After cardiac valve replacement patients were blindly randomized into two groups, both receiving aspirin (330 mg) and dipyridamole (75 mg) twice daily and the oral anticoagulant acenocoumarol (Sintrom). An international normalized ratio of 2.0 to 2.99 was assigned to group A and 3.0 to 4.5 to group B; both groups were subsequently analyzed for thromboembolic and hemorrhagic complications. Final evaluation included 51 and 48 patients, respectively. The follow-up was 626 months for group A (12.3 months/patient) and 486 months for group B (10.1 months/patient). The frequency of thromboembolism was equal in both groups: one transient ischemic attack in group A (a rate of 1.92/100 patient-years) and two transient ischemic attacks in group B (a rate of 4.94/100 patient-years). There was, however, a statistical difference in bleeding complications between the two groups (p less than 0.02). Two patients bled in group A, a rate of 3.9% (3.8/100 patient-years), which represents an incidence of one episode each 25.6 years of treatment; 10 patients bled in group B, a rate of 20.8% (24.7/100 patient-years) representing an incidence of one episode each 4 years of treatment. We conclude that an international normalized ratio of 2 to 3 is safer than a ratio of 3 to 4.5 and confers good protection from thromboembolism when oral anticoagulant therapy is used conjointly with platelet function-inhibiting drugs in patients with mechanical substitute heart valves.