OBJECTIVES :The purpose of this study is to provide economic information comparing olanzapine- and haloperidol-treated patients from the subset of French patients that participated in a large, international, randomised clinical trial in schizophrenia. METHODS : Patients were evaluated from initiation until discontinuation of treatment or completion of the 54-week study. The primary clinical measure was ‘marked clinical response’ (derived from BPRS total scores). The secondary measure was ‘marked clinical improvement’ (derived from CGI severity of illness scores). The primary economic measure was mean per diem, per patient total direct medical costs. RESULTS : A total of 275 French patients were included in the study. Demographics and other baseline differences between olanzapine- and haloperidol-treated patients were not statistically significant. Olanzapine-treated patients (205 ? 142 days) experienced significantly (p<0.001) longer valuation periods than did haloperidol-treated patients (132 ? 129 days). Olanzapine-treated patients (54%) were significantly (p = 0.03) more likely to experience a marked clinical response than were haloperidol-treated patients (40%). Olanzapine-treated patients (69%) were significantly (p = 0.02) more likely to experience a marked clinical improvement than were haloperidol-treated patients (54%). The mean per diem, per patient total direct medical costs were statistically lower (p = 0.033) for the olanzapine-treated patients (F619 ? 509) compared with the haloperiodal-treated patients (F756 ? 478). CONCLUSION : Olanzapine treatment was associated with significantly better clinical outcomes and significantly lower per diem total direct medical costs than was haloperidol treatment. The findings in this study indicate that olanzapine is cost-effective compared with haloperidol for the treatment of schizophrenia. These findings are of increased relevance in France and add to the existing, large body of evidence supporting olanzapine’s cost-effectiveness relative to typical antipsychotics.