OBJECTIVE: To compare the risk of extrapyramidal syndrome (EPS) between risperidone, olanzapine, and haloperidol, taking into account patients' past antipsychotic drug use and past EPS. METHODS: Data were obtained from the PHARMO-database, containing filled prescriptions of 450,000 community-dwelling people in the Netherlands from 1986 through 1999. We defined cohorts of first-time users of haloperidol, risperidone, or olanzapine aged 15 to 54 years. In the first 90 days of treatment, we assessed the occurrence of EPS, defined as first use of any antiparkinsonian agent. We estimated relative risks of EPS for risperidone and olanzapine versus haloperidol using a Cox proportional hazards model. Patients were subdivided according to prior use of antipsychotic and antiparkinsonian drugs. RESULTS: We identified 424 patients starting treatment with haloperidol, 243 with risperidone, and 181 with olanzapine. Prior use of antipsychotic plus antiparkinsonian medication was significantly more frequent among users of risperidone and olanzapine than in those using haloperidol (36.2%, 40.3%, and 4.5%, respectively; p < 0.001). Within most subgroups of comparable treatment history, patients using risperidone and olanzapine showed reduced risks of EPS compared with haloperidol, although some of these findings did not reach statistical significance (RR 0.03-0.22). However, this was not observed for patients using risperidone who had experienced EPS in the past (RR 1.30; 95% CI 0.24 to 7.18). CONCLUSIONS: In general, we observed reduced risks of EPS for risperidone and olanzapine compared with haloperidol within subgroups of patients with a similar treatment history. However, the added value of risperidone in patients who have experienced EPS in the past needs further study.