The availability of an intravenous formulation now makes possible rapid administration of valproate (VIA) loading doses, but estimates of key VIA pharmacokinetic parameters in patients have limited the use of this approach. VPA disposition was characterized in 112 epilepsy patients, with or without enzyme inducing comedications, randomized to either 3.0 or 1.5 mg/kg/min infusions of valproate sodium injection. Maximum dose was less than or equal to15 mg/kg per infusion. Total and unbound plasma VPA concentrations were determined from blood samples obtained prior to and for 6 It following the infusion. Analyses of covariance assessed the effect of induction, weight, age, gender, albumin, creatinine, and infusion rate on pharmacokinetics. Maximum total and unbound VPA concentrations were 94 and 14 mg/l, respectively. Total concentration fell below 50 mg/l within 3 h in induced and 6 h in uninduced patients. VIA unbound fraction decreased from 15% at maximum concentration to 9% at 45 mg/l. The mean (S.D.) distribution volume was 0.21 (0.044) l/kg. Induction status, albumin concentration, and infusion rate significantly affected pharmacokinetics. Measurement of unbound VPA may be useful when alterations in binding are suspected. Infusions up to 3 mg/kg/min produce predictable total VPA concentrations when induction status and albumin levels are considered. (C) 2002 Elsevier Science B.V. All rights reserved.