The purpose of this study was to evaluate long-term survival benefits of bypass surgery and angioplasty versus medical therapy in 9263 patients at Duke University Medical Center between 1984 and 1990 with coronary artery disease confirmed by cardiac catheterization to involve one, two, or three vessels. Clinical data were prospectively entered into an established cardiovascular database, and annual follow-up was 97% complete for a mean interval of 5.3 years and a maximal interval of 10 years. Outcomes were analyzed with the Coronary Artery Surgery Study "method A" to define patient groups treated by medicine (n = 2449), angioplasty (n = 2924), or bypass surgery (n = 3890). Differences among treatment groups in baseline characteristics were adjusted by Cox proportional hazard models. The anatomic severity of coronary artery stenosis best defined survival benefit from bypass surgery and angioplasty versus medical treatment. One or both interventional treatments provided better long-term survival than did medical treatment for all levels of disease severity. All patients with single-vessel disease, except those with at least 95% proximal left anterior descending stenosis, benefited from angioplasty versus bypass. All patients with three-vessel disease and those two-vessel patients with > or = 95% proximal left anterior descending stenosis benefited from bypass surgery versus angioplasty. All other patients with two-vessel disease and those with > or = 95% proximal left anterior descending stenosis only had similar survival with either interventional treatment. The absolute survival benefit was greatest for patients with severe three-vessel disease treated with bypass surgery.