Aleksandr Petrov 1,2, Elena Righetti 3,4, Charlotte Kloft 1,5, Andreas Reichel 6, Wilhelm Huisinga 1,2
1 Graduate Research Training Program PharMetrX (Berlin, Germany), 2 Institute of Mathematics, University of Potsdam (Potsdam, Germany), 3 Fondazione the Microsoft Research-University of Trento Centre for Computational and Systems Biology (COSBI) (Rovereto, Italy), 4 Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento (Trento, Italy), 5 Freie Universität Berlin, Institute of Pharmacy, Department of Clinical Pharmacy and Biochemistry (Berlin, Germany), 6 Preclinical Modeling and Simulation, Preclinical Development, Bayer, Bayer AG (Berlin, Germany)
Introduction and Objective:
Pharmacological effects are driven by drug concentrations at the target site. However, direct measurement of tissue fluid concentrations in humans is often not feasible, and plasma concentrations are commonly used as a surrogate. This assumption may be inaccurate, particularly for the central nervous system (CNS), where endothelial cells form the blood–brain barrier (BBB), which restricts drug penetration into brain tissue. Physiologically based pharmacokinetic (PBPK) modelling enables mechanistic prediction of CNS drug concentrations by integrating physiological parameters with drug-specific parameters obtained from in vitro measurements and in silico predictions [1-4]. For passively diffusing compounds with negligible or low active transport, BBB permeability, expressed as the permeability–surface area product (PS_BBB), represents one of the key determinants of CNS exposure. Multiple experimental and computational methods can be used to estimate PS_BBB, including in situ perfusion, in vitro permeability assays, and in silico models, but these approaches differ in predictive performance and applicability domain [3,4]. Moreover, the impact of PS_BBB on CNS exposure depends on its relationship with cerebral blood flow and compound-specific properties such as protein binding and ionization [5]. The objective of this study was to develop a mechanistically informed framework to support selection and interpretation of PS_BBB values for PBPK CNS modelling of passively diffusing compounds.
Methods
Reference compounds were selected to span a range of molecular weights, lipophilicity, protein binding, and ionization, while excluding strong substrates of major BBB transporters. Permeability values were collected from literature sources, including in situ brain perfusion, in vitro cell-based models (MDCK, Caco-2), in vitro PAMPA, and in silico prediction models [3,5,6]. Permeability values were translated into human PS_BBB input parameters for PBPK CNS modelling. In situ perfusion data were scaled using human brain weight, and cell-based and PAMPA permeability values were scaled using human BBB surface area. Brain extraction (E%) was calculated using the Crone–Renkin equation based on PS_BBB, cerebral blood flow, and compound-specific unbound and neutral fractions [5]. Permeability conditions were classified as permeability-limited (E% < 10%), intermediate (10–95%), or perfusion-limited (E% > 95%). CNS exposure was simulated using a PBPK CNS model consisting of a compound-specific empirical plasma model and a mechanistic CNS model describing distribution between plasma, brain, and cerebrospinal fluid (CSF) [1,3]. Simulated plasma, brain, and spinal CSF concentration–time profiles were compared with observed human CNS data. Model performance was quantified using averaged fold error (AFE) and absolute averaged fold error (AAFE), calculated over the full concentration-time profiles.
Results
Permeability data were collected for 6 compounds, with 4–5 PS_BBB estimates per compound obtained from different methods. PS_BBB values varied over several orders of magnitude within and across compounds. Brain extraction showed that variability in PS_BBB resulted in uncertainty in permeability classification for ethanol, diazepam and paracetamol, with predicted classification varying depending on PS_BBB source. In contrast, ibuprofen, indomethacin and mannitol were consistently classified as permeability-limited (E% < 10%) despite large differences in PS_BBB values. Most PS_BBB estimates resulted in CNS exposure predictions within 2-fold of observed human data. However, prediction accuracy depended on both compound and PS_BBB source. Paracetamol, ibuprofen, indomethacin and mannitol showed the greatest variability in AFE across PS_BBB estimates, indicating moderate-to-high sensitivity to permeability. In contrast, CNS exposure predictions for ethanol and diazepam were largely insensitive to PS_BBB selection, consistent with perfusion-limited behavior. Across methods, in situ perfusion and in silico perfusion-derived PS_BBB values yielded the most accurate and consistent predictions. Cell-based models showed comparable performance, while PAMPA-derived values frequently resulted in underprediction of CNS exposure. These findings enabled development of a mechanistically informed framework linking permeability source, brain extraction classification, and prediction confidence.
Conclusions
The proposed framework enables mechanism-informed selection of PS_BBB values for passively diffusing compounds with negligible or low active transport. Integration of permeability estimates with brain extraction classification allows identification of compounds for which accurate permeability estimation is critical. This framework provides practical guidance for PBPK CNS modelling and supports improved prediction of CNS drug exposure. Extension to actively transported compounds represents an important future direction.
References:
[1] Gaohua L et al. Drug Metab Pharmacokinet. 2016;31(3):224–33.
[2] Saleh M et al. J Pharmacokinet Pharmacodyn. 2021;48:725–741.
[3] Bowman C et al. Biopharm Drug Dispos. 2023;44(1):60–70.
[4] van Valkengoed et al. J Pharmacokinet Pharmacodyn. 2025;52(16).
[5] Smith QR et al. Fluid Barriers CNS. 2024;21(1):100.
[6] Avdeef A. John Wiley & Sons. 2012:575–680.
Reference: PAGE 34 (2026) Abstr 12139 [www.page-meeting.org/?abstract=12139]
Poster: Drug/Disease Modelling - Absorption & PBPK