Shape of the BMI-Mortality Association by Cause of Death, Using Generalized Additive Models: NHIS 1986-2002

The relationship between body mass index (BMI) and mortality is curvilinear; with very high and low BMIs associated with increased odds of dying. The precise shape of the association for the majority of adults in the center of the BMI distribution, however, has not been established. This paper employs generalized additive models (GAMs) to describe the relationship between BMI and cause-specific mortality. GAMs allow modeling the shape of the relationship without imposing a priori constraints on the functional form of the predictors, which helps avoid model misspecification and results in an accurate but still parsimonious description of the predictors’ effects. We examine the association between BMI and all-cause and cause-specific mortality (cardiovascular, respiratory, cancer, and diabetes). Analyses are based on the National Health Interview Survey 1986-2000 linked to the National Death Index for non-Hispanic white adults age 50-80. Results suggest that the BMI-mortality association is more V-shaped than U-shaped for both men and women. The association differs substantially by cause of death: from essentially flat for cancer, V-shaped for cardiovascular, inverted J-shape for respiratory, to monotonically increasing for diabetes.