The pharmacokinetics of intravenously administered valproic acid (VPA) were investigated in 16 healthy male volunteers in a single-dose, fasting, four-period, randomized, double-blind, placebo-controlled, parallel design study. Subjects were randomly assigned to be infused a single dose of sodium valproate equivalent to lo00 mg VPA or placebo over each of four different time periods. Valproate concentrations in plasma were determined using gas chromatography with flame ionization detection. The pharmacokinetic parameters were determined by both non-compartmental and modeldependent techniques. Analyses of variance (ANOVAs) were performed to detect any statistical differences among the regimens. Overall, the pharmacokinetics of valproate were similar after infusions of 5, 10, 30, and 60 min, with an average terminal-phase half-life of 15-9 h. There were modest differences in overall clearances among the regimens, with the 5 min infusion producing a mean area under the plasma concentrationtime curve (AUC; 1877 pg-h ml-') that was significantly (13 to 16 per cent) higher than the means for the longer infusions (1614-1656 pg-h ml-I). Differences in distribution were also noted as a function of infusion duration. The shortest duration produced a significantly smaller terminal volume of distribution (12.8 vs 14-2-15.1 1) and more rapid tissue equilibration. The a-phase rate constant declined from a mean of 5.1 h-' after the 5min infusion to a mean of 0.9h-' after the 60min infusion. The distributional differences are almost certainly related to the saturable protein binding of valproate. However, the lower clearance after the 5 min infusion indicates that there may have also been partial saturation of one of the metabolic pathways of valproate during the distributive phase, and that the increase infu was smaller than the decrease in CLint, such that the product of f;CLint showed a net decrease.