Development of a PK library of parent drug/metabolite integrated models
N. Lenuzza (1), S. Troncale(1,*), G. Nicolas (2), M. Delaforge (2), H. Bénech (2) and E. Thévenot (1)
(1) CEA, LIST, Data Analysis Tools Laboratory, 91191 Gif-Sur-Yvette CEDEX, FRANCE; (2) CEA, Pharmacology and Immunoanalysis Unit, DSV/iBiTecS, 91191 Gif-Sur-Yvette CEDEX, FRANCE; (*) present address: Institut Curie, Paris, FRANCE
Objectives: Variations of drug metabolism in patients can prevent therapeutic efficacy or lead to toxicity. The CIME cocktail [1] has been designed to measure in vivo the phenotypes of 10 major enzymes involved in drug metabolism (Cytochromes P450, glucuronyl transferase, active transporters) by simultaneous quantification of 10 substrates and their main metabolites, and subsequent estimation of appropriate PK parameters by non-linear mixed-effect modelling. However, no models combining both the parent drug and its metabolite are yet available in the main PK software. We have thus developed and validated such a library in the PK-PD MONOLIX reference software [2].
Methods: Analytical solutions were obtained by using the Laplace transform, and identifiability was assessed by similarity transform approach or by integro-differential algebra using DAISY software [3]. Both solutions and equations were implemented in MLXTRAN. To illustrate the use of our library, we carried out a simulation study with the acetaminophen parent drug and its metabolite acetaminophen glucuronide (which are both measured in the CIME approach). The selected model (2-compartments/1-compartment with Tlag for the parent drug absorption) was chosen to generate data, according to different sampling schemes (rich or sparse), a log-normal distribution of random effects and a proportional residual error model. Parameter estimations were performed with MONOLIX and results were evaluated in terms of bias, standard error (SE) and relative efficiency to evaluate the reliability of estimations. To assess the robustness of MONOLIX estimations, misspecified models were also evaluated (i.e. normal distribution of random effect, or additive residual error).
Results: Models accounting for 1/1, 1/2, 2/1 and 2/2 compartments, with or without Tlag, first-pass effect and equal volumes, for either a single-dose intravenous or oral administration, were implemented. Our simulation study shows good convergence and reliable, robust estimations of parameters with MONOLIX. Interestingly, by using ordinary differential equations (ODE) instead of analytical solutions, we observed smaller SE in the case of rich data and higher efficiency in the case of sparse data.
Conclusions: A comprehensive library of integrated parent drug/metabolite models has been developed and validated. Such models will be of high interest to quantify drug metabolism and drug-drug interactions in human.
References:
[1] Videau et al. (2010). Biochemical and analytical development of the CIME cocktail for drug fate assessment in humans. Rapid Commun Mass Spectrom. 24:2407-19.
[2] MONOLIX 2.4 User Guide. http://software.monolix.org.
[3] Bellu et al. (2007). DAISY: A new software tool to test global identifiability of biological and physiological systems. Comput Methods Programs Biomed. 88:52-61.