In a randomised multicentre study, the prophylactic efficacy of lithium and carbamazepine was compared in 144 patients with bipolar disorder (74 vs. 70 patients; observation period: 2.5 years; lithium serum level: 0.63 +/- 0.12 mmol/l, carbamazepine dose: 621 +/- 186 mg/day). Hospitalisations, recurrences, need of psychotropic comedication and adverse effects prompting discontinuation were defined as treatment failures. Survival analyses regarding hospitalisations and recurrences showed no statistically significant differences between both drugs. Results were distinctly in favour of lithium, considering recurrences combined with comedication (P = 0.041) and/or adverse effects (P = 0.007). Whereas adverse effects prompting discontinuation were more frequent under carbamazepine (9 vs. 4, ns), lithium patients reported more often slight/moderate side effects (61% vs. 21% after 2.5 years; P = 0.0006). In completers, recurrences occurred in 28% (lithium) vs. 47% (carbamazepine) of the patients (P = 0.06). Lithium seems to be superior to carbamazepine in maintenance treatment of bipolar disorder, in particular when applying broader outcome criteria including psychotropic comedication and severe side effects.