The present study aimed to estimate the incidence of diabetes in general practice patients who were treated with haloperidol, olanzapine, risperidone or quetiapine monotherapy and in subjects who were not exposed to antipsychotics. The design was a retrospective, up to 2 years, cohort study, with age-, sex- and length of observation-matching between subjects who were exposed and not exposed to antipsychotic drugs. Data were taken from the Health Search database, which contains information from 550 Italian general practitioners. Participants comprised 2071 subjects taking haloperidol, 266 taking olanzapine, 567 taking risperidone and 109 taking quetiapine, in addition to 6026 age- and sex-matched subjects who were not using antipsychotic drugs during the period of observation. Inclusion was limited to initially non-diabetic and antipsychotic drug-free individuals. The main outcome measure was the incidence of drug-treated diabetes. After age and sex correction by Cox regression analysis, the four groups treated with antipsychotics significantly differed from untreated subjects in hazard ratios for diabetes. The ratios for the haloperidol, olanzapine, risperidone and quetiapine groups were 12.4 (95% confidence interval 6.3-24.5), 20.4 (6.9-60.3), 18.7 (8.2-42.8) and 33.7 (9.2-123.6), respectively, with no significant differences when compared to each other.