Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Body Mass Index and Risk of Suicide Among Men

Body Mass Index and Risk of Suicide Among Men Abstract Background Body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) has been linked to depression and the risk of suicide attempts and deaths in conflicting directions. Methods In a prospective cohort study of 46 755 men free of cancer enrolled in the Health Professionals Follow-up Study, participants reported their height, weight, diet, and physical activity on repeated occasions beginning in 1986 and were followed up until death or until February 2002. A subsample of 1829 men reported their mental health–related quality of life with the Mental Component Summary Scale of the 36-Item Short-Form Health Survey in 1998. Results A total of 131 men died from suicide during follow-up. A higher BMI was related to a graded decline in the suicide mortality rate, from 52 per 100 000 person-years among men with a BMI of less than 21 to 13 per 100 000 among men with a BMI of 30 or higher; the adjusted hazard ratio per 1-U BMI increment was 0.89 (95% confidence interval, 0.84-0.95; P<.001). The relationship was consistent when baseline or updated measures of BMI were used and with adjustment for medical illness, dietary factors, antidepressant use, physical activity, or social support. Height and physical activity were not strongly associated with risk. Analyses of mental health–related quality of life showed a similar positive relationship with BMI. Conclusions Among men, risk of death from suicide is strongly inversely related to BMI, but not to height or to physical activity. Although obesity cannot be recommended on the basis of its detrimental effects, further research into the mechanisms of lower risk among overweight and obese men may provide insights into effective methods of suicide prevention. Obesity and depression, extremely common causes of morbidity,1,2 have an intriguing and complex interrelationship.3,4 Although depression has been associated with subsequent obesity in some studies,5 obesity has also been associated with depressive symptoms6 that may improve with weight loss.7,8 Antidepressant-related weight changes9 and activation of the hypothalamic-adrenal axis in both disorders4 complicate this interrelationship even further. Another complicating factor in the association of obesity with depression is the “jolly fat” hypothesis, first proposed 3 decades ago, which suggests that overweight is inversely related to anxiety and depressive symptoms.10-12 Although the mechanism of this relationship is unclear, some have speculated that a high intake of carbohydrates may ameliorate or prevent depressive symptoms but also lead to weight gain.13 Despite interest in the relationship of weight and depression, the relationship of weight and body habitus with the risk of suicide has received far less attention. Suicide remains the 11th leading cause of death in the United States,14 but identification of remediable risk factors for suicide has been difficult. Based on the links between obesity and depression, it might be hypothesized that overweight and obese individuals would have a higher risk of suicide. A small number of studies with limited information on anthropomorphic measures and potentially confounding factors have suggested exactly the opposite, at least among men,15,16 although an inverse association has not been detected in all studies.17,18 To determine the associations of body mass index (BMI) and other anthropomorphic measures with the risk of death from suicide (hereafter referred to as suicide death), we studied more than 45 000 participants of the Health Professionals Follow-up Study (HPFS). The HPFS is a prospective cohort study of male health care professionals in the United States for whom 16-year follow-up on mortality,19 including death related to suicide, and repeated, validated information on BMI, diet, and physical activity is available. Methods The HPFS is a prospective investigation of 51 529 US male dentists, pharmacists, veterinarians, optometrists, osteopathic physicians, and podiatrists aged 40 to 75 years who returned a mailed questionnaire regarding diet and medical history in 1986. Follow-up questionnaires were sent biennially to update information on exposures and newly diagnosed illnesses. For this analysis, we excluded men with cancer (other than nonmelanoma skin cancer) at baseline (n = 2002) and men whose initial questionnaires were missing information on BMI (n = 1094) or had inadequate dietary information or other technical deficiencies (n = 1678), leaving 46 755 men available for analysis. Anthropomorphic measures Participants reported body weight and height at baseline in 1986 and body weight alone in every biennial questionnaire afterward. At baseline, participants also reported their estimated weight 5 years previously (ie, in 1981) and their estimated weight at 21 years of age. As a measure of weight cycling, participants also reported the difference in their maximal and minimal weight in the preceding 2 years beginning in 1990. Information on waist and hip circumference was collected in a brief follow-up questionnaire sent out in 198720 that included paper tape measures, detailed instructions, and illustrations to assist the men in measuring standing waist and hip circumferences. Participants were instructed to measure their waist circumference at the umbilicus and their hip circumference as the largest diameter between the waist and thighs. We calculated the BMI (measured as weight in kilograms divided by height in meters squared) and waist-hip ratio for each participant. Self-reported anthropomorphic measures have been compared with standardized measures conducted by a technician in a subset of this cohort.21 Self-reported and measured weight were strongly correlated (r = 0.97); the mean difference was an underreporting by 1.06 kg. The corresponding correlation coefficients between self-reported and measured waist and hip circumferences were 0.95 and 0.88, respectively. We categorized BMI into 6 categories spanning approximately 2 U each as in previous analyses22; in sensitivity analyses, we also examined it in quintiles, in World Health Organization (WHO) categories, and as a continuous variable after winsorization at the upper 95th percentile to minimize the influence of outliers.23 We examined the height and waist-hip ratio in quintiles. Physical activity Physical activity was derived from the biennial questionnaires, which included a series of questions on the average total time per week spent on specific activities during the previous year. Total weekly energy expenditure was expressed as metabolic equivalent–hours.24 The validity and reproducibility of physical activity measurements were determined with a past-week recall, four 1-week activity diaries, and a step test among a subset of participants, as reported elsewhere.25 The correlations between diary-based and questionnaire-based activity scores were 0.28 for nonvigorous activity and 0.58 for vigorous activity. We categorized physical activity into approximate quintiles as in previous analyses.26 Other covariates We assessed dietary factors using a 131-item food frequency questionnaire, which has been validated with dietary records.27 In 1988, participants responded to a series of questions regarding social support, including the number of close friends and close relatives they had, the number of friends and relatives they saw at least monthly, how often they attended religious gatherings, and how many hours per week they spent in community or social groups.28 Participants reported use of antidepressant medication every 2 years, beginning with the 1990 questionnaire; doses were not reported. As in previous analyses,29 region of residence was categorized into the following 4 geographic regions, according to US Census definitions: West, Midwest, South, and Northeast. Participants reported prevalent and incident illnesses at every questionnaire; cases of cancer, coronary heart disease, and diabetes are subsequently adjudicated by physicians blinded to questionnaire information.20,22,30 Participants also reported their employment and marital status every 2 years; response options were full-time, part-time, retired, or disabled for employment, and married, divorced, widowed, or never married for marital status. Assessment of suicide deaths We confirmed deaths when reported by families, postal officials, or the National Death Index, with a combined follow-up rate in a companion study exceeding 98%.31 Physicians reviewed death certificates and hospital or pathology reports to classify individual causes of death and were unaware of participants' reported questionnaire results. Deaths caused by self-inflicted or external injury were classified according to the underlying causes listed on the death certificate. For this study, we included deaths classified with codes E950 through E959 of the International Classification of Diseases, Eighth Revision. We also separately examined suicide deaths associated with firearm use (code E955) and other causes. Quality of life In 1998, a supplemental questionnaire was administered to men with prostate cancer and control participants who were free of prostate cancer32,34; the questionnaire included the Medical Outcomes Study 36-Item Short-Form Health Survey.35 For each case, 2 healthy controls matched on age were selected; the response rate among the controls was 93%. Of the 2217 controls who returned questionnaires, 2103 had baseline information on BMI and nutrient intake, and 1829 had complete information on the Mental Component Summary (MCS) and Physical Component Summary scales, summary measures of mental and physical health–related quality of life, respectively. The scales have population means of 50 and SDs of 10, with higher scores indicating better quality of life in that domain. Statistical analysis We calculated person-years for each participant from the date of return of the 1986 questionnaire to the date of death or to February 2002. Analyses incorporating the waist-hip ratio or social support began at the return of the 1988 questionnaire, and the corresponding analyses of antidepressant use began at the return of the 1990 questionnaire. We calculated incidence rates by dividing the number of suicides in each category by person-years of follow-up in that category. In multivariate analyses, we simultaneously controlled for age, smoking (never, former, and 1-14, 15-24, and ≥25 cigarettes per day),36 alcohol use (in 4 categories), geographic region, physical activity (in 5 categories), and marital status (in 4 categories). In sensitivity analyses, we also adjusted individually for intakes of energy, glycemic load,37 caffeine,38 and ω-3 fatty acids39 (all in quintiles); profession (in 6 categories); measures of social support (in 5 categories each); employment status; or incident major illnesses (cancer, diabetes, and coronary heart disease). Nutrient intakes were updated every 4 years; other covariates (other than region and profession) were updated every 2 years. For multivariate analysis, we used Cox regression with time-dependent covariates; we noted no violations of the proportional hazards assumption. Our analyses used updated measures of BMI, in which we prospectively assessed the relative risk of suicide death in 2-year increments, based on the weight reported in the preceding questionnaire. Thus, we used the 1986 questionnaire to determine the risk of suicide death during the 1986-1988 period, the 1988 questionnaire for the 1988-1990 period, etc. In all analyses, we excluded men with prevalent malignancy at baseline and stopped updating BMI at the diagnosis of incident malignancy; in sensitivity analyses, we further excluded follow-up after diagnosis of malignancy during follow-up. We also examined baseline BMI alone and with a 2-year lag to minimize the inclusion of individuals with preexisting illnesses. For quality of life, we used linear regression to examine the relationship of BMI in 1986 or 1998 with MCS and Physical Component Summary scores, adjusting for the same covariates as in Cox models; however, of these covariates, only age and BMI were independently associated with MCS scores. Results Baseline characteristics of the HPFS cohort according to BMI are shown in Table 1; similar characteristics according to physical activity have been reported previously.26 A total of 131 men died of suicide during follow-up, ie, a suicide mortality rate of 19 per 100 000 person-years. The median age at suicide death was 64 years (interquartile range, 56-74 years). Table 2 shows incidence rates and hazard ratios (HRs) for suicide according to BMI. There was a stepwise decrease in the suicide mortality rate with increasing BMI, with a roughly 4-fold gradient in risk even after adjustment. Additional adjustment for incident disease or social support did not influence the results; incident cancer was strongly related to suicide risk (HR, 2.23; 95% confidence interval [CI], 1.27-3.93), but diabetes was not (HR, 1.31; 95% CI, 0.68-2.53), even without adjustment for BMI (HR, 1.23; 95% CI, 0.64-2.36). Results also tended to be similar for suicide deaths related to firearms and other causes. Additional adjustment for energy intake, caffeine, or ω-3 fatty acid intake also did not influence our findings (data not shown). Analyses restricted to nonsmokers, those using a 2-year lag or baseline BMI, and those adjusted for antidepressant use (a potential intermediate) all attenuated the HRs to a moderate degree, although at least 2-fold gradients in risk remained in all cases. Antidepressant use itself was also strongly associated with risk (HR, 5.94; 95% CI, 3.03-11.64). Table 2 further shows analyses using alternate cut points for BMI. Using the WHO definition of normal weight, there was an inverse association of BMI and risk through BMI values below 35.0; risk did not appear to be lower at the highest levels of BMI, but only 5 cases of suicide occurred in this subgroup. Table 3 shows the relationships of waist-hip ratio, height, and physical activity with risk of suicide. There was an inverse trend between waist-hip ratio and suicide risk, although it was less strong than that seen for BMI and not statistically significant. There was no evidence of any association between height or physical activity and risk. Table 4 shows the relationship of a 1-U increase in BMI with risk of suicide, overall and in sensitivity analyses. Risk was 11% lower for each 1-U increase in the full cohort. Neither censoring at the time of incident malignancy nor incorporating a 4-year lag between the assessment of exposure and onset of follow-up altered this finding. Even in analyses excluding men in the lowest BMI category to minimize confounding related to comorbid illness, there was a 10% lower risk for a 1-U increase in BMI. The inverse association of BMI with suicide was also not altered by additional adjustment for glycemic load (HR, 0.89; 95% CI, 0.84-0.95), and it did not differ in men whose glycemic load was above or below the median level; glycemic load per se was not associated with risk. The association was also identical when adjusted for work status, although disability itself was associated with a higher risk (HR, 3.20; 95% CI, 1.20-8.58). Stratified analyses suggested generally similar relationships in all subgroups, with no statistically significant interactions; although the association appeared to be stronger in younger men, even when BMI and age were modeled as continuous variables, the interaction was not significant (P=.14). When we analyzed change in BMI during follow-up controlling for previous BMI, a 1-U increase in BMI was associated with a nonsignificant 11% lower subsequent risk. When we further excluded decreases in BMI of 0.5 U or more to minimize the contribution of weight loss among participants before suicide, a 1-U increase in BMI was associated with a nonsignificant 17% lower subsequent risk (HR, 0.83; 95% CI, 0.56-1.21; P = .32). We also evaluated the association of self-reported BMI at 21 years of age with risk of suicide mortality during follow-up. The HR associated with a 1-U increase in BMI at 21 years of age was 0.95 (95% CI, 0.89-1.01); the corresponding HR for a 1-U gain in BMI from 21 years of age to baseline was 0.93 (95% CI, 0.87-1.00). Large amounts of weight change within a 2-year period tended to be associated with a higher risk, although estimates were not statistically significant (HR for the maximal weight difference within the last 2 years of ≥6.8 kg vs ≤1.8 kg, 1.61; 95% CI, 0.79-3.31) and did not alter the association of BMI with risk (HR per 1-U increment in BMI, 0.90; 95% CI, 0.84-0.97). Table 5 shows the relationship of BMI and mental health–related quality of life using the baseline BMI in 1986 or the BMI at the time of the questionnaire in 1998. The Pearson correlation coefficients of 1986 and 1998 BMI with MCS scores were 0.11 and 0.09, respectively (P<.001 for both). Similar to the results seen for suicide risk, MCS scores tended to be higher with increasing BMI, even when measured at baseline 12 years before the administration of the 36-Item Short-Form Health Survey. In contrast, 1986 and 1998 BMIs were inversely correlated with Physical Component Summary scores, with Pearson correlation coefficients of −0.15 and −0.11, respectively (P<.001 for both). Comment In this prospective cohort study of male health care professionals, higher BMI was associated with a lower risk of suicide mortality. There was a stepwise gradient in risk, the results were largely unchanged in sensitivity analyses, and there was a similar relationship of BMI with mental health–related quality of life. A small number of previous studies support an inverse relationship of BMI with risk of suicide death, although data on suicide attempts generally suggest the opposite. In a case-control analysis of the General Practice Research Database, hypertensive adults with a BMI of 25 or higher had an adjusted odds ratio for suicide of 0.33 (95% CI, 0.11-0.98) compared with leaner hypertensive adults.15 Using the military records of nearly 1.3 million Swedish male teens, Magnusson and colleagues16 found an inverse association of teenage BMI with the risk of suicide, with an HR of 0.85 (95% CI, 0.79-0.91) for each 5-U increment in BMI, although no information on adult anthropometric measures was available. However, suicide attempts tended to be positively associated with BMI in a representative German sample40 and among teenagers41 and female physicians42 in the United States. Dong and colleagues43 also found positive associations of marked obesity with suicide attempts in 2 parallel studies, but more typical levels of obesity were not assessed. The increased prevalence of overweight and obesity in the United States in the 1990s was accompanied by a 6% decrease in suicide mortality rates but little apparent change in the rate of suicide attempts.44 The mechanisms by which adult BMI could be inversely linked to the risk of suicide death are as yet uncertain. One possibility is the role of insulin in dictating mood. The drop in maternal insulin levels after birth has been hypothesized to precipitate postpartum depression,45 perhaps related to the ability of insulin to stimulate hypothalamic serotonin release.46 In a study of 30 patients with type 2 diabetes, 4 weeks of insulin administration led to marked improvements in depressive symptoms and subjective well-being, even with minimally altered glycemic control.47 Given that BMI has a continuous and positive relationship with insulin levels at all levels of weight,48 even among children,49 and that weight gain over time has a similar relationship,50 it is plausible that BMI could be linked to lower risk of suicide through circulating levels of insulin. At the same time, we generally found little evidence to support the hypothesis that specific dietary factors mediate the relationship of obesity and suicide. Although carbohydrate intake in particular has been postulated to improve mood and increase weight, adjustment for glycemic load had no effect on our estimates of risk. Glycemic load is a measure of the acute glycemic response51 and is not necessarily related to fasting insulin levels.52 Taken with our findings on BMI, these results suggest that physiological responses to glycemic load may be less relevant for suicide than is adiposity per se. Our findings in this population of men may not readily generalize to all women. For example, in an analysis of the 1992 National Longitudinal Alcohol Epidemiologic Survey, Carpenter and colleagues53 found that a 10-U increment in BMI was associated with increased odds of suicide attempts among women (odds ratio, 1.22; 95% CI, 0.91-1.63), but significantly decreased odds among men (odds ratio, 0.45; 95% CI, 0.40-0.51). Several cross-sectional studies have suggested that men and women differ in their relationship between BMI and depressive symptoms.10,12,54 It is possible that the stigma attached to obesity may be particularly acute among women,55 although this has not been borne out in all studies.56 Alternately, the association between endogenous estrogen levels and depression may complicate the picture for women. Estrogen alters serotonergic neurotransmission and receptor distribution57,58 and has been suggested to ameliorate depression in some clinical trials.59 In support of this hypothesis, BMI may be inversely associated with depressive symptoms among postmenopausal but not premenopausal women.11 The relationship of BMI and suicide tended to be stronger among younger than older men, although a formal test of interaction did not yield significant results. This may reflect the role of chance, as in any subgroup analysis. We also could not readily evaluate suicide among men in their third and fourth decades of life, in whom suicide accounts for a disproportionately large proportion of deaths, albeit at a lower absolute rate than among older men. However, BMI tends to be a better measure of adiposity in the young, whereas it is disproportionately influenced by underlying illness and changes in muscle mass in older adults.60 If the weaker relation in older men is confirmed, it would tend to support further the importance of adiposity per se in lowering the risk of suicide. The potential interrelationships of BMI, depressive symptoms and major depressive disorders, antidepressant medication, and risk of suicide are complex and likely bidirectional. On the one hand, BMI may have direct positive effects on mood that mediate the observed association of BMI with suicide, such as through elevated insulin levels. On the other hand, depression and the medications used to treat it may influence eating and exercise habits and subsequently BMI. This bidirectionality is also likely to be true for other emotional and relational factors for which we did not have information. We had information on antidepressant use, a potential contributor to suicide even in adults,61 and social support, neither of which appeared to confound the association of BMI with the risk of suicide death to any substantial degree. However, we did not have systematic information on depression or depressive symptoms. Future studies that incorporate repeated measures of both depression and BMI in large samples of participants and methodological approaches that account for the possibility that depression acts as both a mediator and a confounder for the effect of BMI on suicide risk62,63 will ultimately be required to clarify these relationships definitively. The HPFS is also subject to other limitations. We did not have information on suicide attempts, and hence we could not ascertain whether BMI is related to the risk of suicide attempts or their case-fatality rate in this population. This latter issue will require additional study, given the somewhat contradictory results of previous studies relating BMI to suicide attempts and suicide death. Participants in the HPFS are not necessarily representative of the total US population, which limits the generalizability of our results but provides a sample of uniform sex, educational attainment, and occupational status. Moreover, the rate of suicide death among these middle-aged and older men is similar to that found in other large cohort studies.64 As in other cohort studies, our definition of suicide death was based on death certificates, an approach that appears to be valid.65 However, treating practitioners may have misclassified some cases of suicide death, particularly in cases where other causes of death may have been plausible (eg, among men with known chronic disease), although cases of suicide death related to firearm use may be less prone to underreporting and were related inversely with BMI. Although we relied on self-reported measures of height and weight in this study, we have previously validated them in this population and all assessments were made prospectively. A related limitation is the relatively truncated range of BMI noted among these men, and we cannot evaluate the possibility that greater degrees of obesity (eg, BMI of ≥40) are associated with a higher risk of suicide, as suggested elsewhere.43 Our analyses were necessarily limited by the relative rarity of death related to suicide. Although our primary analyses yielded reasonably precise results, our stratified analyses—and particularly stratified analyses that examined categories of BMI—yielded considerably wider CIs. Larger studies with greater numbers of cases of suicide mortality will be needed to overcome this limitation. In conclusion, higher BMI was strongly associated with a lower risk of suicide mortality among men, with a clear gradient in risk between the highest and lowest BMI levels. Given the strength of this relationship, research to identify its mechanisms could lead to important insights into prevention of this important cause of death. Correspondence: Kenneth J. Mukamal, MD, MPH, MA, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 1309 Beacon St, Second Floor, Boston, MA 02446 (kmukamal@bidmc.harvard.edu). Accepted for Publication: November 13, 2006. Author Contributions: Dr Mukamal had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Mukamal and Rimm. Acquisition of data: Rimm. Analysis and interpretation of data: Mukamal, Kawachi, Miller, and Rimm. Drafting of the manuscript: Mukamal. Critical revision of the manuscript for important intellectual content: Kawachi, Miller, and Rimm. Statistical analysis: Mukamal and Miller. Obtained funding: Mukamal and Rimm. Financial Disclosure: None reported. Funding/Support: This study was supported by grants AA014900, HL35464, CA55075, AA11181, HL35464, and CA55075 from the National Institutes of Health. Role of the Sponsor: The funding sources for this work had no role in the analysis, presentation, or publication of results. References 1. Kessler RCChiu WTDemler OMerikangas KRWalters EE Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 2005;62617- 627PubMedGoogle Scholar 2. Flegal KM Epidemiologic aspects of overweight and obesity in the United States. Physiol Behav 2005;86599- 602PubMedGoogle Scholar 3. Evans DLCharney DSLewis L et al. Mood disorders in the medically ill: scientific review and recommendations. Biol Psychiatry 2005;58175- 189PubMedGoogle Scholar 4. Stunkard AJFaith MSAllison KC Depression and obesity. Biol Psychiatry 2003;54330- 337PubMedGoogle Scholar 5. Franko DLStriegel-Moore RHThompson DSchreiber GBDaniels SR Does adolescent depression predict obesity in black and white young adult women? Psychol Med 2005;351505- 1513PubMedGoogle Scholar 6. Roberts REDeleger SStrawbridge WJKaplan GA Prospective association between obesity and depression: evidence from the Alameda County Study. Int J Obes Relat Metab Disord 2003;27514- 521PubMedGoogle Scholar 7. Mamplekou EKomesidou VBissias CPapakonstantinou AMelissas J Psychological condition and quality of life in patients with morbid obesity before and after surgical weight loss. Obes Surg 2005;151177- 1184PubMedGoogle Scholar 8. Waters GSPories WJSwanson MSMeelheim HDFlickinger EGMay HJ Long-term studies of mental health after the Greenville gastric bypass operation for morbid obesity. Am J Surg 1991;161154- 158PubMedGoogle Scholar 9. Zimmermann UKraus THimmerich HSchuld APollmacher T Epidemiology, implications and mechanisms underlying drug-induced weight gain in psychiatric patients. J Psychiatr Res 2003;37193- 220PubMedGoogle Scholar 10. Crisp AHMcGuiness B Jolly fat: relation between obesity and psychoneurosis in general population. BMJ 1976;17- 9PubMedGoogle Scholar 11. Jasienska GZiomkiewicz AGorkiewicz MPajak A Body mass, depressive symptoms and menopausal status: an examination of the “jolly fat” hypothesis. Womens Health Issues 2005;15145- 151PubMedGoogle Scholar 12. Palinkas LAWingard DLBarrett-Connor E Depressive symptoms in overweight and obese older adults: a test of the “jolly fat” hypothesis. J Psychosom Res 1996;4059- 66PubMedGoogle Scholar 13. Wurtman RJWurtman JJ Carbohydrates and depression. Sci Am 1989;26068- 75PubMedGoogle Scholar 14. Hoyert DLKung HCSmith BL Deaths: preliminary data for 2003. Natl Vital Stat Rep 2005;531- 48PubMedGoogle Scholar 15. Gasse CDerby LEVasilakis CJick H Risk of suicide among users of calcium channel blockers: population based, nested case-control study. BMJ 2000;3201251PubMedGoogle Scholar 16. Magnusson PKRasmussen FLawlor DATynelius PGunnell D Association of body mass index with suicide mortality: a prospective cohort study of more than one million men. Am J Epidemiol 2006;1631- 8PubMedGoogle Scholar 17. Jiang GXRasmussen FWasserman D Short stature and poor psychological performance: risk factors for attempted suicide among Swedish male conscripts. Acta Psychiatr Scand 1999;100433- 440PubMedGoogle Scholar 18. Allebeck PBergh C Height, body mass index and mortality: do social factors explain the association? Public Health 1992;106375- 382PubMedGoogle Scholar 19. Kawachi IColditz GAAscherio A et al. A prospective study of social networks in relation to total mortality and cardiovascular disease in men in the USA. J Epidemiol Community Health 1996;50245- 251PubMedGoogle Scholar 20. Chan JMRimm EBColditz GAStampfer MJWillett WC Obesity, fat distribution, and weight gain as risk factors for clinical diabetes in men. Diabetes Care 1994;17961- 969PubMedGoogle Scholar 21. Rimm EBStampfer MJColditz GAChute CGLitin LBWillett WC Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990;1466- 473PubMedGoogle Scholar 22. Giovannucci ERimm EBLiu Y et al. Body mass index and risk of prostate cancer in US health professionals. J Natl Cancer Inst 2003;951240- 1244PubMedGoogle Scholar 23. Fernandez JREtzel CBeasley TMShete SAmos CIAllison DB Improving the power of sib pair quantitative trait loci detection by phenotype winsorization. Hum Hered 2002;5359- 67PubMedGoogle Scholar 24. Ainsworth BEHaskell WLLeon AS et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;2571- 80PubMedGoogle Scholar 25. Chasan-Taber SRimm EBStampfer MJ et al. Reproducibility and validity of a self-administered physical activity questionnaire for male health professionals. Epidemiology 1996;781- 86PubMedGoogle Scholar 26. Tanasescu MLeitzmann MFRimm EBWillett WCStampfer MJHu FB Exercise type and intensity in relation to coronary heart disease in men. JAMA 2002;2881994- 2000PubMedGoogle Scholar 27. Feskanich DRimm EBGiovannucci EL et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993;93790- 796PubMedGoogle Scholar 28. Eng PMRimm EBFitzmaurice GKawachi I Social ties and change in social ties in relation to subsequent total and cause-specific mortality and coronary heart disease incidence in men. Am J Epidemiol 2002;155700- 709PubMedGoogle Scholar 29. Michaud DSClinton SKRimm EBWillett WCGiovannucci E Risk of bladder cancer by geographic region in a US cohort of male health professionals. Epidemiology 2001;12719- 726PubMedGoogle Scholar 30. Rimm EBGiovannucci ELWillett WC et al. Prospective study of alcohol consumption and risk of coronary disease in men. Lancet 1991;338464- 468PubMedGoogle Scholar 31. Rich-Edwards JWCorsano KAStampfer MJ Test of the National Death Index and Equifax Nationwide Death Search. Am J Epidemiol 1994;1401016- 1019PubMedGoogle Scholar 32. Bacon CGGiovannucci ETesta MGlass TAKawachi I The association of treatment-related symptoms with quality-of-life outcomes for localized prostate carcinoma patients. Cancer 2002;94862- 871PubMedGoogle Scholar 33. Bacon CGGiovannucci ETesta MKawachi I The impact of cancer treatment on quality of life outcomes for patients with localized prostate cancer. J Urol 2001;1661804- 1810PubMedGoogle Scholar 34. Bacon CGKawachi I Quality-of-life differences among various populations of localized prostate cancer patients: 2001. Curr Urol Rep 2002;3239- 243PubMedGoogle Scholar 35. Ware JE JrSherbourne CD The MOS 36-Item Short-Form Health Survey (SF-36), I: conceptual framework and item selection. Med Care 1992;30473- 483PubMedGoogle Scholar 36. Miller MHemenway DRimm E Cigarettes and suicide: a prospective study of 50 000 men. Am J Public Health 2000;90768- 773PubMedGoogle Scholar 37. Salmeron JAscherio ARimm EB et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997;20545- 550PubMedGoogle Scholar 38. Kawachi IWillett WCColditz GAStampfer MJSpeizer FE A prospective study of coffee drinking and suicide in women. Arch Intern Med 1996;156521- 525PubMedGoogle Scholar 39. Freeman MP Omega-3 fatty acids in psychiatry: a review. Ann Clin Psychiatry 2000;12159- 165PubMedGoogle Scholar 40. Brunner JBronisch TPfister HJacobi FHofler MWittchen HU High cholesterol, triglycerides, and body-mass index in suicide attempters. Arch Suicide Res 2006;101- 9PubMedGoogle Scholar 41. Eaton DKLowry RBrener NDGaluska DACrosby AE Associations of body mass index and perceived weight with suicide ideation and suicide attempts among US high school students. Arch Pediatr Adolesc Med 2005;159513- 519PubMedGoogle Scholar 42. Frank EDingle AD Self-reported depression and suicide attempts among US women physicians. Am J Psychiatry 1999;1561887- 1894PubMedGoogle Scholar 43. Dong CLi WDLi DPrice RA Extreme obesity is associated with attempted suicides: results from a family study. Int J Obes (Lond) 2006;30388- 390PubMedGoogle Scholar 44. Kessler RCBerglund PBorges GNock MWang PS Trends in suicide ideation, plans, gestures, and attempts in the United States, 1990-1992 to 2001-2003. JAMA 2005;2932487- 2495PubMedGoogle Scholar 45. Chen THLan THYang CYJuang KD Postpartum mood disorders may be related to a decreased insulin level after delivery. Med Hypotheses 2006;66820- 823PubMedGoogle Scholar 46. Orosco MRouch CGerozissis K Activation of hypothalamic insulin by serotonin is the primary event of the insulin-serotonin interaction involved in the control of feeding. Brain Res 2000;87264- 70PubMedGoogle Scholar 47. Reza MTaylor CDTowse KWard JDHendra TJ Insulin improves well-being for selected elderly type 2 diabetic subjects. Diabetes Res Clin Pract 2002;55201- 207PubMedGoogle Scholar 48. Shiwaku KAnuurad EEnkhmaa B et al. Predictive values of anthropometric measurements for multiple metabolic disorders in Asian populations. Diabetes Res Clin Pract 2005;6952- 62PubMedGoogle Scholar 49. Janssen IKatzmarzyk PTSrinivasan SR et al. Combined influence of body mass index and waist circumference on coronary artery disease risk factors among children and adolescents. Pediatrics 2005;1151623- 1630PubMedGoogle Scholar 50. Norman JEBild DLewis CELiu KWest DS The impact of weight change on cardiovascular disease risk factors in young black and white adults: the CARDIA study. Int J Obes Relat Metab Disord 2003;27369- 376PubMedGoogle Scholar 51. Wolever TMJenkins DJJenkins ALJosse RG The glycemic index: methodology and clinical implications. Am J Clin Nutr 1991;54846- 854PubMedGoogle Scholar 52. Liese ADSchulz MFang F et al. Dietary glycemic index and glycemic load, carbohydrate and fiber intake, and measures of insulin sensitivity, secretion, and adiposity in the Insulin Resistance Atherosclerosis Study. Diabetes Care 2005;282832- 2838PubMedGoogle Scholar 53. Carpenter KMHasin DSAllison DBFaith MS Relationships between obesity and DSM-IV major depressive disorder, suicide ideation, and suicide attempts: results from a general population study. Am J Public Health 2000;90251- 257PubMedGoogle Scholar 54. Heo MPietrobelli AFontaine KRSirey JAFaith MS Depressive mood and obesity in US adults: comparison and moderation by sex, age, and race. Int J Obes (Lond) 2006;30513- 519PubMedGoogle Scholar 55. Miller CTRothblum EDBarbour LBrand PAFelicio D Social interactions of obese and nonobese women. J Pers 1990;58365- 380PubMedGoogle Scholar 56. Myers ARosen JC Obesity stigmatization and coping: relation to mental health symptoms, body image, and self-esteem. Int J Obes Relat Metab Disord 1999;23221- 230PubMedGoogle Scholar 57. Fink GSumner BEMcQueen JKWilson HRosie R Sex steroid control of mood, mental state and memory. Clin Exp Pharmacol Physiol 1998;25764- 775PubMedGoogle Scholar 58. Rybaczyk LABashaw MJPathak DRMoody SMGilders RMHolzschu DL An overlooked connection: serotonergic mediation of estrogen-related physiology and pathology. BMC Womens Health 2005;512PubMedGoogle Scholar 59. Soares CNAlmeida OPJoffe HCohen LS Efficacy of estradiol for the treatment of depressive disorders in perimenopausal women: a double-blind, randomized, placebo-controlled trial. Arch Gen Psychiatry 2001;58529- 534PubMedGoogle Scholar 60. Flint AJRimm EB Commentary: obesity and cardiovascular disease risk among the young and old: is BMI the wrong benchmark? Int J Epidemiol 2006;35187- 189PubMedGoogle Scholar 61. Gunnell DSaperia JAshby D Selective serotonin reuptake inhibitors (SSRIs) and suicide in adults: meta-analysis of drug company data from placebo controlled, randomised controlled trials submitted to the MHRA's safety review. BMJ 2005;330385[published correction appears in BMJ. 2006;333:30]PubMeddoi:10.1136/bmj.330.7488.385Google Scholar 62. Hernan MABrumback BARobins JM Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures. Stat Med 2002;211689- 1709PubMedGoogle Scholar 63. Hernan MACole SRMargolick JCohen MRobins JM Structural accelerated failure time models for survival analysis in studies with time-varying treatments. Pharmacoepidemiol Drug Saf 2005;14477- 491PubMedGoogle Scholar 64. Tanskanen ATuomilehto JViinamaki HVartiainen ELehtonen JPuska P Joint heavy use of alcohol, cigarettes and coffee and the risk of suicide. Addiction 2000;951699- 1704PubMedGoogle Scholar 65. Moyer LABoyle CAPollock DA Validity of death certificates for injury-related causes of death. Am J Epidemiol 1989;1301024- 1032PubMedGoogle Scholar http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Archives of Internal Medicine American Medical Association

Loading next page...
 
/lp/american-medical-association/body-mass-index-and-risk-of-suicide-among-men-f0J9DgXj8a

References (68)

Publisher
American Medical Association
Copyright
Copyright © 2007 American Medical Association. All Rights Reserved.
ISSN
0003-9926
DOI
10.1001/archinte.167.5.468
pmid
17353494
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) has been linked to depression and the risk of suicide attempts and deaths in conflicting directions. Methods In a prospective cohort study of 46 755 men free of cancer enrolled in the Health Professionals Follow-up Study, participants reported their height, weight, diet, and physical activity on repeated occasions beginning in 1986 and were followed up until death or until February 2002. A subsample of 1829 men reported their mental health–related quality of life with the Mental Component Summary Scale of the 36-Item Short-Form Health Survey in 1998. Results A total of 131 men died from suicide during follow-up. A higher BMI was related to a graded decline in the suicide mortality rate, from 52 per 100 000 person-years among men with a BMI of less than 21 to 13 per 100 000 among men with a BMI of 30 or higher; the adjusted hazard ratio per 1-U BMI increment was 0.89 (95% confidence interval, 0.84-0.95; P<.001). The relationship was consistent when baseline or updated measures of BMI were used and with adjustment for medical illness, dietary factors, antidepressant use, physical activity, or social support. Height and physical activity were not strongly associated with risk. Analyses of mental health–related quality of life showed a similar positive relationship with BMI. Conclusions Among men, risk of death from suicide is strongly inversely related to BMI, but not to height or to physical activity. Although obesity cannot be recommended on the basis of its detrimental effects, further research into the mechanisms of lower risk among overweight and obese men may provide insights into effective methods of suicide prevention. Obesity and depression, extremely common causes of morbidity,1,2 have an intriguing and complex interrelationship.3,4 Although depression has been associated with subsequent obesity in some studies,5 obesity has also been associated with depressive symptoms6 that may improve with weight loss.7,8 Antidepressant-related weight changes9 and activation of the hypothalamic-adrenal axis in both disorders4 complicate this interrelationship even further. Another complicating factor in the association of obesity with depression is the “jolly fat” hypothesis, first proposed 3 decades ago, which suggests that overweight is inversely related to anxiety and depressive symptoms.10-12 Although the mechanism of this relationship is unclear, some have speculated that a high intake of carbohydrates may ameliorate or prevent depressive symptoms but also lead to weight gain.13 Despite interest in the relationship of weight and depression, the relationship of weight and body habitus with the risk of suicide has received far less attention. Suicide remains the 11th leading cause of death in the United States,14 but identification of remediable risk factors for suicide has been difficult. Based on the links between obesity and depression, it might be hypothesized that overweight and obese individuals would have a higher risk of suicide. A small number of studies with limited information on anthropomorphic measures and potentially confounding factors have suggested exactly the opposite, at least among men,15,16 although an inverse association has not been detected in all studies.17,18 To determine the associations of body mass index (BMI) and other anthropomorphic measures with the risk of death from suicide (hereafter referred to as suicide death), we studied more than 45 000 participants of the Health Professionals Follow-up Study (HPFS). The HPFS is a prospective cohort study of male health care professionals in the United States for whom 16-year follow-up on mortality,19 including death related to suicide, and repeated, validated information on BMI, diet, and physical activity is available. Methods The HPFS is a prospective investigation of 51 529 US male dentists, pharmacists, veterinarians, optometrists, osteopathic physicians, and podiatrists aged 40 to 75 years who returned a mailed questionnaire regarding diet and medical history in 1986. Follow-up questionnaires were sent biennially to update information on exposures and newly diagnosed illnesses. For this analysis, we excluded men with cancer (other than nonmelanoma skin cancer) at baseline (n = 2002) and men whose initial questionnaires were missing information on BMI (n = 1094) or had inadequate dietary information or other technical deficiencies (n = 1678), leaving 46 755 men available for analysis. Anthropomorphic measures Participants reported body weight and height at baseline in 1986 and body weight alone in every biennial questionnaire afterward. At baseline, participants also reported their estimated weight 5 years previously (ie, in 1981) and their estimated weight at 21 years of age. As a measure of weight cycling, participants also reported the difference in their maximal and minimal weight in the preceding 2 years beginning in 1990. Information on waist and hip circumference was collected in a brief follow-up questionnaire sent out in 198720 that included paper tape measures, detailed instructions, and illustrations to assist the men in measuring standing waist and hip circumferences. Participants were instructed to measure their waist circumference at the umbilicus and their hip circumference as the largest diameter between the waist and thighs. We calculated the BMI (measured as weight in kilograms divided by height in meters squared) and waist-hip ratio for each participant. Self-reported anthropomorphic measures have been compared with standardized measures conducted by a technician in a subset of this cohort.21 Self-reported and measured weight were strongly correlated (r = 0.97); the mean difference was an underreporting by 1.06 kg. The corresponding correlation coefficients between self-reported and measured waist and hip circumferences were 0.95 and 0.88, respectively. We categorized BMI into 6 categories spanning approximately 2 U each as in previous analyses22; in sensitivity analyses, we also examined it in quintiles, in World Health Organization (WHO) categories, and as a continuous variable after winsorization at the upper 95th percentile to minimize the influence of outliers.23 We examined the height and waist-hip ratio in quintiles. Physical activity Physical activity was derived from the biennial questionnaires, which included a series of questions on the average total time per week spent on specific activities during the previous year. Total weekly energy expenditure was expressed as metabolic equivalent–hours.24 The validity and reproducibility of physical activity measurements were determined with a past-week recall, four 1-week activity diaries, and a step test among a subset of participants, as reported elsewhere.25 The correlations between diary-based and questionnaire-based activity scores were 0.28 for nonvigorous activity and 0.58 for vigorous activity. We categorized physical activity into approximate quintiles as in previous analyses.26 Other covariates We assessed dietary factors using a 131-item food frequency questionnaire, which has been validated with dietary records.27 In 1988, participants responded to a series of questions regarding social support, including the number of close friends and close relatives they had, the number of friends and relatives they saw at least monthly, how often they attended religious gatherings, and how many hours per week they spent in community or social groups.28 Participants reported use of antidepressant medication every 2 years, beginning with the 1990 questionnaire; doses were not reported. As in previous analyses,29 region of residence was categorized into the following 4 geographic regions, according to US Census definitions: West, Midwest, South, and Northeast. Participants reported prevalent and incident illnesses at every questionnaire; cases of cancer, coronary heart disease, and diabetes are subsequently adjudicated by physicians blinded to questionnaire information.20,22,30 Participants also reported their employment and marital status every 2 years; response options were full-time, part-time, retired, or disabled for employment, and married, divorced, widowed, or never married for marital status. Assessment of suicide deaths We confirmed deaths when reported by families, postal officials, or the National Death Index, with a combined follow-up rate in a companion study exceeding 98%.31 Physicians reviewed death certificates and hospital or pathology reports to classify individual causes of death and were unaware of participants' reported questionnaire results. Deaths caused by self-inflicted or external injury were classified according to the underlying causes listed on the death certificate. For this study, we included deaths classified with codes E950 through E959 of the International Classification of Diseases, Eighth Revision. We also separately examined suicide deaths associated with firearm use (code E955) and other causes. Quality of life In 1998, a supplemental questionnaire was administered to men with prostate cancer and control participants who were free of prostate cancer32,34; the questionnaire included the Medical Outcomes Study 36-Item Short-Form Health Survey.35 For each case, 2 healthy controls matched on age were selected; the response rate among the controls was 93%. Of the 2217 controls who returned questionnaires, 2103 had baseline information on BMI and nutrient intake, and 1829 had complete information on the Mental Component Summary (MCS) and Physical Component Summary scales, summary measures of mental and physical health–related quality of life, respectively. The scales have population means of 50 and SDs of 10, with higher scores indicating better quality of life in that domain. Statistical analysis We calculated person-years for each participant from the date of return of the 1986 questionnaire to the date of death or to February 2002. Analyses incorporating the waist-hip ratio or social support began at the return of the 1988 questionnaire, and the corresponding analyses of antidepressant use began at the return of the 1990 questionnaire. We calculated incidence rates by dividing the number of suicides in each category by person-years of follow-up in that category. In multivariate analyses, we simultaneously controlled for age, smoking (never, former, and 1-14, 15-24, and ≥25 cigarettes per day),36 alcohol use (in 4 categories), geographic region, physical activity (in 5 categories), and marital status (in 4 categories). In sensitivity analyses, we also adjusted individually for intakes of energy, glycemic load,37 caffeine,38 and ω-3 fatty acids39 (all in quintiles); profession (in 6 categories); measures of social support (in 5 categories each); employment status; or incident major illnesses (cancer, diabetes, and coronary heart disease). Nutrient intakes were updated every 4 years; other covariates (other than region and profession) were updated every 2 years. For multivariate analysis, we used Cox regression with time-dependent covariates; we noted no violations of the proportional hazards assumption. Our analyses used updated measures of BMI, in which we prospectively assessed the relative risk of suicide death in 2-year increments, based on the weight reported in the preceding questionnaire. Thus, we used the 1986 questionnaire to determine the risk of suicide death during the 1986-1988 period, the 1988 questionnaire for the 1988-1990 period, etc. In all analyses, we excluded men with prevalent malignancy at baseline and stopped updating BMI at the diagnosis of incident malignancy; in sensitivity analyses, we further excluded follow-up after diagnosis of malignancy during follow-up. We also examined baseline BMI alone and with a 2-year lag to minimize the inclusion of individuals with preexisting illnesses. For quality of life, we used linear regression to examine the relationship of BMI in 1986 or 1998 with MCS and Physical Component Summary scores, adjusting for the same covariates as in Cox models; however, of these covariates, only age and BMI were independently associated with MCS scores. Results Baseline characteristics of the HPFS cohort according to BMI are shown in Table 1; similar characteristics according to physical activity have been reported previously.26 A total of 131 men died of suicide during follow-up, ie, a suicide mortality rate of 19 per 100 000 person-years. The median age at suicide death was 64 years (interquartile range, 56-74 years). Table 2 shows incidence rates and hazard ratios (HRs) for suicide according to BMI. There was a stepwise decrease in the suicide mortality rate with increasing BMI, with a roughly 4-fold gradient in risk even after adjustment. Additional adjustment for incident disease or social support did not influence the results; incident cancer was strongly related to suicide risk (HR, 2.23; 95% confidence interval [CI], 1.27-3.93), but diabetes was not (HR, 1.31; 95% CI, 0.68-2.53), even without adjustment for BMI (HR, 1.23; 95% CI, 0.64-2.36). Results also tended to be similar for suicide deaths related to firearms and other causes. Additional adjustment for energy intake, caffeine, or ω-3 fatty acid intake also did not influence our findings (data not shown). Analyses restricted to nonsmokers, those using a 2-year lag or baseline BMI, and those adjusted for antidepressant use (a potential intermediate) all attenuated the HRs to a moderate degree, although at least 2-fold gradients in risk remained in all cases. Antidepressant use itself was also strongly associated with risk (HR, 5.94; 95% CI, 3.03-11.64). Table 2 further shows analyses using alternate cut points for BMI. Using the WHO definition of normal weight, there was an inverse association of BMI and risk through BMI values below 35.0; risk did not appear to be lower at the highest levels of BMI, but only 5 cases of suicide occurred in this subgroup. Table 3 shows the relationships of waist-hip ratio, height, and physical activity with risk of suicide. There was an inverse trend between waist-hip ratio and suicide risk, although it was less strong than that seen for BMI and not statistically significant. There was no evidence of any association between height or physical activity and risk. Table 4 shows the relationship of a 1-U increase in BMI with risk of suicide, overall and in sensitivity analyses. Risk was 11% lower for each 1-U increase in the full cohort. Neither censoring at the time of incident malignancy nor incorporating a 4-year lag between the assessment of exposure and onset of follow-up altered this finding. Even in analyses excluding men in the lowest BMI category to minimize confounding related to comorbid illness, there was a 10% lower risk for a 1-U increase in BMI. The inverse association of BMI with suicide was also not altered by additional adjustment for glycemic load (HR, 0.89; 95% CI, 0.84-0.95), and it did not differ in men whose glycemic load was above or below the median level; glycemic load per se was not associated with risk. The association was also identical when adjusted for work status, although disability itself was associated with a higher risk (HR, 3.20; 95% CI, 1.20-8.58). Stratified analyses suggested generally similar relationships in all subgroups, with no statistically significant interactions; although the association appeared to be stronger in younger men, even when BMI and age were modeled as continuous variables, the interaction was not significant (P=.14). When we analyzed change in BMI during follow-up controlling for previous BMI, a 1-U increase in BMI was associated with a nonsignificant 11% lower subsequent risk. When we further excluded decreases in BMI of 0.5 U or more to minimize the contribution of weight loss among participants before suicide, a 1-U increase in BMI was associated with a nonsignificant 17% lower subsequent risk (HR, 0.83; 95% CI, 0.56-1.21; P = .32). We also evaluated the association of self-reported BMI at 21 years of age with risk of suicide mortality during follow-up. The HR associated with a 1-U increase in BMI at 21 years of age was 0.95 (95% CI, 0.89-1.01); the corresponding HR for a 1-U gain in BMI from 21 years of age to baseline was 0.93 (95% CI, 0.87-1.00). Large amounts of weight change within a 2-year period tended to be associated with a higher risk, although estimates were not statistically significant (HR for the maximal weight difference within the last 2 years of ≥6.8 kg vs ≤1.8 kg, 1.61; 95% CI, 0.79-3.31) and did not alter the association of BMI with risk (HR per 1-U increment in BMI, 0.90; 95% CI, 0.84-0.97). Table 5 shows the relationship of BMI and mental health–related quality of life using the baseline BMI in 1986 or the BMI at the time of the questionnaire in 1998. The Pearson correlation coefficients of 1986 and 1998 BMI with MCS scores were 0.11 and 0.09, respectively (P<.001 for both). Similar to the results seen for suicide risk, MCS scores tended to be higher with increasing BMI, even when measured at baseline 12 years before the administration of the 36-Item Short-Form Health Survey. In contrast, 1986 and 1998 BMIs were inversely correlated with Physical Component Summary scores, with Pearson correlation coefficients of −0.15 and −0.11, respectively (P<.001 for both). Comment In this prospective cohort study of male health care professionals, higher BMI was associated with a lower risk of suicide mortality. There was a stepwise gradient in risk, the results were largely unchanged in sensitivity analyses, and there was a similar relationship of BMI with mental health–related quality of life. A small number of previous studies support an inverse relationship of BMI with risk of suicide death, although data on suicide attempts generally suggest the opposite. In a case-control analysis of the General Practice Research Database, hypertensive adults with a BMI of 25 or higher had an adjusted odds ratio for suicide of 0.33 (95% CI, 0.11-0.98) compared with leaner hypertensive adults.15 Using the military records of nearly 1.3 million Swedish male teens, Magnusson and colleagues16 found an inverse association of teenage BMI with the risk of suicide, with an HR of 0.85 (95% CI, 0.79-0.91) for each 5-U increment in BMI, although no information on adult anthropometric measures was available. However, suicide attempts tended to be positively associated with BMI in a representative German sample40 and among teenagers41 and female physicians42 in the United States. Dong and colleagues43 also found positive associations of marked obesity with suicide attempts in 2 parallel studies, but more typical levels of obesity were not assessed. The increased prevalence of overweight and obesity in the United States in the 1990s was accompanied by a 6% decrease in suicide mortality rates but little apparent change in the rate of suicide attempts.44 The mechanisms by which adult BMI could be inversely linked to the risk of suicide death are as yet uncertain. One possibility is the role of insulin in dictating mood. The drop in maternal insulin levels after birth has been hypothesized to precipitate postpartum depression,45 perhaps related to the ability of insulin to stimulate hypothalamic serotonin release.46 In a study of 30 patients with type 2 diabetes, 4 weeks of insulin administration led to marked improvements in depressive symptoms and subjective well-being, even with minimally altered glycemic control.47 Given that BMI has a continuous and positive relationship with insulin levels at all levels of weight,48 even among children,49 and that weight gain over time has a similar relationship,50 it is plausible that BMI could be linked to lower risk of suicide through circulating levels of insulin. At the same time, we generally found little evidence to support the hypothesis that specific dietary factors mediate the relationship of obesity and suicide. Although carbohydrate intake in particular has been postulated to improve mood and increase weight, adjustment for glycemic load had no effect on our estimates of risk. Glycemic load is a measure of the acute glycemic response51 and is not necessarily related to fasting insulin levels.52 Taken with our findings on BMI, these results suggest that physiological responses to glycemic load may be less relevant for suicide than is adiposity per se. Our findings in this population of men may not readily generalize to all women. For example, in an analysis of the 1992 National Longitudinal Alcohol Epidemiologic Survey, Carpenter and colleagues53 found that a 10-U increment in BMI was associated with increased odds of suicide attempts among women (odds ratio, 1.22; 95% CI, 0.91-1.63), but significantly decreased odds among men (odds ratio, 0.45; 95% CI, 0.40-0.51). Several cross-sectional studies have suggested that men and women differ in their relationship between BMI and depressive symptoms.10,12,54 It is possible that the stigma attached to obesity may be particularly acute among women,55 although this has not been borne out in all studies.56 Alternately, the association between endogenous estrogen levels and depression may complicate the picture for women. Estrogen alters serotonergic neurotransmission and receptor distribution57,58 and has been suggested to ameliorate depression in some clinical trials.59 In support of this hypothesis, BMI may be inversely associated with depressive symptoms among postmenopausal but not premenopausal women.11 The relationship of BMI and suicide tended to be stronger among younger than older men, although a formal test of interaction did not yield significant results. This may reflect the role of chance, as in any subgroup analysis. We also could not readily evaluate suicide among men in their third and fourth decades of life, in whom suicide accounts for a disproportionately large proportion of deaths, albeit at a lower absolute rate than among older men. However, BMI tends to be a better measure of adiposity in the young, whereas it is disproportionately influenced by underlying illness and changes in muscle mass in older adults.60 If the weaker relation in older men is confirmed, it would tend to support further the importance of adiposity per se in lowering the risk of suicide. The potential interrelationships of BMI, depressive symptoms and major depressive disorders, antidepressant medication, and risk of suicide are complex and likely bidirectional. On the one hand, BMI may have direct positive effects on mood that mediate the observed association of BMI with suicide, such as through elevated insulin levels. On the other hand, depression and the medications used to treat it may influence eating and exercise habits and subsequently BMI. This bidirectionality is also likely to be true for other emotional and relational factors for which we did not have information. We had information on antidepressant use, a potential contributor to suicide even in adults,61 and social support, neither of which appeared to confound the association of BMI with the risk of suicide death to any substantial degree. However, we did not have systematic information on depression or depressive symptoms. Future studies that incorporate repeated measures of both depression and BMI in large samples of participants and methodological approaches that account for the possibility that depression acts as both a mediator and a confounder for the effect of BMI on suicide risk62,63 will ultimately be required to clarify these relationships definitively. The HPFS is also subject to other limitations. We did not have information on suicide attempts, and hence we could not ascertain whether BMI is related to the risk of suicide attempts or their case-fatality rate in this population. This latter issue will require additional study, given the somewhat contradictory results of previous studies relating BMI to suicide attempts and suicide death. Participants in the HPFS are not necessarily representative of the total US population, which limits the generalizability of our results but provides a sample of uniform sex, educational attainment, and occupational status. Moreover, the rate of suicide death among these middle-aged and older men is similar to that found in other large cohort studies.64 As in other cohort studies, our definition of suicide death was based on death certificates, an approach that appears to be valid.65 However, treating practitioners may have misclassified some cases of suicide death, particularly in cases where other causes of death may have been plausible (eg, among men with known chronic disease), although cases of suicide death related to firearm use may be less prone to underreporting and were related inversely with BMI. Although we relied on self-reported measures of height and weight in this study, we have previously validated them in this population and all assessments were made prospectively. A related limitation is the relatively truncated range of BMI noted among these men, and we cannot evaluate the possibility that greater degrees of obesity (eg, BMI of ≥40) are associated with a higher risk of suicide, as suggested elsewhere.43 Our analyses were necessarily limited by the relative rarity of death related to suicide. Although our primary analyses yielded reasonably precise results, our stratified analyses—and particularly stratified analyses that examined categories of BMI—yielded considerably wider CIs. Larger studies with greater numbers of cases of suicide mortality will be needed to overcome this limitation. In conclusion, higher BMI was strongly associated with a lower risk of suicide mortality among men, with a clear gradient in risk between the highest and lowest BMI levels. Given the strength of this relationship, research to identify its mechanisms could lead to important insights into prevention of this important cause of death. Correspondence: Kenneth J. Mukamal, MD, MPH, MA, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 1309 Beacon St, Second Floor, Boston, MA 02446 (kmukamal@bidmc.harvard.edu). Accepted for Publication: November 13, 2006. Author Contributions: Dr Mukamal had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Mukamal and Rimm. Acquisition of data: Rimm. Analysis and interpretation of data: Mukamal, Kawachi, Miller, and Rimm. Drafting of the manuscript: Mukamal. Critical revision of the manuscript for important intellectual content: Kawachi, Miller, and Rimm. Statistical analysis: Mukamal and Miller. Obtained funding: Mukamal and Rimm. Financial Disclosure: None reported. Funding/Support: This study was supported by grants AA014900, HL35464, CA55075, AA11181, HL35464, and CA55075 from the National Institutes of Health. Role of the Sponsor: The funding sources for this work had no role in the analysis, presentation, or publication of results. References 1. Kessler RCChiu WTDemler OMerikangas KRWalters EE Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry 2005;62617- 627PubMedGoogle Scholar 2. Flegal KM Epidemiologic aspects of overweight and obesity in the United States. Physiol Behav 2005;86599- 602PubMedGoogle Scholar 3. Evans DLCharney DSLewis L et al. Mood disorders in the medically ill: scientific review and recommendations. Biol Psychiatry 2005;58175- 189PubMedGoogle Scholar 4. Stunkard AJFaith MSAllison KC Depression and obesity. Biol Psychiatry 2003;54330- 337PubMedGoogle Scholar 5. Franko DLStriegel-Moore RHThompson DSchreiber GBDaniels SR Does adolescent depression predict obesity in black and white young adult women? Psychol Med 2005;351505- 1513PubMedGoogle Scholar 6. Roberts REDeleger SStrawbridge WJKaplan GA Prospective association between obesity and depression: evidence from the Alameda County Study. Int J Obes Relat Metab Disord 2003;27514- 521PubMedGoogle Scholar 7. Mamplekou EKomesidou VBissias CPapakonstantinou AMelissas J Psychological condition and quality of life in patients with morbid obesity before and after surgical weight loss. Obes Surg 2005;151177- 1184PubMedGoogle Scholar 8. Waters GSPories WJSwanson MSMeelheim HDFlickinger EGMay HJ Long-term studies of mental health after the Greenville gastric bypass operation for morbid obesity. Am J Surg 1991;161154- 158PubMedGoogle Scholar 9. Zimmermann UKraus THimmerich HSchuld APollmacher T Epidemiology, implications and mechanisms underlying drug-induced weight gain in psychiatric patients. J Psychiatr Res 2003;37193- 220PubMedGoogle Scholar 10. Crisp AHMcGuiness B Jolly fat: relation between obesity and psychoneurosis in general population. BMJ 1976;17- 9PubMedGoogle Scholar 11. Jasienska GZiomkiewicz AGorkiewicz MPajak A Body mass, depressive symptoms and menopausal status: an examination of the “jolly fat” hypothesis. Womens Health Issues 2005;15145- 151PubMedGoogle Scholar 12. Palinkas LAWingard DLBarrett-Connor E Depressive symptoms in overweight and obese older adults: a test of the “jolly fat” hypothesis. J Psychosom Res 1996;4059- 66PubMedGoogle Scholar 13. Wurtman RJWurtman JJ Carbohydrates and depression. Sci Am 1989;26068- 75PubMedGoogle Scholar 14. Hoyert DLKung HCSmith BL Deaths: preliminary data for 2003. Natl Vital Stat Rep 2005;531- 48PubMedGoogle Scholar 15. Gasse CDerby LEVasilakis CJick H Risk of suicide among users of calcium channel blockers: population based, nested case-control study. BMJ 2000;3201251PubMedGoogle Scholar 16. Magnusson PKRasmussen FLawlor DATynelius PGunnell D Association of body mass index with suicide mortality: a prospective cohort study of more than one million men. Am J Epidemiol 2006;1631- 8PubMedGoogle Scholar 17. Jiang GXRasmussen FWasserman D Short stature and poor psychological performance: risk factors for attempted suicide among Swedish male conscripts. Acta Psychiatr Scand 1999;100433- 440PubMedGoogle Scholar 18. Allebeck PBergh C Height, body mass index and mortality: do social factors explain the association? Public Health 1992;106375- 382PubMedGoogle Scholar 19. Kawachi IColditz GAAscherio A et al. A prospective study of social networks in relation to total mortality and cardiovascular disease in men in the USA. J Epidemiol Community Health 1996;50245- 251PubMedGoogle Scholar 20. Chan JMRimm EBColditz GAStampfer MJWillett WC Obesity, fat distribution, and weight gain as risk factors for clinical diabetes in men. Diabetes Care 1994;17961- 969PubMedGoogle Scholar 21. Rimm EBStampfer MJColditz GAChute CGLitin LBWillett WC Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990;1466- 473PubMedGoogle Scholar 22. Giovannucci ERimm EBLiu Y et al. Body mass index and risk of prostate cancer in US health professionals. J Natl Cancer Inst 2003;951240- 1244PubMedGoogle Scholar 23. Fernandez JREtzel CBeasley TMShete SAmos CIAllison DB Improving the power of sib pair quantitative trait loci detection by phenotype winsorization. Hum Hered 2002;5359- 67PubMedGoogle Scholar 24. Ainsworth BEHaskell WLLeon AS et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;2571- 80PubMedGoogle Scholar 25. Chasan-Taber SRimm EBStampfer MJ et al. Reproducibility and validity of a self-administered physical activity questionnaire for male health professionals. Epidemiology 1996;781- 86PubMedGoogle Scholar 26. Tanasescu MLeitzmann MFRimm EBWillett WCStampfer MJHu FB Exercise type and intensity in relation to coronary heart disease in men. JAMA 2002;2881994- 2000PubMedGoogle Scholar 27. Feskanich DRimm EBGiovannucci EL et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993;93790- 796PubMedGoogle Scholar 28. Eng PMRimm EBFitzmaurice GKawachi I Social ties and change in social ties in relation to subsequent total and cause-specific mortality and coronary heart disease incidence in men. Am J Epidemiol 2002;155700- 709PubMedGoogle Scholar 29. Michaud DSClinton SKRimm EBWillett WCGiovannucci E Risk of bladder cancer by geographic region in a US cohort of male health professionals. Epidemiology 2001;12719- 726PubMedGoogle Scholar 30. Rimm EBGiovannucci ELWillett WC et al. Prospective study of alcohol consumption and risk of coronary disease in men. Lancet 1991;338464- 468PubMedGoogle Scholar 31. Rich-Edwards JWCorsano KAStampfer MJ Test of the National Death Index and Equifax Nationwide Death Search. Am J Epidemiol 1994;1401016- 1019PubMedGoogle Scholar 32. Bacon CGGiovannucci ETesta MGlass TAKawachi I The association of treatment-related symptoms with quality-of-life outcomes for localized prostate carcinoma patients. Cancer 2002;94862- 871PubMedGoogle Scholar 33. Bacon CGGiovannucci ETesta MKawachi I The impact of cancer treatment on quality of life outcomes for patients with localized prostate cancer. J Urol 2001;1661804- 1810PubMedGoogle Scholar 34. Bacon CGKawachi I Quality-of-life differences among various populations of localized prostate cancer patients: 2001. Curr Urol Rep 2002;3239- 243PubMedGoogle Scholar 35. Ware JE JrSherbourne CD The MOS 36-Item Short-Form Health Survey (SF-36), I: conceptual framework and item selection. Med Care 1992;30473- 483PubMedGoogle Scholar 36. Miller MHemenway DRimm E Cigarettes and suicide: a prospective study of 50 000 men. Am J Public Health 2000;90768- 773PubMedGoogle Scholar 37. Salmeron JAscherio ARimm EB et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997;20545- 550PubMedGoogle Scholar 38. Kawachi IWillett WCColditz GAStampfer MJSpeizer FE A prospective study of coffee drinking and suicide in women. Arch Intern Med 1996;156521- 525PubMedGoogle Scholar 39. Freeman MP Omega-3 fatty acids in psychiatry: a review. Ann Clin Psychiatry 2000;12159- 165PubMedGoogle Scholar 40. Brunner JBronisch TPfister HJacobi FHofler MWittchen HU High cholesterol, triglycerides, and body-mass index in suicide attempters. Arch Suicide Res 2006;101- 9PubMedGoogle Scholar 41. Eaton DKLowry RBrener NDGaluska DACrosby AE Associations of body mass index and perceived weight with suicide ideation and suicide attempts among US high school students. Arch Pediatr Adolesc Med 2005;159513- 519PubMedGoogle Scholar 42. Frank EDingle AD Self-reported depression and suicide attempts among US women physicians. Am J Psychiatry 1999;1561887- 1894PubMedGoogle Scholar 43. Dong CLi WDLi DPrice RA Extreme obesity is associated with attempted suicides: results from a family study. Int J Obes (Lond) 2006;30388- 390PubMedGoogle Scholar 44. Kessler RCBerglund PBorges GNock MWang PS Trends in suicide ideation, plans, gestures, and attempts in the United States, 1990-1992 to 2001-2003. JAMA 2005;2932487- 2495PubMedGoogle Scholar 45. Chen THLan THYang CYJuang KD Postpartum mood disorders may be related to a decreased insulin level after delivery. Med Hypotheses 2006;66820- 823PubMedGoogle Scholar 46. Orosco MRouch CGerozissis K Activation of hypothalamic insulin by serotonin is the primary event of the insulin-serotonin interaction involved in the control of feeding. Brain Res 2000;87264- 70PubMedGoogle Scholar 47. Reza MTaylor CDTowse KWard JDHendra TJ Insulin improves well-being for selected elderly type 2 diabetic subjects. Diabetes Res Clin Pract 2002;55201- 207PubMedGoogle Scholar 48. Shiwaku KAnuurad EEnkhmaa B et al. Predictive values of anthropometric measurements for multiple metabolic disorders in Asian populations. Diabetes Res Clin Pract 2005;6952- 62PubMedGoogle Scholar 49. Janssen IKatzmarzyk PTSrinivasan SR et al. Combined influence of body mass index and waist circumference on coronary artery disease risk factors among children and adolescents. Pediatrics 2005;1151623- 1630PubMedGoogle Scholar 50. Norman JEBild DLewis CELiu KWest DS The impact of weight change on cardiovascular disease risk factors in young black and white adults: the CARDIA study. Int J Obes Relat Metab Disord 2003;27369- 376PubMedGoogle Scholar 51. Wolever TMJenkins DJJenkins ALJosse RG The glycemic index: methodology and clinical implications. Am J Clin Nutr 1991;54846- 854PubMedGoogle Scholar 52. Liese ADSchulz MFang F et al. Dietary glycemic index and glycemic load, carbohydrate and fiber intake, and measures of insulin sensitivity, secretion, and adiposity in the Insulin Resistance Atherosclerosis Study. Diabetes Care 2005;282832- 2838PubMedGoogle Scholar 53. Carpenter KMHasin DSAllison DBFaith MS Relationships between obesity and DSM-IV major depressive disorder, suicide ideation, and suicide attempts: results from a general population study. Am J Public Health 2000;90251- 257PubMedGoogle Scholar 54. Heo MPietrobelli AFontaine KRSirey JAFaith MS Depressive mood and obesity in US adults: comparison and moderation by sex, age, and race. Int J Obes (Lond) 2006;30513- 519PubMedGoogle Scholar 55. Miller CTRothblum EDBarbour LBrand PAFelicio D Social interactions of obese and nonobese women. J Pers 1990;58365- 380PubMedGoogle Scholar 56. Myers ARosen JC Obesity stigmatization and coping: relation to mental health symptoms, body image, and self-esteem. Int J Obes Relat Metab Disord 1999;23221- 230PubMedGoogle Scholar 57. Fink GSumner BEMcQueen JKWilson HRosie R Sex steroid control of mood, mental state and memory. Clin Exp Pharmacol Physiol 1998;25764- 775PubMedGoogle Scholar 58. Rybaczyk LABashaw MJPathak DRMoody SMGilders RMHolzschu DL An overlooked connection: serotonergic mediation of estrogen-related physiology and pathology. BMC Womens Health 2005;512PubMedGoogle Scholar 59. Soares CNAlmeida OPJoffe HCohen LS Efficacy of estradiol for the treatment of depressive disorders in perimenopausal women: a double-blind, randomized, placebo-controlled trial. Arch Gen Psychiatry 2001;58529- 534PubMedGoogle Scholar 60. Flint AJRimm EB Commentary: obesity and cardiovascular disease risk among the young and old: is BMI the wrong benchmark? Int J Epidemiol 2006;35187- 189PubMedGoogle Scholar 61. Gunnell DSaperia JAshby D Selective serotonin reuptake inhibitors (SSRIs) and suicide in adults: meta-analysis of drug company data from placebo controlled, randomised controlled trials submitted to the MHRA's safety review. BMJ 2005;330385[published correction appears in BMJ. 2006;333:30]PubMeddoi:10.1136/bmj.330.7488.385Google Scholar 62. Hernan MABrumback BARobins JM Estimating the causal effect of zidovudine on CD4 count with a marginal structural model for repeated measures. Stat Med 2002;211689- 1709PubMedGoogle Scholar 63. Hernan MACole SRMargolick JCohen MRobins JM Structural accelerated failure time models for survival analysis in studies with time-varying treatments. Pharmacoepidemiol Drug Saf 2005;14477- 491PubMedGoogle Scholar 64. Tanskanen ATuomilehto JViinamaki HVartiainen ELehtonen JPuska P Joint heavy use of alcohol, cigarettes and coffee and the risk of suicide. Addiction 2000;951699- 1704PubMedGoogle Scholar 65. Moyer LABoyle CAPollock DA Validity of death certificates for injury-related causes of death. Am J Epidemiol 1989;1301024- 1032PubMedGoogle Scholar

Journal

Archives of Internal MedicineAmerican Medical Association

Published: Mar 12, 2007

There are no references for this article.