To the Editor: In their study of deaths associated
with underweight, overweight, and obesity, Dr Flegal and colleagues1 conclude that excess mortality due to obesity and
overweight is much lower than previously reported. We believe that their analysis
is flawed and misleading.
A major challenge in such studies is that low weight is often due to
underlying chronic disease, which may exist for many years before death. Thus,
lean persons are a mix of smokers, healthy active persons, and those with
chronic illness (due to the direct effects of disease on weight and sometimes
purposeful weight loss motivated by diagnosis of a serious illness). Their
analysis does not successfully disentangle this diverse group. In the main
analyses, the study apparently did not exclude persons with known chronic
disease at baseline. The authors did conduct some analyses to address reverse
causation, such as excluding early follow-up and those persons who had lost
weight before baseline (data not shown), and they examined never-smokers separately;
however, these approaches should be used simultaneously, not one at a time.2 Unfortunately, the size of the National Health and
Nutrition Examination Survey (NHANES) data set is too small to yield precise
estimates from these simultaneous analyses. Even just restricting the analysis
to never-smokers led to the wide confidence intervals seen in Table 2 in the
study by Flegal et al.