Limited sampling strategy for a population pharmacokinetic modelling of cocktail of phenotyping drugs
Thu Thuy Nguyen (1,2), Henri Bénech (3), Marcel Delaforge (3), France Mentré (1), Natacha Lenuzza (2)
(1) IAME-UMR 1137, INSERM and University Paris Diderot, Paris, France; (2) CEA, LIST, LADIS, Gif-sur-Yvette, France; (3) CEA, DSV, iBiTecS, Gif-sur-Yvette, France
Objectives: Cocktail approach using a combination of probes to phenotype several cytochromes P450 and transporters, is of high interest in anticipating drug-drug interactions and personalized medicine . Phenotyping indexes (PI) which are obtained from the area under the concentration-time curves, can be derived from a few samples using nonlinear mixed effect models (NLMEM) and maximum a posteriori estimation. We aimed to 1) propose a limited sampling strategy, allowing correct estimation of PI for several probes while providing as much flexibility as possible in sampling timing; 2) illustrate this strategy for two drugs often used in phenotyping tests: midazolam (probe for CYP3A) and digoxin (P-glycoprotein).
Methods: Data of a previous study with ten healthy volunteers  were analyzed to develop population parent/metabolite models for several drugs. In order to optimize joint design for a cocktail, we proposed to use the compound D-optimality  by maximizing a weighted sum of log determinants of the expected population Fisher information matrix (FIM)  for these models. Sampling windows were computed around the optimal fixed times, based on recursive random sampling and Monte-Carlo simulation [4,5], satisfying an expected joint loss of efficiency below 10% for each molecule. We illustrated this strategy to find a sparse and flexible design common to three compounds: midazolam, its metabolite 1-OH-midazolam and digoxin. The obtained design was evaluated by clinical trial simulations for estimation of population and individual PI.
Results: A two-compartment model (first order absorption, linear elimination) adequately fitted the concentrations of digoxin while a joint two-compartment-parent/one-compartment-metabolite model was found for midazolam and 1-OH-midazolam. The common design ξ was composed of six samples (0.25, 1, 2.5, 5, 12, 48h post-administration) instead of nine samples if optimizing separately design for each drug, giving predicted relative standard errors below 20% for both PI. The window design achieved an efficiency above 90% relative to ξ for population analysis of each drug and showed good performance for estimation of individual PI.
Conclusions: By combining NLMEM, compound design and sampling windows based on FIM, we were able to determine sparse and flexible samples allowing correct estimation of PI for three compounds . This approach can be used to efficiently design studies with cocktails including more drugs.
 Lenuzza N, Duval X, Nicolas G, Thévenot E, Job S, Videau O, Narjoz C, Loriot MA, Beaune P, Becquemont L, Mentré F, Funck-Brentano C, Alavoine L, Arnaud P, Delaforge M, Bénech H. Safety and pharmacokinetics of the CIME combination of drugs and their metabolites after a single oral dosing in healthy volunteers. Eur J Drug Metab Pharmacokinet 2014 [Epub ahead of print].
 Atkinson AC. DT-optimum designs for model discrimination and parameter estimation. J. Stat. Plan. Inference 2008; 138(1): 56–64.
 Mentré F, Mallet A, Baccar D. Optimal design in random-effects regression models. Biometrika 1997; 84(2): 429–442.
 Foo LK, McGree J, Duffull S. A general method to determine sampling windows for nonlinear mixed effects models with an application to population pharmacokinetic studies. Pharm Stat 2012; 11(4): 325–333.
 Ogungbenro K, Aarons L. An effective approach for obtaining optimal sampling windows for population pharmacokinetic experiments. J Biopharm Stat 2009; 19(1): 174–189.
 Nguyen TT, Bénech H, Delaforge M, Lenuzza N. Design optimisation for pharmacokinetic modeling of a cocktail of phenotyping drugs. Pharm Stat 2015 [Epub ahead of print].