Concentration-Response Modelling of Adjunctive Lamotrigine Extended-Release for Primary Generalised Tonic-Clonic Seizures
Chao Chen (1), Jonathan Bullman (1), John Messenheimer (2)
(1) Clinical Pharmacology Modelling and Simulation, (2) Medicines Development Centre, GlaxoSmithKline
Background: Lamotrigine (LTG) immediate-release (IR) is indicated for several seizure types, including partial seizures and Primary Generalised Tonic-Clonic (PGTC) seizures, in adult and paediatric patients. Efficacy and safety of an extended-release (XR) formulation as an adjunctive therapy for PGTC seizures were investigated in a placebo-controlled Phase III trial. Daily doses were escalated to the recommended maintenance levels for the IR, depending on the known nature of PK interaction with the background drugs. Sparse PK samples were collected throughout the trial; and seizure frequency was calculated for baseline and treatment periods.
Objectives: This was a retrospective analysis to explore PGTC seizure frequency as a function of LTG concentration following XR administration as an adjunctive therapy.
Methods: A previously developed population PK model, along with covariates body weight and PK interaction, was fitted to the trial data. Model parameters and steady-state average serum concentration for individual patients were estimated. Seizure frequency data were modelled as an Emax or linear function of the concentration, with an exponential error term. The model included baseline and placebo terms. Between-subject variability was explored for structure parameters. NONMEM version 5 was used for the analyses.
Results: The PK dataset included up to six samples per patient from 66 patients. Although absorption and distribution parameters could not be estimated with good precision, likely due to the flatness of the curve and sparseness of the data, individual concentrations were well estimated. Consistent with the dosing strategy, the spread of the estimated steady-state concentration was limited: 90% of the values were within a five-fold range. The data could not support an Emax model, which would be mechanistically and clinically meaningful. A small bias was observed, in the pattern that was expected when a linear function was forced through the data. Nonetheless, within the observed concentration range, the linear model described the data reasonably well for both placebo and active treatments.
Conclusion: These results suggest a potential for further therapeutic benefits at higher exposure in some patients. The analysis revealed certain limitations associated with retrospective analysis, thus raising the issue of cost-effective application of modelling approach and highlighting the importance of design-by-simulation.