A model-based approach to dose selection in early paediatric development
Massimo Cella(1), Gijs Santen(1), Frederik Gorter de Vries(1),Meindert Danhof(1), Oscar Della Pasqua(1,2)
(1)Division of Pharmacology, LACDR, Leiden University, Leiden, The Netherlands, (2)Clinical Pharmacology & Discovery Medicine, GlaxoSmithKline, Greenford, UK
Introduction: The use of pharmacokinetic-pharmacodynamic (PKPD) modelling has proven to be a powerful tool in clinical dose finding in adults. However, empiricism seems to rule the rationale for dose selection and adjustment even in cases where the concentration-effect relationship is known to be similar in adult and in children. A bridging strategy based on integrated pharmacokinetic modelling could provide better dosing recommendation for the paediatric indication.
Objectives: The aim of the current investigation was to show the suitability of combining prior information from dense pharmacokinetic sampling in adults with data sparse sampling in paediatric studies. The approach has been tested using abacavir, an antiviral used in the treatment of HIV infection.
Methods: The pharmacokinetics of abacavir was evaluated using nonlinear mixed-effects modelling in NONMEM v5. Using pharmacokinetic data in adults as prior for the fitting procedure, a sensitivity analysis based on a series of simulation scenarios was performed to explore the impact of sampling frequency and group size on the estimation of pharmacokinetic parameters in children. For each scenario, a reference cohort of 200 adults with data from serial sampling was simulated and combined with sparse data from a hypothetical adaptive design study in which 5, 10 or 20 children were included, each with limited data from sparse sampling. Population mean values and confidence intervals were estimated for clearance, volume of distribution and absorption rate constant. In children, the parameter point estimate was assumed to deviate from adults across a range that varied from -100% to +300%.
Results: Based on a one-compartment model with first order absorption and no covariate effect of weight on clearance, differences in the paediatric subpopulation as low as 10% can be identified if study cohort includes 20 children, each with 5 samples. The threshold for detecting differences between children and adult increases to approximately 15%, if the cohort is reduced to 5 or 10 children. Sample size does not seem to affect parameter estimation. The relative errors associated with clearance, volume of distribution and absorption rate constant were less than 5%, 7% and 3%, respectively in the worst case scenario.
Conclusions: Our results show that pharmacokinetic parameter distribution can be accurately estimated for a new population by integrated data analysis. Modelling of abacavir data confirmed our findings in the sensitivity analysis. Subsequent evaluation of the method is currently ongoing and includes bootstrapping and refitting of the data. The method seems to provide a feasible alternative to the evaluation of pharmacokinetics in children in early drug development.