In this follow-up of a large, low-income population of black adults and white adults among whom obesity is common, the association between adult BMI and all-cause mortality differed by race. Being obese at cohort entry was not associated with elevated mortality among blacks; indeed, the lowest hazard ratios were observed among those with BMI of 30–34.9, and hazard ratios were near or even below 1.0 among those with extreme obesity. Among socioeconomically comparable whites, however, extreme obesity was associated with mortality increases up to 50% higher than among normal weight individuals.
Only a few cohort studies have had adequate samples of both blacks and whites to compare BMI–all-cause mortality patterns by race. This report includes one of the largest populations of blacks (nearly 53,000) thus far studied; further, the nearly 24,000 whites of similar socioeconomic status in the SCCS provided a comparison group to examine patterns across race while minimizing socioeconomic-related confounding. In the American Cancer Society's Cancer Prevention Study II that included 12,000 blacks, nonsmokers with no history of disease were found to have a significantly increased mortality risk among whites with extreme obesity (HR = 2.58 for males and 2.00 for females), but the association was weaker and nonsignificant among black males (HR = 1.35) and females (HR = 1.21) (20
). In the National Institutes of Health (NIH)-AARP cohort of adults aged 50–71 years, including 20,200 blacks, associations between BMI and mortality were also lower among blacks than whites (HRs for BMI >40 were 1.68 for black males, 1.82 for white males, 1.70 for black females, and 1.95 for white females) (9
). In a combined analysis of the First National Health and Nutrition Examination Survey Epidemiologic Follow-up Study and the National Health Interview Survey populations (including 19,000 blacks), the BMI associated with the lowest mortality risk was 3.1 kg/m2
higher among black men than white men and 1.5 kg/m2
higher among black women than white women (10
). Hence, the emerging pattern is generally consistent with our findings of a weaker association between mid-to-later-life BMI and all-cause mortality among blacks than whites. However, the Black Women's Health Study reported elevated mortality among 51,695 black women for all BMI categories of 25 or higher with the highest hazard ratio observed in extremely obese black women (HR = 2.19) (14
), a finding more in line with results from recent pooled analyses of whites (6
) than other cohort studies with large black populations.
The reasons for differing adult BMI-mortality patterns between blacks and whites are unknown. Socioeconomic differences have been postulated (12
) but seem unlikely to account for elevated BMI as a risk factor only for whites in this study, because income and education distributions were fairly similar between black participants and white participants (and residual differences were adjusted in the analysis). Among both blacks and whites, however, we did observe that mortality risks associated with obesity tended to be lower among those with lower socioeconomic status, perhaps because other detriments to health are more prominent among low income populations. Similar differences by education were also observed among black women in the Black Women's Health Study, where excess mortality among the obese was limited to those with greater than 12 years of education (14
). Other factors related to overall mortality that might differ by race could play a role in the black-white difference in the BMI-mortality relation. We examined the effect of cigarette smoking by adjusting for smoking in the analyses and by restricting analyses to nonsmokers, but in both, the differential BMI mortality patterns between blacks and whites persisted.
In cause-specific mortality analyses, the most pronounced differences between blacks and whites were seen for cardiovascular disease mortality, where more than 2-fold increases were seen in whites in obesity classes II and III, while only slight and nonsignificant increases were observed in blacks of similarly elevated BMI. The stronger association between BMI and cardiovascular disease (compared with cancer or other combined causes of death) has been observed in previous large studies of primarily white individuals (6
), as well black women (14
). The difference seems likely to be due to stronger correlations between obesity and factors directly associated with increased risk of cardiovascular disease (more so than for cancer or other nonexternal causes) including dyslipidemia, hypertension, and insulin resistance (21
). We have previously reported that the prevalence of diabetes, a strong cardiovascular disease risk factor, rose markedly with increasing BMI (22
). In ongoing analyses, we have also observed a higher prevalence of atrial fibrillation, another cardiovascular disease risk factor, among SCCS participants with high BMI (Loren Lipworth, Vanderbilt University, personal communication, 2011). Notably, for both our analyses of diabetes and atrial fibrillation within the SCCS, the association with obesity was stronger among whites than blacks, which may help to explain the larger racial difference seen for cardiovascular disease mortality than for cancer or other causes.
For all cancers, the differential pattern by race in the BMI-mortality association was evident only among males, but this finding should be viewed as preliminary given the relatively small number of cancers among whites. The major cause of cancer death among SCCS participants was lung cancer, which tends to be inversely associated with obesity (4
). Thus, the influence of lung cancer on total cancer mortality likely dampened the association with obesity and accentuated the difference between the cardiovascular disease-BMI and cancer-BMI associations. Further, the absence of a strong association between cancer mortality and BMI tends to minimize the possibility for a racial difference, further accentuating the strong racial difference in cardiovascular disease mortality.
The different patterns by race for the BMI-overall mortality associations, as well as the more pronounced racial difference for cardiovascular disease mortality, may indicate that equivalency of BMI does not mean equivalency of the specific body composition parameters most associated with mortality. Although BMI is highly correlated with overall body fat, the distribution of fat is not well captured by BMI (23
). As blacks have been shown to have less mean visceral fat than whites at the same level of BMI (24
) and because visceral fat is more strongly associated with metabolic disorders and cardiovascular disease than subcutaneous fat (23
), measures of adiposity such as waist circumference or waist/hip ratio may be better predictors of health outcomes than BMI, especially when examining populations with diverse racial backgrounds (25
). In the SCCS, despite having measured data on waist circumference and waist/hip ratio for only a small subset of the cohort with limited follow-up, we observed that waist circumference was significantly associated with all-cause mortality only among whites, similar to the associations we observed with BMI, whereas trends in hazard ratios associated with waist circumference among blacks were flat. Thus, while BMI may be a less sensitive indicator of the body size characteristics (i.e., abdominal obesity) predictive of increased mortality among blacks, further study is needed to help identify factors underlying the racial differences in mortality patterns observed here and in other studies, as well as to clarify whether excess weight is actually less detrimental to health in blacks.
Future research would also benefit from the assessment of differences between blacks and whites in genetic variants thought to be associated with obesity and potential mortality. Within the SCCS, we recently reported that mean serum levels of adiponectin, an adipose tissue-derived protein that plays a critical role in several physiologic pathways related to disease risk (28
), decreased monotonically with increasing BMI among white but not black women (29
). Furthermore, we found that single nucleotide polymorphisms in adiponectin-related genes were associated with the serum levels only among whites (30
). These preliminary findings suggest that genetically determined biologic attributes may be contributing to the differing patterns between obesity and overall mortality, but additional study is needed to clarify the roles of allelic variation and the potential mechanisms involved.
Limitations of this study first include the use of self-reported height and weight for a large proportion of participants. While height tends to be overreported and weight underreported (31
), data from the 1999 to 2004 National Health and Nutrition Examination Survey show that, despite errors in self-report, BMI categories based on self-reported values still generally demonstrate good agreement with BMI categories from measured values (32
). Furthermore, in the SCCS, BMI values calculated from self-reported height and weight were highly correlated with values calculated from abstracted community health center medical records overall (n
= 14,000; Pearson's correlation > 0.95) and within strata of race, income, and education, indicating that the self-reported values are generally of high quality. An additional limitation is the difficulty in controlling for existing disease, particularly with the relatively short follow-up currently available in the SCCS. It is likely that the higher mortality observed in underweight participants at least partially reflects preexisting disease processes. We attempted to address this issue by excluding the first 12 months of follow-up, but as the cohort is followed for longer periods, we anticipate that the high risks associated with lower weight will begin to dissipate, a trend that has already been seen in follow-up greater than 5 years after cohort entry.
This study also has several notable strengths. First, although the SCCS population is not reflective of the socioeconomic or racial distributions in the general US population because of the recruitment strategy through community health centers and the resulting overrepresentation of low-income individuals, it is a unique cohort in which to study health effects in blacks compared with whites because of both the large number of blacks and the comparability of socioeconomic status between the racial groups. The SCCS data add an important new dimension to the existing body of literature that for the most part has examined associations between obesity and mortality in middle or upper income cohorts, while the SCCS by contrast is composed of largely low-income individuals among whom obesity is more prevalent. Second, because of the large size of this cohort and the extensive data on smoking collected during the baseline questionnaire, we were able to carefully control for smoking, an important confounder of the mortality-obesity relation.
In summary, we found that racial differences in BMI-mortality associations exist between African Americans and whites of similarly low socioeconomic levels. These findings should stimulate additional research to untangle and identify the complex mechanisms underlying the observed racial differences and help in tailoring strategies to combat the adverse health effects of obesity in all population groups.