Body Weight, Weight Change, and Mortality Risk

Laura Blue, Princeton University

This study estimates the relationship between body mass index (BMI) and mortality among adults aged 50-79. It further estimates the relationship between changes in BMI and mortality, and the joint association of BMI and BMI change with mortality. Data come from five longitudinal surveys, each based in a different high- or middle-income country: Australia, Costa Rica, England, Taiwan, and the USA. All five surveys measure BMI over time, rather than relying on participants’ self-reports. This study uses P-splines (penalized B-splines) to treat predictors as continuous variables in logistic regression analyses without strong functional-form assumptions. In all five study populations, both low and high BMIs are associated with mortality, as is BMI loss. Adjustment for BMI change rotates the BMI–mortality curve, suggesting that people with stable weight experience lower mortality at low BMIs and higher mortality at high BMIs than is observed among the population at large.