This study reviews 146 consecutive patients who underwent tricuspid valve replacement (TVR) with 69 bioprostheses (porcine and bovine pericardial) and 77 mechanical ball, disc or bileaflet valves between 1967 and 1987. The mean age was 51.4 +/- 12.1 years. Preoperatively, 97% were in New York Heart Association (NYHA) functional class III or more and over 40% had undergone previous cardiac surgery. Hospital mortality was high (16.1%). Incremental risk factors for hospital death were preoperative icterus (P < 0.01), hepatomegaly (P = 0.02), NYHA functional class IV (P = 0.02) and male sex (P = 0.04) (univariate analysis). Ninety-eight percent of the hospital survivors were followed up for a mean of 92 months. Cumulative follow-up added up to 955 patient-years. There were 70 late deaths. The actuarial survival rate was 74% at 60 months and less than 25% at 14 years. Incremental risk factors for late death indicated by univariate analysis were the type of tricuspid prosthesis (Smel-off-Cutter and Kay-Shiley prostheses) (P = 0.04), the type of operative myocardial protection (normothermia and coronary perfusion) (P = 0.05) and preoperative NYHA functional class IV (P = 0.05). We conclude that TVR carries a high operative risk and poor long-term survival, both influenced by preoperative and perioperative variables. Bearing in mind the poor prognosis for TVR, we prefer a large-sized bioprosthesis, in view of its initial good durability and low risk of valve-related events. However, in patients with good life expectancy, a bileaflet mechanical prosthesis may be an acceptable alternative.