Reduction of postoperative pain by injecting opioid into the knee joint is believed to support the hypothesis of peripheral opioid receptor activation in inflammation. The study design consisted of a systematic review of randomised controlled trials (RCTs). Main outcomes were pain intensity and the use of supplementary analgesics. Efficacy of intra-articular bupivacaine against placebo was used as an index of internal sensitivity. Evidence of efficacy was sought in both early (0-6 h after intra-articular injection) and late (6-24 h) periods. Thirty-six RCTs in knee surgery were found. Six had both a local anaesthetic control and placebo; four showed internal sensitivity. All four sensitive studies had at least one outcome showing efficacy of intra-articular morphine against placebo. Six studies compared intra-articular morphine with intravenous or intramuscular morphine or with intra-articular saline without a bupivacaine control. Four of the six studies showed greater efficacy for intra-articular morphine. There was no dose-response evident. No quantitative analysis of pooled data was done. We conclude that intra-articular morphine may have some effect in reducing postoperative pain intensity and consumption of analgesics. These studies had significant problems in design, data collection, statistical analysis and reporting. Trials of better methodological quality are needed for a conclusive answer that intra-articular morphine is analgesic, and that any analgesia produced is clinically useful.