Without having read this study in detail, other possible things tend to muck up such studies. Suppose you define mortality as "chance of dying within a year"

Without having read this study in detail, other possible things tend to muck up such studies. Suppose you define mortality as "chance of dying within a year" 1) People who start dying (eg by developing cancer) might take more than a year, and generally they lose weight when they do so. So even if obesity caused you to be more likely to get cancer and die, it wouldn't show up in this sort of data. 2) People who weigh less may be otherwise unhealthy for other reasons -- this doesn't imply that gaining weight would benefit them. In this sense weight is an effect of some other cause. 3) We may not be measuring "overweight" correctly -- scale weight is different from BMI which is different from body fat% which is different from body density which is different from body volume which is different from other things closely associated with "weight" (and specifically the stuff you gain/lose by eating)