Sodium dichloroacetate (DCA) or placebo, two infusions 30 min in duration and 8 h apart, was administered to healthy subjects under double-blind conditions. The objectives were to characterize accurately the tolerability of DCA, its pharmacokinetics, and the reduction of resting serum lactate concentration by DCA. A hybrid, one-compartment pharmacokinetic model fitted best, with zero-order elimination mean of 27.9 micrograms/ml/h at concentrations above about 80 to 120 micrograms/ml, and with first-order elimination (mean kelim = 0.54) at lower serum concentrations of DCA. Resting serum lactate was dose-independently, maximally reduced within 15 min of the end of all active infusions. The duration of suppression of resting serum lactate was dose-dependent, from 4.5 h (30 mg/kg) to > 8 h (100 mg/kg). Second infusions (15-50 mg/kg) again promptly and maximally reduced resting serum lactate. Hysteresis loops were asymmetrical for all doses but exhibited change in shape that was dose-dependent; no good pharmacokinetic-pharmacodynamic model could be fitted that was consistent between doses. Infusions were well tolerated, 100 mg/kg + 50 mg/kg being the highest doses. Somnolence, the only dose-related adverse event, was reported by 3 of 37 subjects at times corresponding to the highest serum DCA concentrations. This study demonstrates the tolerability of i.v. DCA, proposes a simple pharmacokinetic model for its elimination, characterizes the dose-response relationship in terms of time course of effect, shows the dissociation between elimination of DCA and offset of response and will guide further studies of DCA in patients with head injury or stroke.