To determine the association of duration of warfarin anticoagulant therapy after hospitalization for venous thromboembolism with the probability of recurrent thromboembolism and with the risks of hemorrhage, medical records of 2,422 patients hospitalized in 1970 through 1980 with pulmonary embolism, thrombophlebitis, or both were reviewed. Multivariate life-table analyses were performed for 370 patients who had positive results of venography or pulmonary angiography, or who had lung scanning evidence of a "high probability" of pulmonary embolism and no history of the disease. For these patients, warfarin therapy for more than six weeks was not associated with a lower risk of recurrent thromboembolism when compared with warfarin therapy for one through six weeks (risk of recurrence for seven to 26 weeks of treatment 0.8; 95 percent confidence limits 0.3 and 2.5; risk of recurrence for more than 26 weeks of treatment 1.1; 95 percent confidence limits 0.4 and 3.1). The longer the warfarin therapy, the higher the risk of medically important complications from therapy. From one week through five years, the probability of major hemorrhage increased almost linearly: 10 percent for 12 weeks, 18 percent at one year, 26 percent at two years, and 41 percent at five years. This study suggests that intensive, long-term warfarin anticoagulation, in patients with a first episode of venous thromboembolism and no predisposing condition, is associated with more toxicity than efficacy and should be abandoned.