Thromboembolism, valve thrombosis and haemorrhagic events have been compared in 356 Starr-Edwards (SE) 1260, 113 Björk-Shiley (BS), and 178 aortic bioprostheses operated upon between 1968 and 1982, and reviewed by the same group with less than 2% of patients lost of follow-up. Expressed in actuarial rate at 7 years the percentage of patients free of thromboembolism event is 87% for SE, 86% for BS, 94% for bioprostheses; the linearized rate is 2.9%/patient/year for SE, 2.2 for BS, 1.9 for bioprostheses (NS). Valve thrombosis was not observed in bioprostheses; 97.9% of patients with mechanical valves were free of valve thrombosis at 6.5 years. Haemorrhagic risk was lower with bioprostheses than with mechanical valves 0.2% patient/year vs 2.33 (P less than 0.005). The most important factor influencing thromboembolic and haemorrhagic risks is the quality of anticoagulant therapy. Other contributing factors are the date of the operation and associated mitral disease.