The role of dexamethasone in the prevention of postoperative nausea and vomiting (PONV) is unclear. We reviewed efficacy and safety data of dexamethasone for prevention of PONV. A systematic search (MEDLINE, EMBASE, Cochrane Library, hand searching, bibliographies, all languages, up to April 1999) was done for full reports of randomized comparisons of dexamethasone with other antiemetics or placebo in surgical patients. Relevant end points were prevention of early PONV (0 to 6 h postoperatively), late PONV (0 to 24 h), and adverse effects. Data from 1,946 patients from 17 trials were analyzed: 598 received dexamethasone; 582 received ondansetron, granisetron, droperidol, metoclopramide, or perphenazine; 423 received a placebo; and 343 received a combination of dexamethasone with ondansetron or granisetron. With placebo, the incidence of early and late PONV was 35% and 50%, respectively. Sixteen different regimens of dexamethasone were tested, most frequently, 8 or 10 mg IV in adults, and 1 or 1.5 mg/kg IV in children. With these doses, the number needed to treat to prevent early and late vomiting compared with placebo in adults and children was 7.1 (95% CI 4.5 to 18), and 3.8 (2.9 to 5), respectively. In adults, the number needed to treat to prevent late nausea was 4.3 (2.3 to 26). The combination of dexamethasone with ondansetron or granisetron further decreased the risk of PONV; the number needed to treat to prevent late nausea and vomiting with the combined regimen compared with the 5-HT3 receptor antagonists alone was 7.7 (4.8 to 19) and 7.8 (4.1 to 66), respectively. There was a lack of data from comparisons with other antiemetics for sensible conclusions. There were no reports on dexamethasone-related adverse effects. When there is a high risk of postoperative nausea and vomiting, a single prophylactic dose of dexamethasone is antiemetic compared with placebo, without evidence of any clinically relevant toxicity in otherwise healthy patients. Late efficacy seems to be most pronounced. It is very likely that the best prophylaxis of postoperative nausea and vomiting currently available is achieved by combining dexamethasone with a 5-HT3 receptor antagonist. Optimal doses of this combination need to be identified.