Genetic effect on a complex parent-metabolite joint PK model developed with NONMEM and MONOLIX
J. Bertrand (1), C.M. Laffont (2), E. Comets (1), M. Chenel (3), F. Mentré (1)
(1)UMR738, INSERM, Paris, France; Université Paris Diderot, Paris, France ; (2)UMR181, Physiopathologie et Toxicologie Expérimentales INRA, ENVT, Toulouse, France ; (3)Institut de Recherches Internationales Servier, Courbevoie, France
Objective: To investigate the influence of genetic covariates on the PK of an antipsychotic agent under development and its active metabolite. Here, we address the case where the active metabolite is back-transformed into the parent drug leading to identifiability problems and numerical difficulties [1].
Methods: We jointly modelled the plasma concentrations of both the parent drug and the metabolite collected in 101 patients on two occasions (after 4 and 8 weeks of treatment) at 1, 3, 6 and 24 hours (trough) following once a day administration. For each patient, genotypes were obtained for 5 CYP2D6 and 2 CYP2C19 polymorphisms.
Four different structural models were compared based on BIC [2], using the FOCE with interaction algorithm implemented in NONMEM 5 and 6 and the SAEM algorithm implemented in MONOLIX 2.4. Models were first written as ordinary differential equations (ODE) systems, but closed form (CF) solutions were subsequently derived to facilitate further analyses.
Once selected the structural model, addition of between- and within-subject variances was investigated using likelihood ratio tests. Genetic covariates were finally included following an ascendant selection using Wald test. A permutation approach was performed to assess the p-values of the covariates remaining in the final model [3].
External evaluation with normalized prediction discrepancies was performed using a study involving healthy volunteers [4].
Results: The model selection was similar using either NONMEM or MONOLIX and ODE or CF. The final PK model included two compartments with a back-transformation and a first-pass effect where a fraction FP of the dose reaches the circulation as parent and a fraction 1-FP reaches the circulation as metabolite. Volumes of parent drug and metabolite were set to be equal, and a dose-dependent decrease in bioavailability was taken into account. Five of the 8 model parameters had between-subject variances significantly different from 0, and only the clearance of the parent had a significantly non null within-subject variance. The clearance of the metabolite through processes other than back-transformation was decreased by 35% in CYP2D6 poor metabolizers.
Conclusion: A similar model was selected in all configurations. Using the SAEM algorithm we could estimate all population parameters as well as between- and within-subject variabilities with standard errors for this rather complex model. CYP2D6 polymorphisms appeared to affect the PK of the metabolite.
References:
[1]. Cheng, H. and W.J. Jusko, Pharmacokinetics of reversible metabolic systems. Biopharm Drug Dispos, 1993. 14: p. 721-66.
[2]. Bertrand, J., E. Comets, and F. Mentre, Comparison of model-based tests and selection strategies to detect genetic polymorphisms influencing pharmacokinetic parameters. J Biopharm Stat, 2008. 18: p. 1084-102.
[3]. Bertrand, J., et al., Pharmacogenetics and population pharmacokinetics: impact of the design on three tests using the SAEM algorithm. J Pharmacokinet Pharmacodyn, 2009. 36: p. 317-39.
[4]. Comets, E., K. Brendel, and F. Mentre, Computing normalised prediction distribution errors to evaluate nonlinear mixed-effect models: the npde add-on package for R. Comput Methods Programs Biomed, 2008. 90: p. 154-66.
