Background Patients with mechanical heart valve prostheses may experience valve thrombosis and subsequent systemic embolism for which they are treated with oral anticoagulant therapy. It is essential to know reliable estimates of the risks and benefits of this therapy in order to answer a number of clinical questions rationally. We sought to obtain more precise estimates of the risks and benefits by combining the data from individual studies by using meta-analysis. Methods and Results We searched for studies in which the incidences were reported of embolic or bleeding complications in patients with mechanical heart valve prostheses. They were collected from the Medline and Current Contents database and by cross-references between 1970 and 1992. Since most studies vary greatly in many respects, we used a number of inclusion criteria, thus selecting comparable studies of acceptable quality only. The influence of antithrombotic therapy, valve position, and valve type was analyzed by univariate and by multivariate analysis with Poisson regression techniques. Forty-six studies were found, including 13 088 patients studied for 53 647 patient-years. We found an incidence of major embolism in the absence of antithrombotic therapy of 4 per 100 patient-years. With antiplatelet therapy this risk was 2.2 per 100 patient-years, and with cumarin therapy it was reduced to 1 per 100 patient-years. This risk varied with the type and the site of the prosthesis. A prosthesis in mitral position increased the risk almost twice as compared with the aortic position. Tilting disc valves and bileaflet valves showed a lower incidence of major embolism than caged ball valves. An incidence of major bleeding was found in patients treated with cumarin derivatives of 1.4 per 100 patient-years. The incidence of bleeding became significantly higher with the addition of antiplatelet therapy, although this did not decrease the risk of thromboembolism any further. Conclusions These data provide a reference for future studies and give adequate risk estimates for clinical decision making.
