The concentrations of valproate (VPA) and six of its pharmacologically active, unsaturated metabolites (E-delta 2-VPA, Z-delta 3-VPA, E-delta 3-VPA, E,E-delta 2,3'-VPA, delta 4-VPA, and E-delta 2,4-VPA) were measured in serum and cortical brain samples from 24 patients undergoing epilepsy surgery. Collectively, the six metabolites were present at concentrations < 13% of VPA brain concentrations. Because the six unsaturated metabolites were present at such low brain concentrations, we concluded that these metabolites probably did not contribute significantly to the anticonvulsant effect of VPA. Results from a parallel pharmacodynamic study in rats in which VPA was administered three times daily for 8 weeks supported this conclusion. Only three unsaturated metabolites (E-delta 2-VPA, delta 3-VPA, E,E-delta 2,3'-VPA) were detected in rat brain. No correlation was observed between the time course of anticonvulsant effect [as measured by the timed intravenous pentylenetetrazol (PTZ) test] and the time course of VPA or metabolite concentrations in rat brain. Despite the structural similarity of VPA and its metabolites, striking differences were observed in their serum protein binding and blood-brain distribution properties. In the human brain, VPA and delta 4-VPA exhibited brain-to-free serum concentration ratios that were less than unity. In contrast, compounds with the double bond at the 2- or 3-position had brain:free concentration ratios that were much higher than unity. The structure-distribution relationship observed with VPA and its unsaturated metabolites suggested that these branched-chain fatty acids differ in their asymmetric transport across the blood-brain barrier (BBB).