Three RCTs (1,175 patients) were included.
All 3 RCTs were of a high quality (rated A on the Cochrane quality criteria and 5 on the Jadad scale).
Olanzapine and risperidone were associated with significantly greater benefit than placebo; the OR (3 RCTs, 911 patients) was 0.59 (95% CI: 0.44, 0.78) and the NNT was 8 (95% CI: 5, 18). There was no evidence of significant heterogeneity (chi-squared 1.71, d.f.=2, Z=3.70). Both drugs were shown separately to be significantly better than placebo. The OR for risperidone versus placebo (2 RCTs, 705 patients) was 0.63 (95% CI: 0.46, 0.85) and there was no evidence of significant heterogeneity (chi-squared 0.94, d.f.=1, Z=2.97); the OR for olanzapine versus placebo (1 RCT, 206 patients) was 0.59 (95% CI: 0.24, 0.87).
More than 80% of the patients complained of at least one adverse event. Olanzapine and risperidone were associated with significantly more EPS and somnolence than placebo. There was no evidence of significant heterogeneity among studies for either outcome. For EPS, the OR was 2.04 (95% CI: 1.24, 3.33) and the NNH was 13 (95% CI: 8, 40); for somnolence, the OR was 1.74 (95% CI: 1.18, 2.57) and the NNH was 10 (95% CI: 7, 22). There were no significant differences in withdrawal rates between the active drugs and placebo (OR 1.31, 95% CI: 1, 1.71).
For risperidone versus haloperidol (1 RCT, 344 patients), there was no significant difference in clinical improvement, adverse events or drop-outs between the treatment groups.