Physiologically based pharmacokinetic modeling of disposition and drug-drug interactions for valproic acid and divalproex.

Valproic acid (VPA) is an older first-line antiepileptic drug with a complex pharmacokinetic (PK) profile, currently under investigation for several novel neurologic and non-neurologic indications. Our study objective was to design and validate a mechanistic model of VPA disposition in adults and children; and evaluate its predictive performance of drug-drug interactions (DDIs). This study expands upon existing physiologically based pharmacokinetic (PBPK) models for VPA by incorporating UGT enzyme kinetics and an advanced dissolution, absorption, and metabolism (ADAM) model for extended-release (ER) formulation. PBPK models for VPA IR and ER formulations were constructed using Simcyp Simulator (Version 15). First-order absorption was used for the immediate-release (IR) formulation and the ADAM model, including a controlled-release profile, for ER. Data from twenty-one published clinical studies were used to assess model performance. The model accurately predicted the concentration-time profiles of IR formulation for single-dose and steady-state doses ranging from 200mg to 1000mg. Similarly profiles were also simulated for ER formulation after a single-dose and steady-state doses of 500mg and 1000mg, respectively. In addition, simulated PK profiles agreed well with the observed data from studies in which VPA ER formulation was given to pediatric patients and VPA IR formulation to adult patients with cirrhosis. The model was further validated with individual adult data from a Phase I clinical trial consisting of eight cohorts after IV infusion of VPA with doses ranging from 15 to 150mg/kg. Co-administrations of VPA as an enzyme-inhibitor with victim drug phenytoin or lorazepam, as well as a substrate with enzyme inducer carbamazepine or phenobarbital, were simulated with the model to evaluate drug-drug interaction. The simulated serum concentration-time profiles were within the 5th and 95th percentiles, and the majority of the predicted area-under-the-curve (AUC) and peak plasma concentration (C