A total of 100 osteoarthritis patients undergoing unilateral total knee arthroplasty were randomly assigned to receive either a multimodal analgesia protocol, comprising oral celecoxib and tramadol before and after surgery and intra-articular injection of large doses of morphine, ropivacaine, adrenaline and betamethasone during surgery (trial group), or oral and intra-articular placebo (control group). All patients received patient-controlled analgesia for 48 h after surgery. Morphine consumption up to 48 h after surgery was significantly lower in the trial than in the control group. Compared with the control group, the trial group had significantly lower visual analogue scale (VAS) scores for pain at rest from 6 h to 7 days after surgery and significantly lower VAS scores during activity from 24 h to 7 days after surgery. Active straight leg raise and active 90 degrees knee flexion were achieved sooner and range of knee movement at postoperative days 1 - 15 were significantly greater in the trial group. Postoperative wound healing, infection, blood pressure, heart rate, rash, respiratory depression, urinary retention and deep vein thrombosis were similar in the two groups, but nausea and vomiting were significantly less frequent in the trial group.