BACKGROUND: Because of the debate about whether second-generation antipsychotic drugs are better than first-generation antipsychotic drugs, we did a meta-analysis of randomised controlled trials to compare the effects of these two types of drugs in patients with schizophrenia. METHODS: We compared nine second-generation antipsychotic drugs with first-generation drugs for overall efficacy (main outcome), positive, negative and depressive symptoms, relapse, quality of life, extrapyramidal side-effects, weight gain, and sedation. FINDINGS: We included 150 double-blind, mostly short-term, studies, with 21 533 participants. We excluded open studies because they systematically favoured second-generation drugs. Four of these drugs were better than first-generation antipsychotic drugs for overall efficacy, with small to medium effect sizes (amisulpride -0.31 [95% CI -0.44 to -0.19, p<0.0001], clozapine -0.52 [-0.75 to -0.29, p<0.0001], olanzapine -0.28 [-0.38 to -0.18, p<0.0001], and risperidone -0.13 [-0.22 to -0.05, p=0.002]). The other second-generation drugs were not more efficacious than the first-generation drugs, even for negative symptoms. Therefore efficacy on negative symptoms cannot be a core component of atypicality. Second-generation antipsychotic drugs induced fewer extrapyramidal side-effects than did haloperidol (even at low doses). Only a few have been shown to induce fewer extrapyramidal side-effects than low-potency first-generation antipsychotic drugs. With the exception of aripiprazole and ziprasidone, second-generation antipsychotic drugs induced more weight gain, in various degrees, than did haloperidol but not than low-potency first-generation drugs. The second-generation drugs also differed in their sedating properties. We did not note any consistent effects of moderator variables, such as industry sponsorship, comparator dose, or prophylactic antiparkinsonian medication. INTERPRETATION: Second-generation antipsychotic drugs differ in many properties and are not a homogeneous class. This meta-analysis provides data for individualised treatment based on efficacy, side-effects, and cost.