Objective: To estimate the relationship between body mass index (BMI, kg/m2) and all-cause mortality with information from the published scientific literature.
Design: Meta-analysis using a hierarchical, mixed model. The analysis included random effects for information sources and fixed effects for factors that may modify the BMI-mortality relationship such as smoking, control for disease, and country of origin, which allowed combining information from diverse studies.
Main outcome measures: Predicted probability of death over a given duration of follow-up plotted by BMI for sex-age cohorts of white race.
Results: An extensive search identified nineteen prospective cohort studies that met inclusion criteria. A U-shaped relationship between BMI and mortality was demonstrated for 50-year-old men followed for 30 years. Mortality risk increased with low and high BMI (< 23 or > 28) in groups of non-smokers without evidence of disease upon study entry. Limited information from studies of women indicated that, with 10 year follow-up, there was little relationship between BMI and mortality for (1) non-smokers and for (2) mixtures of smokers and non-smokers.
Conclusion: This quantitative analysis of existing studies revealed increased mortality at moderately low BMI for white men comparable to that observed at extreme overweight, which does not appear to be due to smoking or existing disease. Attention to the health risks of underweight is needed, and body weight recommendations for optimum longevity need to be considered in light of these risks.