Consensus panel recommendations regarding choice of an antipsychotic agent for schizophrenia differ markedly, but most consider second-generation antipsychotics (SGAs) as a homogeneous group. It has been suggested that SGAs seem falsely more efficacious than first-generation antipsychotics (FGAs) as a result of reduced efficacy due to use of a high-dose comparator, haloperidol. We performed (1) a meta-analysis of randomized efficacy trials comparing SGAs and FGAs, (2) comparisons between SGAs, (3) a dose-response analysis of FGAs and SGAs, and (4) an analysis of the effect on efficacy of an overly high dose of an FGA comparator. Literature search of clinical trials between January 1953 and May 2002 of patients with schizophrenia from electronic databases, reference lists, posters, the Food and Drug Administration, and other unpublished data. We included 124 randomized controlled trials with efficacy data on 10 SGAs vs FGAs and 18 studies of comparisons between SGAs. Two of us independently extracted the sample sizes, means, and standard deviation of the efficacy data. Using the Hedges-Olkin algorithm, the effect sizes of clozapine, amisulpride, risperidone, and olanzapine were 0.49, 0.29, 0.25, and 0.21 greater than those of FGAs, with P values of 2 x 10-8, 3 x 10-7, 2 x 10-12, and 3 x 10-9, respectively. The remaining 6 SGAs were not significantly different from FGAs, although zotepine was marginally different. No efficacy difference was detected among amisulpride, risperidone, and olanzapine. We found no evidence that the haloperidol dose (or all FGA comparators converted to haloperidol-equivalent doses) affected these results when we examined its effect by drug or in a 2-way analysis of variance model in which SGA effectiveness is entered as a second factor. Some SGAs are more efficacious than FGAs, and, therefore, SGAs are not a homogeneous group.