A prospective, randomized trial was conducted to compare the effectiveness and safety of warfarin given in two regimens in prevention of venous thrombosis after total knee replacement. Adult patients scheduled for primary or revision total knee replacement were randomly assigned to receive either a "two-step" warfarin regimen beginning 10-14 days pre-operatively or, alternatively, to begin warfarin the night before surgery. Post-operatively, the dose was adjusted in both groups to achieve a target International Normalized Ratio (INR) of 2.2 and prophylaxis was continued until venography on post-operative days five through nine. Bleeding was assessed by surgical blood loss, transfusion requirements, changes in hematocrit, and clinically identified bleeding complications. The occurrence of deep vein thrombosis was nearly the same in the two treatment groups, 39% in patients randomized to the two-step regimen as compared to 38% in those beginning the night before surgery. The occurrence of proximal vein thrombosis was also similar, 5% versus 7% (p = NS). Patients in the two-step group received 1.33 +/- 1.26 transfusions compared to 0.95 +/- 1.22 in the night before group (p < 0.05) and also had a lower nadir post-operative hematocrit of 26.7 +/- 3.1 as compared to 28.5 +/- 3.2 (p < 0.0001). Major bleeding complications were associated with excessively prolonged INRs and occurred in five patients in the two-step group and two in the night before group. Patients in both groups who developed thrombosis had a significantly lower INR on post-operative days two and three compared to those without thrombosis. We conclude that a prophylactic warfarin regimen for prevention of deep vein thrombosis after total knee replacement beginning the night before surgery is more convenient and may be associated with less bleeding than a regimen beginning warfarin 10-14 days pre-operatively. Careful control of anticoagulant intensity is needed to achieve maximum effectiveness and avoidance of bleeding complications.