We compared time to first new myocardial infarction during a 6-year follow-up in patients in the registry of the Coronary Artery Surgery Study who had three-vessel coronary artery disease and Canadian Cardiovascular Society Class III-IV angina pectoris. There were 679 medically treated patients and 1921 surgically treated patients in this nonrandomized comparison. A broad definition of myocardial infarction incorporating electrocardiographic and clinical criteria was used to include as many new infarctions as possible. Patients were stratified by left ventricular wall motion score and number of proximal coronary artery stenoses; after adjustment for these variables, 86% of surgical and 73% of medical patients were free of new myocardial infarction at 6 years (p less than 0.0001). This advantage of surgical treatment was observed in subgroups of patients with at least one proximal 70% (or greater) stenosis in the left anterior descending coronary artery and moderate or severe impairment of left ventricular function, as well as those patients with two proximal coronary artery narrowings. In a multivariate (Cox) analysis of preoperative clinical, hemodynamic, and angiographic factors, early operation was the strongest predictor of freedom from new myocardial infarction.