In a randomized, prospective trial of 100 patients, we have studied the safety and efficacy of warfarin sodium in comparison with that of dextran 40 in the prevention of venous thrombosis in patients at high risk for deep vein thrombosis after elective total hip or knee replacement. Warfarin was given in a new two-step regimen designed to avoid bleeding complications while still preventing venous thrombosis. A low dose of warfarin was started ten to 14 days preoperatively, and the prothrombin time was regulated to between 1.5 and 3 seconds longer than control at the time of surgery; immediately after surgery, the dose was increased to prolong the prothrombin time to 1.5 times control. The overall incidence of venous thrombosis as documented by venography was less in the 53 patients treated with warfarin than in the 37 treated with dextran (21% v 51%), as was the incidence of thrombi in the femoral or popliteal veins (2% v 16%). Objective measures of blood loss showed no difference between patients treated with warfarin or dextran, and excessive postoperative bleeding was infrequent and similar in both treatment groups. This study demonstrates that two-step warfarin therapy provides highly effective prophylaxis of postoperative venous thrombosis after elective hip or knee prosthetic surgery without excessive risk of perioperative bleeding.