Predictive Performance of Population Pharmacokinetic Models of Levetiracetam in Children and Evaluation of Dosing Regimen

Abstract

Levetiracetam is a broad-spectrum antiepileptic drug that exhibits high interindividual variability in serum concentrations in children. A population pharmacokinetic approach can be used to explain this variability and optimize dosing schemes. The objectives are to identify the best predictive population pharmacokinetic model for children and to evaluate recommended doses using simulations and Bayesian forecasting. A validation cohort included children treated with levetiracetam who had a serum drug concentration assayed during therapeutic drug monitoring. We assessed the predictive performance of all the population pharmacokinetic models published in the literature using mean prediction errors, root mean squared errors, and visual predictive checks. A population model was finally constructed on the data, and dose simulations were performed to evaluate doses. We included 267 levetiracetam concentrations ranging from 2 to 69 mg/L from 194 children in the validation cohort. Six published models were externally evaluated. Most of the models underestimated the variability of our population. A 1-compartment model with first-order absorption and elimination with allometric scaling was finally fitted on our data. In our cohort, 57% of patients had a trough concentration <12 mg/L and 12% <5 mg/L. To reach a trough concentration >5 mg/L, doses ≥30 mg/kg/d for patients ≤50 kg and ≥2000 mg/d for patients >50 kg are required. In our population, a high percentage of children had low trough concentrations. Our population pharmacokinetic model could be used for therapeutic drug monitoring of levetiracetam in children.