonate pairs where physical activity in early pregnancy was not associated with fat mass, fat-free mass or birthweight. However, this study did not look at the tails of the fat mass distribution, and a difference like the one we observed could have been obscured in a comparison of mean fat mass values. While we caution against over-interpreting this result, it is encouraging to think that frequent, moderate-intensity activity might help reduce the prevalence of babies born with a very high amount of fat mass. Further, we found no evidence that exercise was associated with reduced fetal growth (reflected in birth length) or gestational age, which might help reassure everyone that moderate-intensity exercise during pregnancy is in fact safe. Strengths This analysis uses data from one of the largest studies of newborn body composition60 measured with a reference method.61 Further, the population-based nature of the study allowed us to estimate the independent associations of a variety of healthy and unhealthy behaviours in a more representative sample than is often possible. Our use of quantile regression yielded insights that would not have been apparent with multiple linear regression. Binary or multinomial logistic regression is another common alternative for looking at the ends of a distribution, but when the outcome is inherently continuous in nature (e.g. mass), one must first impose one or more cut-offs. The logistic regression model treats any set of two values falling on different sides of a cut-off as similarly different, regardless of the distance between two values on the real number line. Thus quantile regression allows us to investigate the influence of a predictor on the tails of an outcome’s distribution while still respecting the underlying continuous nature of that variable. The potential value of quantile regression for epidemiologists has been highlighted by Beyerlein,62 and quantile regression has been previously applied to anthropometric studies across the life course.63 The ideal expression of infant body composition has been previously debated.64 It is common to focus on percentage fat mass,65 but this has limitations.66 Ratio measures, generally, have poor statistical properties.67 Additionally, because fat mass appears in both the numerator and the denominator of percentage fat mass, increases in absolute fat mass become less obvious as percentage fat mass increases. Further, focusing on fat mass can lead us to ignore the fat-free mass,65 which is often just as important to consider.68,69 Finally, though the supposed value of calculating percentage fat mass is to arrive at a measure that is normalized for overall body size, percentage fat mass is typically correlated with height, which was true in our sample (results not shown). Contrary to common practice, we did not restrict the sample to term infants, nor include gestational age in our models of body composition or birth length. Because gestational age does not precede any of the maternal lifestyle factors we have investigated, it cannot confound associations between lifestyle factors and birth size. It could, however, mediate the influence of lifestyle on birth size, and thus adjusting for it could lead to collider bias when gestational age and birth size share other unmeasured causes. This issue has been previously described by Wilcox, Weinberg and Basso.70 Similarly, we also did not adjust models of fat and fat-free mass for length, since it is likely that these will share unmeasured causes that again could confound associations between lifestyle factors and infant body composition once length is adjusted for. Limitations The main weakness of this research is that it is a secondary analysis of data collected with an observational study design. There is thus considerable potential for meaningful confounding and selection bias. These were hopefully minimized by thoughtful selection of covariates based on theoretical grounds. It is worth noting that models were not modified to better fit the data or in light of any preliminary analyses. There were missing data for both outcomes and predictors in the sample. We used multiple imputation to estimate parameters under an assumption of missing at random (MAR), conditional on other observed covariates, which is more defensible than the assumption that data were missing completely at random (MCAR).71 Our imputation models included the outcomes, which is recommended when there are missing values in the predictors.72 However, we should note that the majority of missing values for body composition were because the PEA POD was not yet available on site, and thus those values are likely MCAR. Thus the main value of multiple imputation in this analysis is an increased sample size and a consequent increase in power to detect effects than had we instead used the complete case analysis, or an impute-then-delete approach.73 Most of the data on maternal lifestyle factors were based on self-report and are likely measured with considerable error. This was particularly true for the dietary data, which were based on questions about the consumption of nine selected food items rather than more established methods for dietary assessment, and so our ability to predict infant body composition from the dietary data was likely quite poor from the outset. Similarly, the survey question about moderate-intensity exercise, though consistent with the Centres for Disease Control (CDC) and World Health Organization (WHO) definitions, is quite broad. Importantly, useful information on gestational diabetes mellitus (GDM) was lacking on over two-thirds of the sample and was not considered in this analysis. However, whereas a GDM diagnosis can impact on subsequent lifestyle factors later in the pregnancy, it could not have affected the early/pre-pregnancy factors we have considered here, and thus could not have confounded the associations we report. Conclusion Despite its clear role in the developmental origins of health and disease, little is known about maternal influences on infant body composition. Taken as a whole, this analysis suggests that supporting healthy maternal lifestyles could reduce the risk of excess fat accumulation in the offspring, without increasing the risk of low body fat or adversely affecting fat-free mass development, length, or gestational age. We suggest that future evaluations of maternal lifestyle interventions include more direct infant body composition measures, and use quantile regression to analyse the subsequent data. The use of quantile regression led to insights that would not have been apparent in the data if using more commonly applied linear or logistic regression models. Supplementary Data Supplementary data are available at IJE online. Funding This work was supported by the Irish Centre for Fetal and Neonatal Translational Research (INFANT) and was funded in part by Science Foundation Ireland (grant number 12/RC/2272). SCOPE Ireland was funded by the Health Research Board, Ireland (CSA 2007/2). The BASELINE cohort was supported by the National Children’s Research Centre, Dublin, Ireland. Analysis of the data was supported by a Health Research Board Interdisciplinary Capacity Enhancement award (ICE/2012/12). Acknowledgements We thank the pregnant women who participated in the SCOPE study and the mothers who allowed their newborn infants participate in the BASELINE study. Conflicts of interest: None. References 1 Adair LS, Dahly DL. Developmental determinants of blood pressure in adults. Annu Rev Nutr  2005; 25: 407– 34. Google Scholar CrossRef Search ADS PubMed  2 Gillman MW. Developmental origins of health and disease. N Engl J Med  2005; 353: 1848– 50. Google Scholar CrossRef Search ADS PubMed  3 Godfrey KM, Lillycrop KA, Burdge GC, Gluckman PD, Hanson MA. Epigenetic mechanisms and the mismatch concept of the developmental origins of health and disease. Pediatr Res  2007; 61: 5 R– 10 R. Google Scholar CrossRef Search ADS   4 Barker DJP. Developmental origins of adult health and disease. J Epidemiol Community Health  2004; 58: 114– 5. Google Scholar CrossRef Search ADS PubMed  5 Hanson MA, Gluckman PD, Ma RCW, Matzen P, Biesma RG. Early life opportunities for prevention of diabetes in low and middle income countries. BMC Public Health  2012; 12: 1025. Google Scholar CrossRef Search ADS PubMed  6 Horton R. Maternal and child undernutrition: an urgent opportunity. Lancet  2008; 371: 179. Google Scholar CrossRef Search ADS PubMed  7 Wells JCK. Historical cohort studies and the early origins of disease hypothesis: making sense of the evidence. Proc Nutr Soc  2009; 68: 179– 88. Google Scholar CrossRef Search ADS PubMed  8 Vrijheid M, Casas M, Bergström A, et al.   European birth cohorts for environmental health research. Environ Health Perspect  2012; 120: 29– 37. Google Scholar CrossRef Search ADS PubMed  9 Sewell MF, Huston-Presley L, Super DM, Catalano P. Increased neonatal fat mass, not lean body mass, is associated with maternal obesity. Am J Obstet Gynecol  2006; 195: 1100– 3. Google Scholar CrossRef Search ADS PubMed  10 Catalano PM, Tyzbir ED, Allen SR, McBean JH, McAuliffe TL. Evaluation of fetal growth by estimation of neonatal body composition. Obstet Gynecol  1992; 79: 46– 50. Google Scholar PubMed  11 Wells JCK. Body composition in infants: evidence for developmental programming and techniques for measurement. Rev Endocr Metab Disord  2012; 13: 93– 101. Google Scholar CrossRef Search ADS PubMed  12 Hull HR, Dinger MK, Knehans AW, Thompson DM, Fields DA. Impact of maternal body mass index on neonate birthweight and body composition. Am J Obstet Gynecol  2008; 198: 416.e1– 6. Google Scholar CrossRef Search ADS   13 IAEA. Body Composition Assessment from Birth to Two Years of Age . Vienna: International Atomic Energy Agency, 2013. 14 Crozier SR, Inskip HM, Godfrey KM, et al.   Weight gain in pregnancy and childhood body composition: findings from the Southampton Women’s Survey. Am J Clin Nutr  2010; 91: 1745– 51. Google Scholar CrossRef Search ADS PubMed  15 Holzhauer S, Hokken Koelega ACS, Ridder M de, et al.   Effect of birthweight and postnatal weight gain on body composition in early infancy: The Generation R Study. Early Hum Dev  2009; 85: 285– 90. Google Scholar CrossRef Search ADS PubMed  16 Hawkes CP, Hourihane JO, Kenny LC, et al.   Gender- and gestational age-specific body fat percentage at birth. Pediatrics  2011; 128: e645– 51. Google Scholar PubMed  17 Andersen GS, Girma T, Wells JCK, et al.   Body composition from birth to 6 months of age in Ethiopian infants: reference data obtained by air-displacement plethysmography. Am J Clin Nutr  2013; 98: 885– 94. Google Scholar CrossRef Search ADS PubMed  18 Lee W, Riggs T, Koo WWK, et al.   Relationship of newborn adiposity to fetal growth outcome based on birthweight or the modified neonatal growth assessment score. J Matern Fetal Neonatal Med  2012; 25: 1933– 40. Google Scholar CrossRef Search ADS PubMed  19 Koenker R, Hallock KF. Quantile Regression. J Econ Perspect  2001; 15: 143– 56. Google Scholar CrossRef Search ADS   20 O’Donovan SM, Murray DM, Hourihane JO, Kenny LC, Irvine AD, Kiely M. Cohort Profile: The Cork BASELINE Birth Cohort Study: Babies after SCOPE: evaluating the longitudinal impact on neurological and nutritional endpoints. Int J Epidemiol  2015; 44: 764– 75. Google Scholar CrossRef Search ADS PubMed  21 North RA, McCowan LME, Dekker GA, et al.   Clinical risk prediction for pre-eclampsia in nulliparous women: development of model in international prospective cohort. BMJ  2011; 342: d1875. Google Scholar CrossRef Search ADS PubMed  22 Fomon SJ, Haschke F, Ziegler EE, Nelson SE. Body composition of reference children from birth to age 10 years. Am J Clin Nutr  1982; 35: 1169– 75. Google Scholar CrossRef Search ADS PubMed  23 Health Services Executive Ireland. Nutrition for Pregnancy . Clinical Practice Guideline No. 27 V 1.1. Limerick, Ireland: HSEI, 2016. 24 Rasmussen KM, Yaktine AL (eds); Institute of Medicine (US) and National Research Council (US) Committee to Reexamine IOM Pregnancy Weight Guidelines. Weight Gain During Pregnancy: Reexamining the Guidelines . Washington, DC: National Academies Press (US), 2009. 25 Chung JGY, Taylor RS, Thompson JMD, et al.   Gestational weight gain and adverse pregnancy outcomes in a nulliparous cohort. Eur J Obstet Gynecol Reprod Biol  2013; 167: 149– 53. Google Scholar CrossRef Search ADS PubMed  26 Clark BK, Sugiyama T, Healy GN, et al.   Validity and reliability of measures of television viewing time and other non-occupational sedentary behaviour of adults: a review. Obes Rev  2009; 10: 7– 16. Google Scholar CrossRef Search ADS PubMed  27 Sugiyama T, Healy GN, Dunstan DW, Salmon J, Owen N. Is television viewing time a marker of a broader pattern of sedentary behavior? Ann Behav Med  2008; 35: 245– 50. Google Scholar CrossRef Search ADS PubMed  28 Galbraith C, Jenkin G, Davis P, Coope P. New Zealand Socio-economic Index 1996 Users’ Guide . Wellington: Statistics New Zealand, 1996. 29 Cox JL, Holden JM, Sagovsky R. Detection of postnatal depression: development of the 10-item Edinburgh Postnatal Depression scale. Br J Psychiatry  1987; 150: 782– 86. Google Scholar CrossRef Search ADS PubMed  30 Adouard F, Glangeaud-Freudenthal NMC, Golse B. Validation of the Edinburgh postnatal depression scale (EPDS) in a sample of women with high-risk pregnancies in France. Arch Womens Ment Health  2005; 8: 89– 95. Google Scholar CrossRef Search ADS PubMed  31 Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav  1983; 24: 385– 96. Google Scholar CrossRef Search ADS PubMed  32 Tukey JW. Exploratory Data Analysis . Reading, MA: Addison-Wesley, 1977. 33 Nohr EA, Vaeth M, Baker JL, et al.   Combined associations of prepregnancy body mass index and gestational weight gain with the outcome of pregnancy. Am J Clin Nutr  2008; 87: 1750– 59. Google Scholar CrossRef Search ADS PubMed  34 Koenker RW, D’Orey V. Algorithm AS 229: computing regression quantiles. J R Stat Soc  1987; 36: 383– 93. 35 Powell J. Estimation of monotonic regression models under quantile restrictions. In: Powell J, Tauchen G (eds.). Nonparametric and Semiparametric Methods in Econometrics . Cambridge, UK: Cambridge University Press, 1989. 36 van Buuren S. Flexible Imputation of Missing Data . Boca Raton FL: Chapman & Hall/CRC, 2012. Google Scholar CrossRef Search ADS   37 Von Hippel PT. How to impute interactions, squares, and other transformed variables. Sociol Methodol  2009; 39: 265– 91. Google Scholar CrossRef Search ADS   38 Rubin DB. Multiple Imputation for Nonresponse in Surveys . New York, NY: Wiley, 1987. Google Scholar CrossRef Search ADS   39 R Development Core Team. R: A Language and Environment for Statistical Computing . Vienna: R Foundation for Statistical Computing, 2004. 40 Koenker R. quantreg: Quantile Regression . R package version 5.11. Vienna: R Foundation for Statistical Computing, 2014. 41 Harrell FE. Hmisc: Harrell Miscellaneous . R package version 3.14–6. Vienna: R Foundation for Statistical Computing, 2014. 42 Wickham H. ggplot2: Elegant Graphics for Data Analysis . New York, NY: Springer, 2009. 43 Catalano PM. Management of obesity in pregnancy. Obstet Gynecol  2007; 109: 419– 33. Google Scholar CrossRef Search ADS PubMed  44 Catalano PM, Hauguel-De Mouzon S. Is it time to revisit the Pedersen hypothesis in the face of the obesity epidemic? Am J Obstet Gynecol  2011; 204: 479– 87. Google Scholar CrossRef Search ADS PubMed  45 Pedersen J. Diabetes and pregnancy: blood sugar of newborn infants. Ugeskr Laeger  1952; 114: 685. Google Scholar PubMed  46 Mitanchez D, Burguet A, Simeoni U. Infants born to mothers with gestational diabetes mellitus: mild neonatal effects, a long-term threat to global health. J Pediatr  2014; 164: 445– 50. Google Scholar CrossRef Search ADS PubMed  47 Hull HR, Thornton JC, Ji Y, et al.   Higher infant body fat with excessive gestational weight gain in overweight women. Am J Obstet Gynecol  2011; 205: 211– 12. Google Scholar CrossRef Search ADS PubMed  48 Carlsen E, Renault K, Nørgaard K, et al.   Newborn regional body composition is influenced by maternal obesity, gestational weight gain and the birthweight standard score. Acta Paediatr  2014; 103: 939– 45. Google Scholar CrossRef Search ADS PubMed  49 Au CP, Raynes-Greenow CH, Turner RM, Carberry AE, Jeffery H. Fetal and maternal factors associated with neonatal adiposity as measured by air displacement plethysmography: a large cross-sectional study. Early Hum Dev  2013; 89: 839– 43. Google Scholar CrossRef Search ADS PubMed  50 Starling AP, Brinton JT, Glueck DH, et al.   Associations of maternal BMI and gestational weight gain with neonatal adiposity in the Healthy Start study. Am J Clin Nutr  2015; 101: 302– 09. Google Scholar CrossRef Search ADS PubMed  51 Friis CM, Qvigstad E, Paasche Roland MC, et al.   Newborn body fat: associations with maternal metabolic state and placental size. PLoS One  2013; 8: e57467. Google Scholar CrossRef Search ADS PubMed  52 Crozier SR, Inskip HM, Godfrey KM, et al.   Weight gain in pregnancy and childhood body composition: findings from the Southampton Women’s Survey. Am J Clin Nutr  2010; 91: 1745– 51. Google Scholar CrossRef Search ADS PubMed  53 McCarthy FP, Khashan AS, Murray D, et al.   Parental physical and lifestyle factors and their association with newborn body composition. BJOG  2016; 123: 1824– 29. Google Scholar CrossRef Search ADS PubMed  54 Sparks JW. Human intrauterine growth and nutrient accretion. Semin Perinatol  1984; 8: 74– 93. Google Scholar PubMed  55 Catalano PM, Thomas AJ, Huston-Presley L, Amini SB. Increased fetal adiposity: a very sensitive marker of abnormal in utero development. Am J Obstet Gynecol  2003; 189: 1698– 704. Google Scholar CrossRef Search ADS PubMed  56 Lindsay CA, Thomas AJ, Catalano PM. The effect of smoking tobacco on neonatal body composition. Am J Obstet Gynecol  1997; 177: 1124– 28. Google Scholar CrossRef Search ADS PubMed  57 Spady DW, Atrens MA, Szymanski WA. Effects of mother’s smoking on their infants’ body composition as determined by total body potassium. Pediatr Res  1986; 20: 716– 19. Google Scholar CrossRef Search ADS PubMed  58 Harrod CS, Reynolds RM, Chasan-Taber L, et al.   Quantity and timing of maternal prenatal smoking on neonatal body composition: the Healthy Start Study. J Pediatr  2014; 165: 707– 12. Google Scholar CrossRef Search ADS PubMed  59 Harrod CS, Chasan-Taber L, Reynolds RM, et al.   Physical activity in pregnancy and neonatal body composition: the Healthy Start study. Obstet Gynecol  2014; 124: 257– 64. Google Scholar CrossRef Search ADS PubMed  60 Hawkes CP, Hourihane JO, Kenny LC, et al.   Gender- and gestational age-specific body fat percentage at birth. Pediatrics  2011; 128: e645– 51. Google Scholar PubMed  61 Ellis KJ, Yao M, Shypailo RJ, et al.   Body composition assessment in infancy: air-displacement plethysmography compared with a reference 4-compartment model. Am J Clin Nutr  2007; 85: 90– 95. Google Scholar CrossRef Search ADS PubMed  62 Beyerlein A. Quantile regression-opportunities and challenges from a user’s perspective. Am J Epidemiol  2014; 180: 330– 31. Google Scholar CrossRef Search ADS PubMed  63 Wei Y, Terry MB. Re: “Quantile Regression - Opportunities and Challenges From a User’s Perspective.” Am J Epidemiol  2014; 181: 152– 53. Google Scholar CrossRef Search ADS PubMed  64 Wells JC. A critique of the expression of paediatric body composition data. Arch Dis Child  2001; 85: 67– 72. Google Scholar CrossRef Search ADS PubMed  65 Wells JCK. Toward body composition reference data for infants, children, and adolescents. Adv Nutr  2014; 5: 320 S– 29 S. Google Scholar CrossRef Search ADS   66 Wells JCK, Cole TJ. the ALSPAC study team. Adjustment of fat-free mass and fat mass for height in children aged 8 y. Int J Obes Relat Metab Disord  2002; 26: 947– 52. Google Scholar CrossRef Search ADS PubMed  67 Kronmal RA. Spurious correlation and the fallacy of the ratio standard revisited. J R Stat Soc A  1993; 156: 379– 92. Google Scholar CrossRef Search ADS   68 Wells JCK. Historical cohort studies and the early origins of disease hypothesis: making sense of the evidence. Proc Nutr Soc  2009; 68: 179– 88. Google Scholar CrossRef Search ADS PubMed  69 Wells JCK. Body composition in infants: evidence for developmental programming and techniques for measurement. Rev Endocr Metab Disord  2012; 13: 93– 101. Google Scholar CrossRef Search ADS PubMed  70 Wilcox AJ, Weinberg CR, Basso O. On the pitfalls of adjusting for gestational age at birth. Am J Epidemiol  2011; 174: 1062– 68. Google Scholar CrossRef Search ADS PubMed  71 Little RJA, Rubin DB. Statistical Analysis with Missing Data . Hoboken, NJ: Wiley, 2002. Google Scholar CrossRef Search ADS   72 Moons KGM, Donders RART, Stijnen T, Harrell FE. Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol  2006; 59: 1092– 101. Google Scholar CrossRef Search ADS PubMed  73 von Hippel PT. 4. Regression with missing Ys: an improved strategy for analyzing multiply imputed data. Sociol Methodol  2007; 37: 83– 117. Google Scholar CrossRef Search ADS   © The Author 2017; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association

Journal

International Journal of EpidemiologyOxford University Press

Published: Feb 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial