Reports among patients with coronary heart disease regarding the association between body mass index (BMI) and long-term mortality are inconsistent, ranging among linear, U-shaped, or inverse (the "obesity paradox") associations. BMI and mortality data were available for 12,466 men with chronic coronary heart disease. BMI was classified as <20 (lean), 20.0 to 22.99, 23.0 to 24.99 (reference), 25.0 to 26.99, 27.0 to 29.99, and >or=30 kg/m(2) (obese). Age-adjusted (direct methods) mortality was investigated within risk factor categories. Adjusted hazard ratios compared with the reference group were estimated using a Cox proportional-hazards model. Two thirds of the patients had BMIs >or=25 kg/m(2). A number of risk factors were progressively more frequent with increasing BMI (age, diabetes, past smoking, and metabolic components). Over a median follow-up period of 12 years, adjusted mortality rates per 1,000 patient-years followed a U-shaped association with BMI. The highest risk was noted in 148 lean (hazard ratio 1.41, 95% confidence interval 1.08 to 1.85) and 1,788 obese (hazard ratio 1.28, 95% confidence interval 1.15 to 1.42) patients. Mortality hazard in patients with BMIs of 20.0 to 29.99 kg/m(2) (84% of patients) did not significantly differ from the reference group (lowest risk). Risk factor presence was associated with higher mortality in every BMI category. Lean patients had a particularly poor prognosis in the presence of past myocardial infarction, smoking, or renal insufficiency. A U-shaped association was found in most subgroups examined. In conclusion, BMI >or=25 kg/m(2) is common in patients with coronary heart disease. A U-shaped association, with highest risk among lean and obese patients, is persistent regardless of risk factor presence. Further data are required to support the need of aggressive weight reduction in patients with BMIs <30 kg/m(2).