The burden of obesity revisited.

A previous mindblog post has noted the obesity paradox, generated by data suggesting that fat people may live longer. Stokes and Preston,however, note an important distinction that may evaporate the apparent paradox:

Analyses of the relation between obesity and mortality typically evaluate risk with respect to weight recorded at a single point in time. As a consequence, there is generally no distinction made between nonobese individuals who were never obese and nonobese individuals who were formerly obese and lost weight. We introduce additional data on an individual’s maximum attained weight and investigate four models that represent different combinations of weight at survey and maximum weight. We use data from the 1988–2010 National Health and Nutrition Examination Survey, linked to death records through 2011, to estimate parameters of these models. We find that the most successful models use data on maximum weight, and the worst-performing model uses only data on weight at survey. We show that the disparity in predictive power between these models is related to exceptionally high mortality among those who have lost weight, with the normal-weight category being particularly susceptible to distortions arising from weight loss. These distortions make overweight and obesity appear less harmful by obscuring the benefits of remaining never obese. Because most previous studies are based on body mass index at survey, it is likely that the effects of excess weight on US mortality have been consistently underestimated.