Sixty-two patients with at least one prosthetic heart valve each underwent a total of 70 reoperations. Their cases were analysed with special regard to the risk of perioperative mortality and to the factors affecting that risk. The age of the patients studied was 47 +/- 14.6 years (mean +/- 1 SD), and 4.75 +/- 4 years had elapsed since their previous operation. A total of 85 prostheses were implanted at the time of initial surgery. Forty-nine of those were mitral, 28 were aortic, and 8 were tricuspid prostheses. There were 48 mechanical valves, 24 bioprostheses and 13 Carpentier rings (9 of the latter being mitral and 4 tricuspid). At reoperation (n = 70) 95 valves required treatment. Eighteen patients died following their first reoperation (18/62, 29%), and two more died after a second reoperation (2/8, 25%). Actuarial probability of hospital survivors being alive 8-9 years following reoperation was 47 +/- 15%. It was found by multiple regression analysis that the duration of cardiopulmonary bypass (p less than 0.01), surgical priority (p less than 0.05) and the patient's age (p less than 0.05) were the only independent predictors of hospital death. A review of the literature and our own results suggest that surgical problems encountered with reoperation have unfavourably affected surgical results in the past. Objective risk factors are currently more related to the operative outcome. For these to be neutralised, patients with dysfunctioning prosthetic heart valves should be operated at an early stage before they reach a critical one. Only patients with stable, mild dysfunction unlikely to worsen acutely may probably be followed-up safely.