Objective: To investigate the nature of the relation between body weight and all-cause mortality.
Design: Prospective cohort study, following up men from 1962 or 1966 (1962/1966) through 1988.
Setting/participants: Harvard University alumni with a mean age of 46.6 years in 1962/1966 and without self-reported, physician-diagnosed coronary heart disease, stroke, or cancer, who completed questionnaires on weight, height, cigarette smoking habit, and physical activity (n = 19,297). We calculated body mass index (weight in kilograms divided by the square of height in meters) using self-reported measures.
Main outcome measure: All-cause mortality (4370 deaths).
Results: In multivariate analysis adjusting for age, cigarette smoking habit, and physical activity, we found a J-shaped relation between body mass index and mortality. Relative risks of dying for men with a body mass index of less than 22.5, 22.5 to less than 23.5, 23.5 to less than 24.5, 24.5 to less than 26.0, and 26.0 or greater were 1.00 (referent), 0.99 (95% confidence interval, 0.89 to 1.20), 0.95 (0.87 to 1.05), 1.01 (0.91 to 1.10), and 1.18 (1.08 to 1.28), respectively (P for linear trend = .0008). Among current smokers, the relation between body mass index and mortality was U-shaped, with lowest risk of death at a body mass index of 23.5 to less than 24.5. During early follow-up (1962/1966 through 1974), we also observed a U-shaped curve, this time with lowest mortality risk at a body mass index of 24.5 to less than 26.0. To minimize confounding by cigarette smoking and bias from antecedent disease and early mortality, we conducted analysis only among never smokers and omitted the first 5 years of follow-up (510 deaths). The corresponding relative risks from this analysis, adjusted for age and physical activity, were 1.00, 1.23 (95% confidence interval, 0.90 to 1.67), 1.06 (0.80 to 1.42), 1.27 (0.96 to 1.68), and 1.67 (1.29 to 2.17), respectively (P for linear trend = .0001).
Conclusions: In these prospective data, body weight and mortality were directly related. After accounting for confounding by cigarette smoking and bias resulting from illness-related weight loss or inappropriate control for the biologic effects of obesity, we found no evidence of excess mortality among lean men. Indeed, lowest mortality was observed among men weighing, on average, 20% below the US average for men of comparable age and height.