2014 - Alicante - Spain

PAGE 2014: Methodology - Covariate/Variability Models
Markus Krauß

Hierarchical Bayesian-PBPK modeling for physiological characterization and extrapolation of patient populations from clinical data

M. Krauß (1/2), Kai Tappe (2), Rolf Burghaus (3), Andreas Schuppert (1/2), L. Kuepfer (2), L. Goerlitz (2)

(1) Aachen Institute for Advanced Study in Computational Engineering Sciences, RWTH Aachen, Aachen, Germany; (2) Bayer Technology Services GmbH, Technology Development, Enabling Technologies, Computational Systems Biology, Leverkusen, Germany; (3) Bayer Pharma AG, Clinical Pharmacometrics, Wuppertal, Germany

Objectives: The identification of interindividual variability from clinical pharmacokinetic data using a hierarchical Bayesian approach in combination with large-scale physiologically-based pharmacokinetic modeling; and the physiological characterization and refinement of populations with potentially critical pharmacokinetic response to drug exposure.

Methods: On the one hand, physiologically-based pharmacokinetic (PBPK) modeling characterizes a mechanistic approach for a highly detailed description of the key absorption, distribution, metabolization and excretion (ADME) processes in the body. Large amounts of prior data about anthropometric and physiological parameters are integrated into PBPK models to mechanistically evaluate the processes governing pharmacokinetics (PK) behavior. On the other hand, a hierarchical statistical model was used in combination with a Bayesian approach to systematically account for parameter variability and uncertainty at the population and the individual level. To cope with the high dimensionality of the approach, Markov chain Monte Carlo approaches were used.

Results: By the consideration of the PK of several drugs in small cohorts of healthy volunteers the advantages of the combined approach of Bayesian-PBPK are illustrated. The ADME processes showed large variability. Additionally, several physiological parameters were informed by the experimental data, which allows refining the physiological database concerning such parameters. The resulting posterior distributions of all integrated parameters were then used for extrapolations of a large population of healthy individuals to evaluate the effective range of the drug as well as critical dosings.

Conclusions: The presented Bayesian-PBPK population approach systematically characterizes and quantifies interindividual variability at the population level and the parameter uncertainty at the individual level. Due to the separation of physiology, ADME processes and drug physicochemistry the translational learning of physiological parameters is possible. This allows a thorough characterization and prediction of specific properties of special populations like pediatric or diseased populations.




Reference: PAGE 23 (2014) Abstr 3151 [www.page-meeting.org/?abstract=3151]
Poster: Methodology - Covariate/Variability Models
Top