The 3 year cumulative survival rate of 1,492 patients with left main coronary artery disease (50 percent or greater stenosis of luminal diameter) enrolled in the Collaborative Study in Coronary Artery Surgery (CASS) was 91 percent for the surgical group and 69 percent for patients treated medically (p less than 0.0001). Mortality was significantly greater in patients with impaired left ventricular function. The difference between medical and surgical therapy was significant for patients who had normal, moderately abnormal and severely impaired left ventricular function and for patients with stenosis of the left main coronary artery of 50 to 59, 60 to 69, 70 to 79 and 80 percent or greater. Aortocoronary bypass surgery did not significantly improve survival in patient subgroups who had (1) a nonstenotic dominant right or balanced coronary circulation, (2) a stenotic dominant right coronary artery and normal left ventricular function, and (3) left main coronary stenosis of 50 to 59 percent and normal or mildly abnormal left ventricular function. The Cox proportional hazards model was used to select baseline variables that were independent predictors of long-time mortality. The model selected left ventricular score, age, congestive heart failure score, hypertension, percent left main coronary arterial stenosis and coronary arterial dominance as the baseline variables most predictive of long-term survival. A clinical and angiographic prognostic risk index developed from these six baseline variables showed significantly improved survival for the surgical cohort in each of four risk categories. In the best and worst risk category, the 3 year survival rate was 97 and 82 percent, respectively, for the surgical group and 85 and 34 percent, respectively, for the medical group (p less than or equal to 0.0002). The data from this observational study show that coronary bypass surgery prolongs life in most patients with left main coronary artery disease, particularly those who have severe narrowing of the left main coronary artery or impaired left ventricular function. The results permit a better understanding of the natural history of left main coronary artery disease and permit a more accurate estimate of long-term survival for individual patients through the use of a clinical-angiographic risk index.