(W1130-02-12) Virtual Bioequivalence (VBE) Outcome May Depend on Choice of Dissolution Inputs in Physiologically Based Pharmacokinetic Modelling

Purpose: Physiologically based pharmacokinetic (PBPK) models can be used to conduct virtual bioequivalence (VBE) studies. There are different ways to input dissolution parameters in PBPK models. Depending on the dissolution inputs it may or may not be possible to incorporate the impact of gut physiological parameters on drug dissolution and absorption [1]. For example, it is simpler to directly enter in vitro biopredictive dissolution profiles into PBPK, but this means the dissolution in the gastrointestinal tract is already determined and the individuals’ gut physiology is not going to affect the rate or extent of the drug product dissolution. This may not be the case where a mechanistic dissolution model is used. Many small molecule tyrosine kinase inhibitors (TKI) are weakly basic with significant pH-dependent solubility [2]. Therefore, physiological parameters such as stomach and small intestine pH and transit time may affect the bioavailability of these drugs. This work aims at investigating the potential impact of dissolution input options on the VBE outcome of two TKIs.

Methods: Ibrutinib and Crizotinib PBPK default models from the Simcyp Simulator V23 (Certara UK) are used for this study. The Advanced Dissolution Absorption and Metabolism (ADAM) model is selected for both drugs. Further, the Diffusion Layer Model (DLM) is selected that can handle within and between subject variability of physiology and formulation. Then two new PBPK models (test) for each drug were developed where the dissolution profiles are determined and entered instead of using the DLM model. The dissolution profiles were manually fitted to the give very close dissolution profiles, Cmax and AUC values when a population representative subject is simulated. Next, the Virtual Bioequivalence (VBE) module was used to evaluate the bioequivalence of the reference and test models for each of these two drugs. A crossover BE study of 2 treatments (T1: reference and T2: test), two periods and 2 sequences (2T2P2S) of T1T2/T2T1 was simulated. For each VBE simulation 10 replicates of 12 subjects from the healthy volunteer population with age range 20-50 years and 50% females were simulated. The duration of simulation for Ibrutinib and Crizotinib was 48 and 144 hours respectively. To simulate an ideal situation withing subject variability (WSV) was not considered, meaning the same physiological values were used in both sequences.

Results: The Tmax, Cmax and AUC values for Ibrutinib (reference using the DLM model) for a population representative were respectively 1.44 (h), 0.035 (mg/L) and 0.275 (mg/L.h) and for Ibrutinib (test model using the dissolution profile input) were 1.44 (h), 0.035 (mg/L) and 0.274 (mg/L.h). The VBE simulation results for Ibrutinib showed that for Cmax only 4 out of 10 replicates were BE while for AUC and AUCinf 5 out of 10 replicates were within the BE ranges. The remaining replicates were undecided meaning they were not BE [2]. The Tmax, Cmax and AUC values for Crizotinib (reference model using the DLM model) for a population representative were respectively 2.16 (h), 0.112 (mg/L) and 2.26 (mg/L.h) and for Crizotinib (test model using the dissolution profile input) were 2.16 (h), 0.112 (mg/L) and 2.26 (mg/L.h). The VBE simulation results for Crizotinib showed that for the three PK parameters (Cmax, AUC, AUCinf) all 10 replicates were well within the BE ranges.

Conclusion: Through simulation it is shown that the outcome of VBE studies for two TKIs may depend on the dissolution inputs to PBPK models. Six out of 10 replicates for Ibrutinib failed BE test when a dissolution profile was used instead of the diffusion layer model. However, for the Crizotinib VBE study whether a dissolution profile or the DLM was used there was BE for all 10 replicates. This demonstrates that the outcome of VBE studies may depend on PBPK model inputs, specifically for cases where the drug dissolution is sensitive to the gastrointestinal pH, transit time or numerous other physiological parameters.

References: [1] Jamei et al., (2020) European Journal of Pharmaceutics and Biopharmaceutics 155:55-68.
[2] Williams et al., (2018) Mol. Pharmaceutics 15:5678-5696.
[3] Bego et al., (2022) American Association of Pharmaceutical Scientists Journal 24:21.

Acknowledgements: Disclaimer: This abstract is presented at the Population Approach Group Eurpe (PAGE), 26-28 June 2024.

Figure 1 – Choice of mechanistic vs empirical models affect how much of physiological variability can be reproduced in the simulation.

Figure 2 –The dissolution input options (Test vs Reference) in PBPK model where the impact of model input is investigated using VBE for two model drugs.