Eighty consecutive patients undergoing unilateral total knee arthroplasty received postoperative analgesia consisting of a continuous epidural infusion of fentanyl and bupivacaine. Nineteen patients (24%) were unable to complete the three-day course of epidural infusion: two thirds for technical reasons and one third because of adverse effects. The remaining 61 patients (76%) successfully completed the 72-hour protocol. The benefits in this group included a shorter hospital stay (8.3 versus 9.5 days, p < 0.01) and earlier return of flexion (p < 0.05) as compared with the group that had the epidural catheter removed prematurely. When compared with previous experiences with continuous epidural analgesia using Duramorph and bupivacaine for the first three days after TKA in 56 patients, the current group of 80 patients had shorter average hospitalization (8.6 versus 9.7 days, p < 0.01) and earlier return of flexion (p < 0.001). Patient acceptance was excellent with either agent. Confusion and pruritis were significantly less common with fentanyl, but the incidence of hypotension was increased. Nausea was problematic in both groups. The incidence of respiratory depression was 5% for either opiate. The degree of hypoventilation and treatment required for it were less severe with fentanyl, however.