BACKGROUND: The relative efficacy of antiemetics for the treatment of postoperative nausea and vomiting (PONV) is poorly understood. METHODS: Systematic search (MEDLINE, Embase, Cochrane Library, bibliographies, any language, to 8.2000) for randomised comparisons of antiemetics with any comparator for the treatment of established PONV. Dichotomous data on prevention of further nausea and vomiting, and on side effects were combined using a fixed effect model. RESULTS: In seven trials (1,267 patients), 11 different antiemetics were tested without placebos; these data were not further analysed. Eighteen trials (3,809) had placebo controls. Dolasetron 12.5-100 mg, granisetron 0.1-3 mg, tropisetron 0.5-5 mg, and ondansetron 1-8 mg prevented further vomiting with little evidence of dose-responsiveness; with all regimens, absolute risk reductions compared with placebo were 20%-30%. The anti-nausea effect was less pronounced. Headache was dose-dependent. Results on propofol were contradictory. The NK1 antagonist GR205171, isopropyl alcohol vapor, metoclopramide, domperidone, and midazolam were tested in one trial each with a limited number of patients. CONCLUSIONS: Of 100 vomiting surgical patients receiving a 5-HT3 receptor antagonist, 20 to 30 will stop vomiting who would not have done so had they received a placebo; less will profit from the anti-nausea effect. There is a lack of evidence for a clinically relevant dose-response; minimal effective doses may be used. There is a discrepancy between the plethora of trials on prevention of PONV and the paucity of trials on treatment of established symptoms. Valid data on the therapeutic efficacy of classic antiemetics, which have been used for decades, are needed.