Comparison of different tools for the optimization of a pediatric clinical trial
N. Gobeau, B. Boulanger, L.Sargentini-Maier
UCB Pharma, Braine-l’Alleud, Belgium
Objective: compare the features, ease of use and results of three publicly available pieces of software (PFIM version 3.0 (1), POPT version 3.0 (2) and PopED version 2.08 (3)) for optimizing a pediatric clinical trial.
Methods: A pediatric clinical trial is planned primarily to characterize the steady-state pharmacokinetics (PK) of a drug under development in children aged 1 month to 16 years old. The design is a fixed 3-step dose up-titration study. Three different tools (PFIM, POPT and PopED) were employed to optimize the sampling times in order to limit the number of samples. The optimization was undertaken based on the assumption that the PK in children was described by a population PK model previously developed in adults, scaled by weight. Due to the different features of the optimization tools, the problem was set up differently to take into account the varying PK parameters with age/weight: optimization per weight category in PFIM; optimization per age category in POPT with 3 competing models between a population with respectively minimum, median and maximum weight; optimization per age category in POPED assuming a uniform weight distribution. The performance of the designs obtained were compared with a 21 samples design initially proposed by the clinicians via simulations as described in Hooker et al. (4).
Results: The number of sampling times could be reduced from 21 to 6 per patient with a total number of patients of 48 without compromising the estimation of the PK parameters. The optimized times designs obtained with the different optimization tools were similar; the main differences between the tools were found to be the implementation of the problem. The performance of the optimized design via simulations, expressed in terms of Relative Mean Square Error (4), was found to be acceptable compared to the one obtained with the rich sampling design.
Conclusions: The optimization software tools tested were found to be useful in helping reduce the number of samples in a pedicatric clinical trial. The main difference between the tools was the implementation of the problem. As for non-linear models in general, the search of optimal designs require appropriate prior knowledge of the PK parameters to make the optimal designs valuable. For that very reason, the pediatric PK model is being refined using parameters estimated using a Physiologically-Based PharmacoKinetic (PBPK) model.
References:
[1] http://www.pfim.biostat.fr/
[2] http://www.winpopt.com/
[3] http://poped.sourceforge.net/
[4] Hooker, A. C., Foracchia, M., Dodds, M. G., and Vicini, P., "An evaluation of population D-optimal designs via pharmacokinetic simulations," Ann Biomed Eng, Vol. 31, No. 1, 2003, pp. 98-111.
