The relation between measures of adiposity and mortality has been a controversial topic. The authors examined prospectively the relation between several measures of adiposity and risks of overall and cause-specific mortality in 39,756 US men aged 40-75 years. During 10 years of follow-up (1986-1996), 1,972 deaths (747 from cancer, 423 from cardiovascular disease, and 802 from other causes) were documented. An elevated risk of death among the leanest men was partly accounted for by excess mortality during early follow-up and high mortality among those with a history of recent weight loss. After exclusion of men with substantial recent weight loss and deaths occurring during the first 4 years of follow-up, overall and cardiovascular disease mortality among men aged <65 years increased linearly with greater body mass index (BMI) (weight (kg)/height (m)2); multivariate relative risks for overall mortality were 1.0 (referent) for a BMI of <23, 1.21 for a BMI of 23-24.9, 1.19 for a BMI of 25-26.9, 1.39 for a BMI of 27-29.9, and 1.97 for a BMI of >30 (test for trend: p< 0.001). Among men aged > or =65 years, there were no significant relations between BMI and overall, cardiovascular disease, or cancer mortality risk. However, waist circumference strongly predicted risk of death from cardiovascular disease among the older men. These findings indicated that the relation between body fat and mortality was influenced by reverse causation and varied by age.