Exercise testing in asymptomatic persons has been criticized for failing to accurately predict those at risk for coronary heart disease (CHD). Previous studies on asymptomatic subjects, however, may not have been large enough or long enough to provide reliable outcome measures. This study examines the ability of a maximal exercise test to predict death from CHD and death from any cause in a population of asymptomatic men. This is a prospective longitudinal study performed between 1970 and 1989, with an average follow-up of 8.4 years. The subjects are 25,927 healthy men, 20 to 82 years of age at baseline (mean 42.9 years) who were free of cardiovascular disease and who were evaluated in a preventive medicine clinic. The main outcome measures are CHD mortality and all-cause mortality. During follow-up there were 612 deaths from all causes and 158 deaths from CHD. The sensitivity of an abnormal exercise test to predict coronary death was 61%. The age-adjusted relative risk of an abnormal exercise test for CHD death was 21 (6.9 to 63.3) in those with no risk factors, 27 (10.7 to 68.8) in those with 1 risk factor, 54 (21.5 to 133.7) in those with 2 risk factors, and 80 (30.0 to 212. 5) in those with >/=3 factors. A maximal exercise test performed in asymptomatic men free of cardiovascular disease does appear to be a worthwhile tool in predicting future risk of CHD death. An abnormal exercise test is a more powerful predictor of risk in those with than without conventional risk factors.