To evaluate the potential impact of patient selection for coronary artery bypass graft surgery on long-term survival, the outcomes of 5,809 consecutive patients with symptomatic coronary disease documented by angiography at Duke University Medical Center were examined. Over the entire study period (1969-1984), surgical therapy was associated with improved survival compared with medical therapy whether or not adjustment was made for imbalances in baseline prognostic factors. When patients were categorized according to coronary anatomy and left ventricular function, patients with multivessel disease and poor left ventricular function had a greater long-term survival benefit with surgery than did patients with less coronary artery disease and better left ventricular function. When 5-year survival rates were examined as a function of operative risk, a direct relation was found between estimated operative risk and the medical-surgical survival difference. For patients with an operative risk of 1%, the expected 5-year mortality with surgical therapy was 3% versus 8% with medical therapy (an absolute survival difference at 5 years of 5%). In comparison, for patients with an operative risk of 5%, the expected 5-year mortality with surgery was 10% versus 23% with medical therapy (an absolute survival difference at 5 years of 13%). Over 50% of patients with significant coronary artery disease undergoing cardiac catheterization have an estimated operative mortality risk under 2.5%. These patients would be expected to have a small survival advantage treated surgically. As operative mortality rates are subjected to increasing public scrutiny, selection of low-risk patients will reduce the overall benefit of the operation to the population.(ABSTRACT TRUNCATED AT 250 WORDS)