Forty-two patients scheduled for total knee arthroplasty (n = 20) or hip arthroplasty (n = 22) were randomly allocated to receive either continuous epidural bupivacaine/morphine for 48 h postoperatively plus oral piroxicam, or general anaesthesia followed by a conventional intramuscular opioid and acetaminophen regimen. Patients undergoing knee- or hip arthroplasty treated with epidural analgesia had significantly lower pain scores during mobilization under the 48 h epidural infusion compared with patients receiving conventional treatment, while no important differences were observed after cessation of the epidural regimen. However, the achieved pain relief had no impact on postoperative convalescence parameters, such as ambulation, patient activity including need for nursing care, fatigue or hospital stay. Late postoperative pain, fatigue and conservative attitudes and routines in the postoperative care, were the most important reasons limiting mobilization and activity. We conclude that effective early (48 h) postoperative pain relief with balanced analgesia does not per se lead to important improvements in convalescence and hospital stay.