Donato Teutonico (1), David Marchionni (1), Marc Lavielle (2,3), Laurent Nguyen (1)
(1) Sanofi, France, Translational Medicine and Early Development; modeling and simulation department (2) Inria, France (3) Ecole Polytechnique, France.
Introduction: Physiologically Based Pharmacokinetic (PBPK) modelling is a powerful tool in drug development since it allows the integration of drug specific information as well as physiological relevant parameters to predict drug concentrations [1]. One of the major limitations of this approach is to rely on average concentrations; when individual data are available, they are usually treated as naïve pooled approach without any consideration of individual information. As consequence, when these models are used for simulation, usually only the a priori known variability (e.g. physiological variability) is considered, while for unknown variability (e.g. intrinsic clearance, dissolution and transit times) an arbitrarily chosen variability is often integrated in the simulation. Open System Pharmacology (OSP) Suite is a fully developed open source PBPK platform that integrate a whole-body PBPK model allowing full flexibility to modify and manipulate such model via scripting languages in R [2].
Objectives: The objective of this exercise was to develop a framework to couple a PBPK model developed with PK-Sim (part of OSP Suite) with an extension of the SAEM algorithm [3] already available in R in the package Saemix [4] to allow the estimation of population parameter values and the corresponding inter-individual variability (IIV).
Methods: A drug SAR, with good oral absorption (capsule formulation) and a clearance partly attributed to CYP3A4, was used as test case. A PBPK model of the SAR compound was build using Phase 1 data after single and repeated dosing. This model integrates in vitro measurement of CYP3A4 metabolizing fraction (Fm). Data from a drug-drug interaction study (DDI) using itraconazole (8 subjects) were used to estimate in vivo CYP3A4 Fm based on the classical approach (naïve-pooled data) and the new proposed hybrid approach coupling the PBPK model with a population approach. As a first step, the demographic covariates available (gender, age, weight, and height) were used to generate individuals mirroring the real subjects enrolled in the DDI study. Then, the control arm with SAR compound without itraconazole was used to estimate the total intrinsic clearance for each subject and its associated interindividual variability (IIV), and the dissolution time of the capsule (population value). Itraconazole concentrations were used to estimate the CYP3A4 abundance (with IIV) as well as the capsule dissolution time (population value). Finally, the interaction arm was used to estimate the CYP3A4 contribution for each subject and consequently the individual fraction metabolized (Fm). To decrease the runtime of the estimation, a reduction in the number of calls to the PBPK model by the SAEM algorithm was also integrated. This optimization was realized by defining an adaptive grid of parameter values to be used by the algorithm and applying a linear interpolation between the grid points.
Results: The PBPK model initially developed with the in vitro Fm overpredict the DDI between SAR and itraconazole, therefore an estimation of the in vivo Fm was required. The estimated parameter values using the individual plasma concentrations (intrinsic clearance, dissolution functions and CYP3A4 contributions) using the SAEM algorithm were consistent with the parameter values used with the classical approach (naïve-pool approach) but in this case it was possible also to estimate the variability associated to clearance parameters and Fm. Moreover, also the precision of the estimated values was available with this hybrid approach. The estimated variability can then be integrated when performing simulation with the model previously developed. The improvement of the SAEM algorithm allowed to significantly decrease runtime by a factor of about 350-times, making this approach more accessible to a routine use during model building.
Conclusions: The proposed approach represents an integration of estimation methods used in population modeling to a whole-body PBPK model with the objective to estimate IIV and population parameters leveraging individual concentration data. Moreover, this integration opens the possibility of combining covariate analysis to PBPK models to explain inter-patient variability.
References:
[1] Kuepfer L., et al. (2016). Applied concepts in PBPK modeling: how to Build a PBPK/PD Model. CPT Pharmacometrics Syst Pharmacol.
[2] Lippert, J., et al. (2019). Open systems pharmacology community—an open access, open source, open science approach to modeling and simulation in pharmaceutical sciences. CPT: pharmacometrics & systems pharmacology
[3] Lavielle, M. (2014). Mixed effects models for the population approach: models, tasks, methods and tools. CRC press.
[4] https://cran.r-project.org/web/packages/saemix/index.html.
Reference: PAGE 32 (2024) Abstr 11055 [www.page-meeting.org/?abstract=11055]
Poster: Methodology - New Modelling Approaches