Levetiracetam exposure-response analysis in children with partial onset seizures
E. Snoeck (1), R. Schoemaker (1), A. Stockis (2), C. Otoul (2), M.L. Sargentini-Maier (2)
(1) Exprimo NV, Mechelen, Belgium; (2) Pharmacometrics Department, UCB Pharma SA, Braine-l’Alleud, Belgium
Objectives : To compare the performance of several seizure count-based exposure-response models for levetiracetam in children with epilepsy and to undertake simulations to provide a rationale for an optimal dosing scheme.
Methods : Exposure-response analysis was carried out using data from a randomized, double-blind placebo controlled, add-on efficacy trial of levetiracetam in children (4-16 years) with partial onset seizures. The trial consisted of an 8-week prospective baseline followed by two 2-week fixed dose up-titration periods, an evaluation period of 10 weeks at 60 mg/kg/day or at the maximum tolerated dose, and a 6-week withdrawal period. Modeling of daily seizures was performed by nonlinear mixed effects modeling in NONMEM VI with the Laplace estimation method. Several statistical models were compared: Poisson, negative binomial, with or without zero-inflation, and with or without Markov elements. A Mixture procedure was used to separate subjects exhibiting reduced or increased seizure frequency from baseline. The drug effect was modeled in improving subjects as an Emax function of dose or AUCs (individual posterior estimates derived from a population pharmacokinetics model).
Results : The zero-inflated negative binomial model including Markov elements was found superior to all other models. In particular, the OFV dropped by >17000 units compared to the base Poisson model. Using individual AUCs instead of doses did not yield any significant improvement. The baseline seizure frequency was 0.64/day when the previous day was seizure-free and 1.06/day when one or more seizures occurred on the previous day. The improving sub-populations amounted to 78% and 52% of the levetiracetam and placebo groups, respectively. The population ED50 was 287 mg/day. Simulations by bootstrapping (using the observed body weights) showed that most of the drug effect was reached at 20 mg/kg/day and that the clinical response in the range of 20-60 mg/kg/day in children was fairly similar to that of 1000-3000 mg/day in adults. Simulations predicted that levetiracetam 60 mg/kg/day is likely to result in a reduction in seizure frequency of at least 55% in half of the improving subjects.
Conclusions: The zero-inflated negative binomial model with Markov elements has superior features for describing daily seizure count data. The daily levetiracetam dose of 20-60 mg/kg is suggested to be optimal in adjunctive therapy of children with refractory partial onset seizures.