Zhilan Huang 1, Frederike Lentz 2, Ulrich Jaehde 1, Ana RodrÃguez-Báez 1,2
1 Department of Clinical Pharmacy, Institute of Pharmacy, University of Bonn (Bonn, Germany), 2 Biostatistics and Special Pharmacokinetics Unit/Research Division, Federal Institute for Drugs and Medical Devices (BfArM) (Bonn, Germany)
Introduction: PBPK models account for developmental differences in physiology and pathology and thereby support precision dosing in paediatric populations. However, parameter uncertainty, as well as insufficient model qualification remain major barriers to their broader application in clinical practice [1]. Consequently, careful evaluation of ontogeny functions and other age-dependent physiological changes within the modeling framework is required to ensure predictive performance [1]. Moreover, inconsistencies exist among published CYP3A4 ontogeny profiles, with some suggesting that CYP3A4 enzymatic activity exceeds adult levels during early life, whereas others describe a gradual, monotonic increase towards adult activity. Therefore, accurate characterization of enzyme developmental changes is essential for reliable prediction of clearance in paediatric PBPK models. Currently, no guidance exists on selecting appropriate ontogeny functions [1], and the original data used to develop ontogeny functions are derived from diverse sources.
Objectives:
• To evaluate the impact of different ontogeny profiles on clearance prediction in PBPK modeling.
• To investigate differences between developmental ontogeny approaches (i.e., in vitro-derived versus in vivo-derived) on model predictive performance.
Methods: A published midazolam adult PBPK model was extrapolated to paediatric populations (0-2 years) using PK-Sim®. Simulations were conducted based on individual patient characteristics corresponding to the external observed pharmacokinetic data. A total of 34 patients were included for clearance estimation, and 7 patients were used for concentration–time profile analysis. To quantify predictive performance across different ontogeny profiles, prediction errors (PE) and mean prediction errors (MPE) of clearance were calculated. Additionally, the model performance for concentration–time profiles was assessed using goodness-of-fit diagnostics, average absolute fold error (AAFE), and geometric mean fold error (GMFE).
Results: A total of five ontogeny profiles were identified and categorized into three developmental approaches based on their data sources: in vitro–, combined in vitro/ in vivo, and in vivo–ontogeny functions. Among them, the model incorporating the in vivo–ontogeny function proposed by Upreti and Wahlstrom [2] demonstrated the best overall predictive performance, with a MPE of clearance of +51% in term neonates, −13% in infants A (1–6 months), −29% in infants B (6 months-1 year) and −35% in toddlers. The PE values remained largely within the ±50% range. In addition, the predicted concentration–time profiles were generally within the standard PBPK criteria, with a 2-fold deviation from the observed data. Clear differences were observed in the results among the various ontogeny profiles, as well as between the different developmental approaches of ontogeny. For example, in the infant group, MPE ranged from −13% to −71%, depending on the selected ontogeny profile.
Three preterm ontogeny profiles were also evaluated in our study. The ontogeny profile obtained from the PK-Sim® ontogeny database [3] showed the best predictive performance, with a MPE of 20% in predicted clearance and the least AAFE (1.10-1.55) in concentration-time data results.
Both clearance and concentration results revealed an age-dependent bias in model predictions. Clearance was systematically underpredicted in toddlers and infants, whereas overprediction was observed in term and preterm neonates.
Conclusions: A systematic comparison of ontogeny profiles was conducted to provide evidence guiding the selection of appropriate ontogeny functions in pediatric PBPK models. Ontogeny functions derived from in vivo data describe CYP3A4 enzymatic maturation more adequately within the PBPK framework. The observed age-dependent bias suggests potential structural limitations of the current midazolam PBPK model or possible mischaracterization of CYP3A4 enzymatic activity within the implemented ontogeny function. In summary, further clinical data are required to refine ontogeny functions in young pediatric patient populations and to strengthen confidence in the clinical application of PBPK models.
References:
[1] Cleary Y et al. CPT Pharmacomet Syst Pharmacol 2026;15:e70174.
[2] Upreti VV and Wahlstrom JL. J Clin Pharmacol 2016;56:266–83.
[3] https://github.com/Open-Systems-Pharmacology/OSPSuite.Documentation/blob/master/PK-Sim%20Ontogeny%20Database%20Version%207.3.pdf
Reference: PAGE 34 (2026) Abstr 12137 [www.page-meeting.org/?abstract=12137]
Poster: Drug/Disease Modelling - Absorption & PBPK