Putative biases in estimating mortality attributable to obesity in the US population

Abstract

Background: A recent analysis concluded that there were fewer excess deaths attributable to obesity in the US population than previously believed. This analysis may not have fully corrected for two putative biases, the regression-dilution and the reversecausation biases. It is not presently known whether correcting for these biases would increase estimates of excess deaths attributable to obesity. Methods: All-cause mortality hazard ratios (HRs) for different body mass index (BMI) categories were calculated and adjusted for confounding factors, using data from the prospective Atherosclerosis Risk in Communities Study. The analysis was based on 12 457 individuals aged 51-70 years and 606 all-cause deaths during a 5.3-year follow-up. The HRs were corrected for the regression-dilution and reverse-causation biases, and compared with data from a previously published study to evaluate the effects of correcting for these putative biases on estimates of excess deaths attributable to obesity in the US population.Results: The uncorrected all-cause mortality HR for obesity (BMIX30) was 1.26 (95% confidence interval (95% CI) ¼ 1.01-1.58), using the 21-25 kg/m 2 as ideal-weight category. Correcting for regression dilution increased the HR to 1.46 (95% CI ¼ 1.17-1.83). Correcting for both putative biases increased it further to 2.70 (95% CI ¼ 1.31-5.57). Such increases in HRs are consonant with increases of several hundred percent in estimates of deaths attributable to obesity in the US. Conclusions: Correcting for putative biases yielded increases in all-cause mortality HRs for obesity that correspond to substantial increases in estimates of excess deaths attributable to obesity in the US population.

Modeling Obesity Histories in Cohort Analyses of Health and Mortality

“…Some studies 19,20 create a variable representing the average level of obesity at various observations (i.e., dividing the sum of O 1 and O 2 by 2). Such a model is algebraically indistinguishable from a duration model when the duration variable is a linear transformation of the mean of O 1 and O 2 .…”

Considering the Inclusion of Metabolic and Cardiovascular Markers in the Panel Study of Income Dynamics

“…Evidence on the morbidity and mortality consequences of the obesity epidemic, however, is mixed. While a very low or very high BMI is associated with mortality, the shape of the relationship for the bulk of the distribution is not clear, with the overweight category recently found to be protective compared to normal weight (Flegal, Graubard, Williamson, & Gail 2007; Greenberg, Fontaine, & Allison 2007). While profiles of cardiovascular risk have improved across all BMI categories (Gregg, Cheng, Cadwell, et al 2005), there is evidence that levels of functional impairment have become worse for obese adults (Alley & Chang 2007).…”

BMI and mortality: sorting through the data to find the public health message

“…In contrast, Greenberg et al 9,10 focus on control of confounding rather than temporal issues and use the term 'regression dilution'. The term in itself is slightly misleading, as, depending on relations with errors in other variables, random error in the exposure can artificially inflate as well as deflate or dilute a regression coefficient.…”

“…Greenberg et al 9,10 note that the use of one measurement of BMI as an assessment of long-term or adult lifetime BMI results in measurement error. The BMI at the time it is assessed may be measured very accurately, but it provides at best an imperfect estimate of lifetime BMI.…”