A three way single blind cross-over comparison of progabide, valproate and placebo, as adjunctive therapy, was undertaken in 64 patients with therapy-resistant partial and generalised seizures. The study was not completed because of the incidence of elevated hepatic enzymes on progabide. Analysis of efficacy showed progabide to be inferior to valproate against all seizure types, particularly against tonic-clonic seizures. Valproate was superior to placebo against all seizure types, partial and tonic-clonic seizures. Progabide did not differ significantly from placebo in any instance. In addition progabide caused elevation of hepatic enzymes which was symptomatic in one case, and was associated with an interaction with phenytoin which resulted in symptoms of intoxication in some cases.