Thirty-six patients scheduled for knee arthroplasty were randomized to general or epidural anesthesia that was prolonged into the postoperative period. All the patients wore graded compression stockings until full ambulation. No other thrombo-prophylactic treatment was given. In diagnosing deep venous thrombosis, bilateral ascending venography was performed 9-11 days after surgery. The incidence of thrombosis was 2/13 in the epidural group versus 10/16 in the general anesthesia group (P less than 0.05).