When we're looking at the obesity mortality literature, it is important to try to figure out how the researchers approached their cohort (study population). Who you're using as the referent and what exclusions you choose have a dramatic impact on the results and have, therefore, been rife with controversy.

The biggest problem with the study using NHANES data in particular stems from the very crude control for smoking. In half of the data, smoking information was not available at all. In the other half of the data, the only choice was "never-smoker" or "have you smoked," with no distinction made between past smoking, current smoking and amount of smoking. This lack of data may somewhat explain the more dramatic U-shaped mortality curve in NHANES.

Differences by race or ethnicity also pose an interesting question. Much of the research on obesity has been performed in populations where minorities are quite underrepresented. For example, minorities make up less than 2 percent of the population in the Nurses Health Study. More data is needed to confirm the relationships in different ethnic and racial minorities.