Shape of BMI–Mortality Risk Associations: Reverse Causality and Heterogeneity in a Representative Cohort of US Adults

OBJECTIVE This study examines BMI-mortality associations and evaluates strategies intended to limit reverse causality. Heterogeneity in BMI-mortality risk associations across subgroups and causes of death is explored. METHODS A cohort of 654,382 adults from the US National Health Interview Survey was constructed. Associations between unit BMI levels and mortality were estimated using Cox proportional hazards models, including and excluding the first 5 years of follow-up, with and without controls for smoking or preexisting conditions, and including and excluding ever-smokers and individuals with preexisting conditions. Stratified analyses by individual characteristics were performed. RESULTS Addressing reverse causality led to reduced risk of mortality among those with low BMI levels ( 40 kg/m2 ) and lowered the estimated nadir risk from 27 kg/m2 to 23 kg/m2 . After excluding ever-smokers and individuals with preexisting conditions, limiting the analysis to >5 years of follow-up produced no substantive changes. Heterogeneous results were observed across individual characteristics, particularly age and causes of death. CONCLUSIONS The exclusion of smokers and individuals with preexisting conditions alters the BMI-mortality risk association and results in a somewhat lower range of BMI with minimum mortality risk.