Miao-Chan Huang1, Julia Macente1, Sofie Heylen1, Chen Ning1, Kristof De Vos1, Neel Deferm2, Pieter Annaert1,3
1Drug Delivery and Disposition, Department of Pharmaceutical and Pharmacological Sciences, KU Leuven, 2Simcyp Division, Certara UK Limited, 3BioNotus GCV
[Introduction and objectives] Bosentan is the first approved oral medication for treating pulmonary arterial hypertension. It enters the liver by passive diffusion and OATP1B1/1B3-mediated active transport [1]. Following a single intravenous (IV) administration, bosentan was mainly eliminated via metabolism as hydroxy bosentan (OH-BOS), desmethyl bosentan (DMB), and hydroxy desmethyl bosentan (OH-DMB), and these metabolites were almost exclusively excreted to feces [2]. Owing to the considerable occurrence of abnormal liver function tests in clinical studies, the labeling of bosentan warns about the risk of drug-induced liver injury, and regular monitoring of liver function is required when using bosentan [3]. Accumulating endeavors were devoted to identifying the underlying mechanism of bosentan-induced liver injury. However, since the hepatic level of bosentan is not available, the plasma level of bosentan was used as a surrogate of clinically relevant exposure in these mechanistic toxicity studies. The scalar for unbound concentration between plasma and liver (Kpuu), estimated by mechanistic approach, for OATP1B1/1B3 substrates was generally greater than one [4]. This implied that using the unbound plasma level of these compounds as a proxy could run the risk of underestimating the toxicity or setting the wrong threshold for key liver injury pathways. Physiologically-based pharmacokinetic (PBPK) modeling and simulation has emerged as a valuable tool to quantitatively envision local exposure to drugs of interest [5,6]. The nature of PBPK modeling takes into account the physiology of the organism and the disposition mechanism of the modeled entity. Additionally, PBPK modeling following a middle-out approach allows for identifying disposition-associated parameters, which compensates for the in vitro-in vivo difference and the simplified disposition pathway implemented in the model. However, it requires a careful and thoughtful strategy in utilizing clinical data for identifying multiple hepatic disposition processes and subsequently predicting hepatic exposure. In this context, we hypothesized that using the excretion data from the clinical mass-balance study in addition to systemic plasma data would assist anchoring the metabolism, thereby accurately estimating the active uptake process of bosentan. In this study, we developed a PBPK model of bosentan with this hypothesis attempting to capture bosentan’s disposition in the liver. The model-derived hepatic disposition was evaluated with hepatic clearance from healthy adults and with perpetrators affecting the transport and/or metabolism of bosentan. We also applied the developed model to provide the bosentan Kpuu for liver matrices. [Methods] The PBPK model of bosentan was developed using PK-Sim® and MoBi® (Open System Pharmacology; Version 11) following a middle-out approach. A simplified elimination pathway of bosentan was implemented in the model. The active hepatic uptake was mediated by OATP1B1/1B3, the major metabolic routes were catalyzed by CYP3A4 and/or CYP2C9 to form OH-BOS and DMB, the minor metabolic pathway was described with a first-order hepatic clearance, and the renal excretion of unchanged bosentan was assumed to occur exclusively via glomerular filtration. Additionally, we included the CYP3A4 autoinduction and the autoinhibition of OATP1B1/1B3 in the model. OH-BOS and DMB were assumed to be not metabolized further, so both also included a contribution of the secondary metabolite OH-DMB. Clinical data of bosentan in healthy adults from 12 clinical studies were collected from the literature for model training, evaluation, and drug-drug interaction (DDI) verification. These data included plasma concentration-time profiles of bosentan and the fraction of bosentan dose excreted unchanged or as individual metabolites. The hepatic disposition by active uptake and metabolism was calibrated using data following IV administration. Data following single oral dose and multiple doses were used for capturing the oral absorption and steady-state pharmacokinetics, respectively. The predictive performance of the model was evaluated by comparing the model prediction against the observation in terms of maximum concentration (Cmax) and area under the curve (AUC) with the prediction-to-observation ratio (FE) within twofold as an acceptance criterium. We calculated the hepatic intrinsic clearance (CLint,h) of bosentan using the hepatic disposition derived from the model (Equation 1) and compared it with the one derived from the clinical study [7] by retrograded calculation where the liver was deemed as a well-stirred compartment. The hepatic disposition in the model was also verified with the DDI between bosentan and ketoconazole or rifampicin. CL_int,h=(CL_act,up+PS_dif,inf)×(CL_int,met)/(PS_dif,eff+CL_int,met)(Equation 1) The developed PBPK model was applied to simulate the unbound level of bosentan in plasma (Cu,plasma) and liver matrices (i.e., tissue, interstitial space, intracellular space)(Cu,liver) following various dosing regimens (i.e., once daily regimen with 62.5 mg, 125 mg, 500 mg, and twice daily regimen with 500 mg for 8 days). The simulated Cu,liver and Cu,plasma at the matched time were utilized to derive the Kpuu for the liver matrices . Kp_uu=C_u,liver/C_u,plasma (Equation 2) [Results] Our model exhibited consistently good predictive performance across systemic plasma and excretion data, with all comparisons giving the AUC FE and Cmax FE within twofold. The model-derived CLint,h was within 1.25 fold of the retrograde calculated clinical value (440 L/h). Meanwhile, the FE of DDI effect with ketoconazole and rifampicin was within 1.5 fold. The hepatic simulations by the developed model indicated the maximal unbound level of bosentan was the highest in the hepatocytes, followed by liver tissue, while the unbound level in plasma overlapped with hepatic interstitial space. This finding echoed the concerns about the inappropriateness of using exposure in plasma as a proxy of the one in the liver. Therefore, we provided the Kpuu at steady state for different liver matrices to convert exposure in plasma to liver. At the time of plasma Cmax (Tmax), the Kpuu was around 1.60, 1.50, and 1.00 for hepatocytes, liver tissue, and hepatic interstitial spaces, respectively. The simulated maximal Kpuu reached maximum before the plasma Tmax; the maximal Kpuu was close to 3.00 for hepatocytes and liver tissue, while it was around 1.90 for hepatic interstitial space. [Conclusion] The developed PBPK model well described bosentan’s concentrations in plasma and excretion data. Furthermore, the CLint,h comparison and the DDI verification suggested this model possessed adequate confidence in capturing the hepatic disposition of bosentan. The simulated Kpuu can be applied to convert the concentrations in plasma to the liver matrices so as to be used in the mechanistic studies as the relevant clinical concentrations.
[1] Treiber et al. Drug Metab Dispos. 2007; 35(8): 1400-1407. [2] Weber et al. Drug Metab Dispos. 1999; 27(7): 810-815. [3] USFDA. Drug approval package: Tracleer (bosentan) tablets. Application No. 21-290. [4] Varma et al. J Pharmacol Exp Ther. 2014; 351(1): 214-223. [5] Hanke et al. Pharm Res. 2021; 38: 1645-1661. [6] Garcia et al. CPT Pharmacometrics Syst Pharmacol. 2024; 13(6): 1029-1043. [7] Weber et al. Clin Pharmacol Ther. 1996; 60(2): 124-137.
Reference: PAGE 33 (2025) Abstr 11473 [www.page-meeting.org/?abstract=11473]
Poster: Drug/Disease Modelling - Other Topics