Two hundred fifty consecutive patients treated for one or two vessel coronary artery disease with either balloon angioplasty or surgical bypass were monitored for 3 years in a study designed to determine the comparative long-term effectiveness of each treatment. The 125 patients having angioplasty were matched with the 125 patients having bypass, so that both groups had a similar number of patients with single or double vessel disease. The two groups did not significantly differ in age, male:female ratio, New York Heart Association class, or risk factors. The ejection fraction was 54 +/- 11 in the angioplasty group and 49 +/- 12 mmHg in the surgical patients (p = 0.0031). Angioplasty was deemed initially successful in 88% (110/125), unsuccessful in 10% (12/125), and in 2% (3/125) the lesion could not be crossed. Emergency bypass was performed in 10% (12/125). Four of the 125 angioplasty patients (3%) died within 30 days. Coronary artery bypass grafting was successfully performed on the matched set of surgical patients with 99% (124/125) discharged well. There was one (1%, 1/125) surgical death. The average hospital stay per patient was 4.8 +/- 3.1 days for angioplasty and 12.1 +/- 4.2 days for bypass grafting (p = 0.0000). Three-year postprocedure follow-up was obtained on 96% (236) of the 245 patients discharged alive. A second angioplasty was required in 18%, and 11 angioplasty patients subsequently required surgical bypass. Overall, 19% (23/121) of the angioplasty patients ultimately required bypass. Four late deaths occurred in the angioplasty group, which brought the early and late mortality rates to 7% (8/121). There were two late surgical deaths, which brought the combined surgical mortality to 2.5% (3/120), p = 0.1263. Patient evaluation reveals that 63% (76/121) of the angioplasty group are alive and in New York Heart Association class I or II 3 years after one or two angioplasty procedures. This figure compares with 92% (110/120) of surgical patients alive and in the same two New York Heart Association classes (p = 0.0000).