Long-term performance of Starr-Edwards silastic ball (SESB, n = 168) and St Jude Medical bi-leaflet (SJMB, n = 93) valves in patients who were alive 30 days after implantation (1980-86) for aortic stenosis was compared. Mean follow-up was 3.0 years (0.1-7.9 years). The SESB and SJMB groups differed as regards female gender (18% vs 47%, P less than 0.0001), NYHA classes III-IV (59% vs 72%, P less than 0.05), coronary artery disease (CAD, 32% vs 62%, P less than 0.01) in patients with coronary arteriography (n = 82 and n = 55, respectively), and prosthetic annulus diameter (26 +/- 1 vs 23 +/- 2 mm, P less than 0.0001). Five-year survival +/- SE in SESB vs SJMB patients was: total population, 89 +/- 3% vs 80 +/- 6% (NS); coronary arteriography population, no CAD, 90 +/- 4% vs 100% (NS), and with CAD, 71 +/- 11% vs 60 +/- 13% (NS; P = 0.01 for CAD). Five-year event-free survival +/- SE in SESB vs SJMB patients was 95 +/- 2% vs 97 +/- 2% (NS) for thromboembolism, 95 +/- 2% vs 89 +/- 4% (NS) for coumadin-related haemorrhage, 98 +/- 1% vs 99 +/- 1% (NS) for endocarditis, 98 +/- 1% vs 94 +/- 5% (NS) for paravalvular leak, 88 +/- 3% vs 79 +/- 6% (NS) for all valve-related complications, and 98 +/- 1% vs 95 +/- 4% (NS) for prosthesis replacement. Thrombotic occlusion or structural failure were not observed. No patients without CAD experienced thromboembolic events. Cox regression analyses (in both total population and coronary arteriography population) of survival as well as the various complications revealed that the type of prosthesis did not have predictive influence. CAD was an independent risk factor for thromboembolism, haemorrhage, and all valve-related complications. Previous systemic hypertension was independently predictive of haemorrhage. The SESB and SJMB prostheses showed comparable and acceptable long-term performance. Only patient-related variables, notably CAD, influenced late results. The proven durability and relatively low price of the SESB valves together with the excellent haemodynamic performance of even small-sized SJMB valves should be considered in the light of the present results.