Shivering is a frequent complication in the postoperative period. The relative efficacy of pharmacological interventions to prevent this phenomenon is not well understood. We performed a systematic search for full reports of randomized comparisons of prophylactic, parenteral, single-dose antishivering interventions with inactive control (placebo or no treatment). Variable doses were converted to fixed doses. Dichotomous data on the absence of shivering were analyzed by using relative benefit (RB) and number needed to treat (NNT) with 95% confidence intervals (CI). Data from 27 trials (1348 adults received an antishivering intervention; 931 were controls) were analyzed. The average incidence of shivering in controls was extremely frequent (52%). Clonidine 65-300 microg (1078 patients), meperidine 12.5-35 mg (250 patients), tramadol 35-220 mg (250 patients), and nefopam 6.5-11 mg (204 patients) were tested in at least 3 trials each. All were more effective than control. For clonidine, meperidine, and nefopam, there was some weak evidence of dose responsiveness. For small-dose clonidine (65-110 microg), the RB compared with control was 1.32 (95% CI, 1.16-1.51); for medium-dose clonidine (140-150 microg), the RB was 1.83 (95% CI, 1.47-2.27); and for large-dose clonidine (220-300 microg), the RB was 1.52 (95% CI, 1.30-1.78). For all clonidine regimens combined, the RB was 1.58 (95% CI, 1.43-1.74), with an NNT of 3.7. For all meperidine regimens combined, the RB was 1.67 (95% CI, 1.37-2.03), with an NNT of 3. For all tramadol regimens combined, the RB was 1.93 (95% CI, 1.56-2.39), with an NNT of 2.2. For all nefopam regimens combined, the RB was 2.62 (95% CI, 2.02-3.40), with an NNT of 1.7. Methylphenidate, midazolam, dolasetron, ondansetron, physostigmine, urapidil, and flumazenil were tested in no more than 3 trials each, with a limited number of patients.