Years of Life Lost Due to Obesity

Obesity has become pandemic in the United States. Currently, 2 in 3 US adults are classified as overweight or obese, compared with fewer than 1 in 4 in the early 1960s.^{1, 2} Although still viewed more as a cosmetic rather than a health problem by the general public, excess weight is a major risk factor for premature mortality, cardiovascular disease, type 2 diabetes mellitus, osteoarthritis, certain cancers, and other medical conditions.³ Obesity accounts for more than 280 000 deaths annually in the United States and will soon overtake smoking as the primary preventable cause of death if current trends continue.⁴ Indeed, obesity is already associated with greater morbidity and poorer health-related quality of life than smoking, problem drinking, or poverty.⁵ Despite this, excess weight has not received the same attention from clinicians and policymakers as have other threats to health such as tobacco use, hypertension, or hypercholesterolemia. Given these circumstances, it is not surprising that obesity rates continue to climb, even as significant reductions in other risk factors have been achieved.⁶

Although powerful social and cultural forces drive the pandemic of excess weight, health care professionals can help to counteract these trends by educating patients about the serious health risks of obesity and prescribing concrete interventions, such as regular physical activity and healthier food choices. In this issue of THE JOURNAL, Fontaine et al⁷ provide clinicians with a valuable metric for concisely and tangibly conveying the deleterious effect of obesity to overweight individuals, years of life lost (YLL) defined as the difference between the number of years that one would be expected to live if one were not obese and life expectancy if one were obese. By combining nationally representative databases in a novel and creative analysis, the authors show that even a moderate amount of excess weight confers a noticeable diminution in life expectancy and that, as degree of overweight increases, a striking and steady contraction of life span occurs. However, these patterns are more apparent among whites. Among blacks, moderate and marked obesity (in both men and women) predicts reduced longevity, but lesser degrees of overweight do not. Indeed, an adverse effect of excess weight is not observed in black women until a body mass index (BMI) of 37 is reached.

With its clear intuitive appeal, the concept of YLL is likely to provide compelling motivation for weight control efforts, at least among whites. Nevertheless, it would be a great disservice to blacks if these results were used to promulgate the concept that excess weight is not harmful to them. The finding that blacks lose less life than whites for a comparable amount of overweight is likely due in large part to methodologic limitations many of which are noted by the authorsthat distort inferences about the role of weight in predicting health outcomes. One such limitation is the failure to account for the effects of preexisting disease. Illness-induced weight loss may obscure positive dose-response relationships between adiposity and mortality.⁸ To reduce potential bias due to this reverse causation, investigators often exclude respondents with known, diagnosed diseases that might cause both weight loss and premature death. However, Fontaine et al did not use this strategy. Even if they had, the problem would not have been solved because certain unrecognized conditions (such as preclinical cancers, depression, substance abuse, diabetes mellitus, or early pulmonary or cardiac failure) can cause insidious weight loss years before clinical diagnosis. Given their generally disadvantaged socioeconomic status compared with whites, blacks are likely to have less access to high-quality health care and thus a higher prevalence of these undiagnosed conditions than are whites. These differences would disproportionately bias the results for blacks toward the null.

Moreover, socioeconomic status, which was not adjusted for in these analyses, is likely to be a powerful confounder of the association between BMI and mortality apart from its effect on undetected disease. In whites, numerous studies show an inverse relationship between socioeconomic status and obesity, especially in women.⁹ However, the data are less consistent for blacks,^{10, 11} perhaps because of the higher prevalence of extreme poverty in this ethnic group. At the extreme lower tail of the income distribution, involuntary food deprivation and food insufficiency come into play. Thus, such confounding may lead to a greater degree of underestimation of the relationship between BMI and mortality among blacks than among whites.

Another confounder of the association between weight and mortality is smoking status. Because smoking is more prevalent among lean individuals and is also a strong independent risk factor for death, failure to adjust for its effects will produce an artifactually elevated mortality among the lean.⁸ Although the authors include smoking in their statistical models, their use of a relatively crude categorical variable (current, former, or never smoker) instead of a more precise measure of exposure (eg, number of cigarettes per day) leads to residual confounding by this variable. In any event, the strongest relationship between weight and mortality is observed in studies of never smokers⁸; it would therefore be instructive to repeat the analyses among this subgroup.

Ethnic variations in the reliability of BMI as a measure of total adiposity¹² or visceral adiposity¹³ may also partly account for the observed discrepancies between whites and blacks. Similarly, age and sex variations in BMI may underlie some of the divergence between young and old, and between women and men, observed in both ethnic groups.¹⁴ In addition, because the life table (available at: http://www.cdc.gov/nchs/data/nvsr/nvsr50/nvsr50_06.pdf) and relative risk estimates are derived from sparser data in blacks than in whites, there is more statistical uncertainty surrounding the YLL estimates for blacks. Because confidence intervals are unavailable, one cannot determine whether the data are consistent with alternate conclusions.

Studies of total mortality are inherently insensitive in pinpointing the weight range associated with optimal health. Excess weight, like almost every other risk indicator, is unlikely to influence all causes of death. Relative risks associated with high BMI are far lower for all-cause mortality than for specific diseases, such as type 2 diabetes, myocardial infarction, and ischemic stroke.^{15, 16} Even a moderately elevated BMI is a significant predictor of hypertension and of diabetes in both whites and blacks, and the incidence rates of these conditions increase similarly in both ethnic groups as BMI increases.¹⁷ Additionally, while only limited data support the notion that intentional weight loss reduces total mortality, such weight loss has been shown to reduce disease-specific risks such as hypertension and type 2 diabetes in whites and in blacks.¹⁸ Indeed, because blacks may experience greater mortality from causes unrelated to obesity (eg, homicide) than do whites, the attenuation of the observed association between BMI and YLL due to these competing risks is likely to be more marked among blacks. Moreover, mortality is only a small part of the substantial burden of disease caused by obesity-related conditions such as diabetes, degenerative arthritis, hypertension, angina, and other nonfatal cardiovascular disease.

Thus, despite the powerful simplicity of the YLL metric as a tool for clinicians, it exposes only a small fraction of the toll of obesity. The use of a single outcomeeven if that outcome is mortalityto determine what constitutes a healthy weight and to set public health goals for weight reduction has limitations. A summary measure that incorporates information on the adverse impact of excess weight on morbidity and quality of life as well as on mortality may be preferable. The number of healthy (ie, disability-free) YLL is an alternative that could be developed and used by health care professionals and policymakers to highlight the total risk of obesity. Computing disability-adjusted life-years is clearly more complicated than calculating YLL, since subjective decisions would be required to determine what constitutes a disability and to develop severity-of-disability ratings by which to adjust the statistic. Until such a measure is developed, however, the YLL metric provides a useful strategy to illustrate the most tangible cost of obesity.