The obesity paradox in the US population.

Background: Recently a number of studies have found a lower risk of dying for obese individuals than for normal-weight individuals. The explanation for these paradoxical findings has not yet been identified. Objective: The objective was to assess whether this paradoxical pattern exists in the US population and whether it can be explained by reverse causation. Design: Survival analyses were used to calculate the RR of all-cause mortality for obesity by using data from 35,673 participants in NHANES I (1971-1975), NHANES II (1976-1980), and NHANES III (1988-1994), which reported 7087 deaths during 3 different 15-y follow-up periods. Results: With normal weight as a referent, a lower relative mortality risk of obesity was found only in NHANES III and only among men with a wide variety of preexisting serious illnesses. For this subgroup, the relative mortality risks in NHANES I, II, and III were 2.22 (95% CI: 1.45, 3.40), 0.89 (95% CI: 0.70, 1.15), and 0.65 (95% CI: 0.47, 0.91), respectively. Whereas the mortality rate among seriously ill normal-weight men did not change significantly between NHANES I and III, it did decrease significantly among seriously ill obese men, suggesting that reverse causation was not responsible for the lower relative mortality risk among seriously ill obese men in NHANES III. Conclusions: Only obese NHANES male participants with a wide variety of serious illnesses experienced lower mortality risk than their normal-weight counterparts and only in NHANES III. Reverse causation seems unlikely to have played a role. These conclusions require confirmation.