To establish the impact of antipsychotic medication on the incidence of diabetes, we have analysed data from the prospective French Cohort study of mortality in schizophrenia. This generated a large database of 3470 patients with schizophrenia initially enrolled in 1993 for which data collection on comorbidity has been collected systematically every three years since. The primary objective of the study was to evaluate the mortality rate in the study cohort. From this database, the prevalence and incidence of diabetes can be studied. In 1993, nine patients in 10 were prescribed conventional antipsychotics and the remainder amisulpride and clozapine. Since the introduction of risperidone and olanzapine, atypical antipsychotics are now used in one-third of patients. Multiple antipsychotic medications are used in many patients, with the exception of those receiving clozapine. At inclusion, 2.2% of patients in the study cohort had a diagnosis of diabetes. Morbidity rates were higher in females than in males (3.4% and 1.6%, respectively) and it was only in females that the standard morbidity ratio with respect to the general population was significantly elevated (2.2; 95% confidence intervals: 1.6 and 2.9). The prevalence of diabetes in the study cohort rose over the course of the study. Multivariate regression analysis was performed in order to identify potential determinants of diabetes. For pre-existing diabetes, four factors were identified: age at first hospitalisation, age, obesity and duration of schizophrenia. The same factors with the exception of length of illness were identified for the incident cases. There was no evidence for an interaction between the class and type of antipsychotic medication and risk of treatment emergent diabetes. However, the use of multiple antipsychotic treatments makes the individualisation of specific risks associated with any other antipsychotic drug impossible.