BackgroundThe increased prevalence of diabetes in schizophrenia is partly attributed to antipsychotic treatment, in particular second-generation antipsychotics, but the evidence has not been systematically reviewed.AimsSystematic review and meta-analysis comparing diabetes risk for different antipsychotics in people with schizophrenia.MethodWe searched MEDLINE, PsycINFO, EMBASE, International Pharmaceutical Abstracts, CINAHL and Web of Knowledge until September 2006. Studies were eligible for inclusion if the design was cross-sectional, case-control, cohort or a controlled trial in individuals with schizophrenia or related psychotic disorders, where second-generation antipsychotics (defined as clozapine, olanzapine, risperidone and quetiapine) were compared with first-generation antipsychotics and diabetes was an outcome. Data were pooled using random effects inverse variance weighted meta-analysis.ResultsOf the studies that met the inclusion criteria (n=14), 11 had sufficient data to include in the meta-analysis. Four of these were retrospective cohort studies. The relative risk of diabetes in patients with schizophrenia prescribed one of the second-generation v. first-generation antipsychotics was 1.32 (95% CI 1.15-1.51). There were insufficient data to include aripiprazole, ziprasidone and amisulpride in this analysis.ConclusionsThere is tentative evidence that the second-generation antipsychotics included in this study are associated with a small increased risk for diabetes compared with firstgeneration antipsychotics in people with schizophrenia. Methodological limitations were found in most studies, leading to heterogeneity and difficulty interpreting data. Regardless of type of antipsychotic, screening for diabetes in all people with schizophrenia should be routine.