In a randomised multicentre study, the prophylactic efficacy of lithium and carbamazepine was compared in schizoaffective disorder. A total of 90 ICD-9 schizoaffective patients were included in the maintenance phase (2.5 years). They were also diagnosed according to RDC and DSM-III-R and classified into subgroups. Mean serum levels were 0.58 +/- 0.12 mmol/l for lithium and 6.4 +/- 1.5 micrograms/ml for carbamazepine (mean dose 643 +/- 179 mg/d). Outcome criteria were hospitalisation, recurrence, concomitant psychotropic medication and adverse effects leading to discontinuation. There were more non-completers under carbamazepine than under lithium (p = 0.02). Survival analyses demonstrated no significant differences between lithium and carbamazepine in treatment outcome. Patient's ratings of side effects (p = 0.003) and treatment satisfaction (p = 0.02) favoured carbamazepine. Following the RDC criteria, patients of the schizodepressive and non-classifiable type did better under carbamazepine (p = 0.055 for recurrence), whereas in the schizomanic patients equipotency of both drugs was found. Applying DSM-III-R, carbamazepine demonstrated a superiority in the patient group with more schizophrenia-like or depressive disorders (p = 0.040 for recurrence), but not in patients fulfilling the DSM-III-R criteria of bipolar disorder. Lithium and carbamazepine seem to be equipotent alternatives in the maintenance treatment of broadly defined schizoaffective disorders. However, in subgroups with depressive or schizophrenia-like features and regarding its long-term tolerability carbamazepine seems to be superior.