Controversy still exists about the proper selection of patients with coronary artery disease and left ventricular dysfunction for coronary bypass surgery. To examine this issue, we studied 710 patients with significant coronary artery disease and left ventricular dysfunction (ejection fraction less than or equal to 40%). Of 301 patients treated surgically, 232 had bypass grafts; 17, left ventricular surgery; and 52, both procedures. At 3 years after treatment, unadjusted survival was 84% for surgical patients and 64% for medical patients. At baseline, medical patients had more left ventricular dysfunction than surgical patients, but surgical patients had more coronary artery disease and angina than medical patients. In Cox survival models, two invasive factors (ejection fraction and extent of coronary artery disease) and three noninvasive indexes (assessment of myocardial infarction, angina, and conduction disturbances) were the five best predictors of survival (p less than 0.001). After adjustment for these factors between the two treatment groups, overall surgical survival at 3 years after treatment was 86%, and medical survival was 68%. Long-term surgical survival benefits appeared greatest in patients with the most severe left ventricular dysfunction, most extensive coronary artery disease, and most severe anginal symptoms. We conclude that surgery provides significant survival benefits for coronary disease patients with left ventricular dysfunction; in general, the sicker the patient, the greater the benefit.