Ketamine, an N-methyl-D-aspartate receptor antagonist, is known to be analgesic and to induce psychomimetic effects. Benefits and risks of ketamine for the control of postoperative pain are not well understood. We systematically searched for randomised comparisons of ketamine with inactive controls in surgical patients, reporting on pain outcomes, opioid sparing, and adverse effects. Data were combined using a fixed effect model. Fifty-three trials (2839 patients) from 25 countries reported on a large variety of different ketamine regimens and surgical settings. Sixteen studies tested prophylactic intravenous ketamine (median dose 0.4 mg/kg, range (0.1-1.6)) in 850 adults. Weighted mean difference (WMD) for postoperative pain intensity (0-10 cm visual analogue scale) was -0.89 cm at 6 h, -0.42 at 12 h, -0.35 at 24 h and -0.27 at 48 h. Cumulative morphine consumption at 24 h was significantly decreased with ketamine (WMD -15.7 mg). There was no difference in morphine-related adverse effects. The other 37 trials tested in adults or children, prophylactic or therapeutic ketamine orally, intramuscularly, subcutaneously, intra-articulary, caudally, epidurally, transdermally, peripherally or added to a PCA device; meta-analyses were deemed inappropriate. The highest risk of hallucinations was in awake or sedated patients receiving ketamine without benzodiazepine; compared with controls, the odds ratio (OR) was 2.32 (95%CI, 1.09-4.92), number-needed-to-harm (NNH) 21. In patients undergoing general anaesthesia, the incidence of hallucinations was low and independent of benzodiazepine premedication; OR 1.49 (95%CI 0.18-12.6), NNH 286. Despite many published randomised trials, the role of ketamine, as a component of perioperative analgesia, remains unclear.