2023 - A Coruña - Spain

PAGE 2023: Drug/Disease Modelling - Oncology
Robin Michelet

Development and Application of a Minimal Physiology-based Pharmacokinetics Modelling Framework to Predict Tumor Penetration and Receptor Occupancy.

Robin Michelet1, Klas Petersson2, Marc Huisman3, C. Willemien Menke-van der Houven van Oordt4,5, Iris H.C. Miedema4,5, Andrea Thiele6, Ghazal Montasseri6, Alejandro Pérez Pitarch7, David Busse7

1qPharmetra LLC, Berlin, Germany; 2 qPharmetra LLC, Stockholm, Sweden; 3Department of Radiology and Nuclear Medicine, Amsterdam UMC, Vrije Universiteit Amsterdam; 4Department of Medical Oncology, Amsterdam UMC, Vrije Universiteit Amsterdam, 5Imaging and Biomarkers, Cancer Center Amsterdam, 6Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach an der Riß, Germany; 7Boehringer Ingelheim Pharma GmbH & Co. KG, Ingelheim am Rhein, Germany


In oncology drug development, rapid dose optimization is pivotal to allow timely patient benefit. For this, the FDA’s Project Optimus [1] is currently focusing on the integration of clinical exposure-response/safety, clinical biomarkers, and preclinical data into a framework for dose-optimization. Pharmacometric tools can support this endeavor by readily amalgamating multi-source data into quantitative frameworks. One challenge is the understanding of drug concentrations at the tumor, and the targeted receptor. Minimal physiologically-based pharmacokinetic (mPBPK) modelling approaches can predict the penetration and target-site kinetics of antitumoral drugs in-silico but lack clinical qualification. The non-invasive technique positron emission tomography (PET)-imaging based quantification of radiolabeled drug concentrations in tissues, organs and tumor lesions (biodistribution (BioD) data) can be used to bridge in-silico predictions to the patients. Here, we present the development and application of a mPBPK platform to integrate BioD data in a quantitative framework to predict tumor penetration and receptor occupancy (RO).  


BioD data for an anti-lymphocyte-activation gene 3 (LAG3) monoclonal antibody (mAb) were used for mPBPK model qualification [2]. In this clinical PET imaging study, 2 head and neck squamous cell carcinoma and 4 non-small cell lung cancer patients, who previously progressed on anti-PD-1 containing treatment, were administered a 4 mg mass dose containing [89Zr]Zr-labeled mAb (tracer dose) and PET scans were obtained 2, 96 and 144 h after. Three weeks later, patients received a 40 or 600 mg blocking dose of the compound together with the same 4 mg of [89Zr]Zr-labeled mAb tracer dose and PET scans were obtained 96 and 144 h after. At all PET scan time points plasma radioactivity was also recorded. These data were combined with the patient’s unlabeled plasma concentration-time profiles and integrated into an mPBPK framework based on an earlier published mPBPK model [3]. Physiologically relevant processes such as target recycling, drug-receptor-complex internalization and T-cell turnover were considered to describe the observations from PET uptake. Furthermore, the extravasation process into the interstitial space was modified according to two-pore theory [4]. A global sensitivity analysis (GSA) based on Sobol’s indices was conducted to find impactful parameters and assess model robustness. Predictive model performance was graphically and numerically assessed by individual plots and prediction vs observation plots, and calculation of absolute average fold error (AAFE), average fold error (AFE) and root sum of squared error (RSSE). Lastly, once an acceptable structural model was identified, key parameters were estimated by minimizing the weighted sum of AAFE, AFE and RSSE. All simulations were performed using the mrgsolve package (v1.0.4) in R/RStudio (v4.0.5/1.4.1106). 


The initial mPBPK model underpredicted tumor penetration (AAFE:1.83, AFE: 1.52, RSSE: 0.42), prompting the addition of the physiologically relevant processes drug-receptor-complex internalization, receptor recycling and T-cell turnover (AAFE:1.61, AFE: 0.972, RSSE: 0.14). The GSA indicated that for this model, the vascular reflection coefficient was a highly impactful parameter (Rank 1 based on Sobol’s index) and needed to be lowered by 27% from the default value. This supported a description of extravasation according to two-pore theory, and the model was, thus, updated according to this paradigm [4]. For this final model structure, different sets of parameters were optimized, and the average values of the best-performing estimations were considered as the final model (AAFE:1.35, AFE: 1.05, RSSE: 0.00622). This model was able to accurately predict LAG3 RO at different dose levels. 


A mPBPK framework was set up to 1) predict tumor penetration and RO of a mAb drug and 2) calibrate and qualify predictions using PET imaging BioD data. Due to the large number of parameters, different model structures and parameterizations can give similar performance, indicating a need for incorporating other sources of data and prior knowledge into the model refinement. A GSA was applied to explore the model and highlight important parameters and processes for the compound under study. Next, the developed platform will be applied to more compounds and targets to increase confidence and could then be applied to inform early development of immuno-oncological drugs.

[1] https://www.fda.gov/about-fda/oncology-center-excellence/project-optimus  
[2] Miedema, I. H. C., Huisman, M. C., Zwezerijnen, G. J. C., Thiele, A., Grempler, R., Pitarch, A. P., Vugts, D. J., de Gruijl, T. D., & van Oordt, C. W. M. D. H. (2021). Tumor uptake of the anti-LAG-3 tracer [89Zr]Zr-BI 754111 in HNSCC and NSCLC patients progressing on previous anti-PD-1 treatment. European Journal of Nuclear Medicine and Molecular Imaging, 48(SUPPL 1), S301-S301. 
[3] Lindauer, A., Valiathan, C., Mehta, K., Sriram, V., de Greef, R., Elassaiss-Schaap, J. and de Alwis, D. (2017), Translational Pharmacokinetic/Pharmacodynamic Modeling of Tumor Growth Inhibition Supports Dose-Range Selection of the Anti–PD-1 Antibody Pembrolizumab. CPT Pharmacometrics Syst. Pharmacol., 6: 11-20. https://doi.org/10.1002/psp4.12130 
[4] Li Z, Shah DK. Two-pore physiologically based pharmacokinetic model with de novo derived parameters for predicting plasma PK of different size protein therapeutics. J Pharmacokinet Pharmacodyn. 2019;46(3):305-318. doi:10.1007/s10928-019-09639-2

The authors met criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE). The authors did not receive payment related to the development of the abstract. Boehringer Ingelheim was given the opportunity to review the abstract for medical and scientific accuracy, as well as intellectual property considerations. The study was supported and funded by Boehringer Ingelheim 

Reference: PAGE 31 (2023) Abstr 10366 [www.page-meeting.org/?abstract=10366]
Oral: Drug/Disease Modelling - Oncology