Arnaud Nativel1, Hippolyte Darré1, Perrine Masson1, Loïc Etheve1, Nicolas Ratto1, Adèle L'Hostis1, Claudio Monteiro1
1Nova In Silico
Introduction First-line EGFR tyrosine kinase inhibitors (TKIs) such as osimertinib offer improved survival in EGFR-mutated NSCLC, yet acquired resistance ultimately leads to disease progression. The Phase III FLAURA2 trial evaluated whether combining platinum-doublet chemotherapy (cisplatin + pemetrexed) with osimertinib could extend benefits by targeting tumor heterogeneity and delaying resistance. Cisplatin, however, poses pharmacological challenges, including high interpatient variability, a narrow therapeutic window with cumulative toxicities, and heterogeneous distribution to metastatic sites (particularly brain, bone, and liver). To address these complexities, we developed a physiologically based pharmacokinetic (PBPK) model of cisplatin capturing key processes (protein binding, active/inactive species, tissue partitioning). We then integrated this PBPK module within a quantitative systems pharmacology (QSP) disease model of EGFR-mutated NSCLC to link cisplatin exposure and tumor response. Here, we report on the prospective validation of this PBPK–QSP approach by simulating outcomes of the FLAURA2 study. Objectives To build a whole-body PBPK model for cisplatin in advanced EGFR-mutated NSCLC, incorporate explicit metastatic sites (liver, bone, brain) with clinically relevant distribution dynamics, and embed the final PBPK module into a mechanistic QSP platform (ISELA-V2 on Jinko) that captures tumor progression and interpatient heterogeneity. We aimed to prospectively reproduce the results of the FLAURA2 Phase III trial (osimertinib ± cisplatin/pemetrexed) and validate the predictive accuracy of this integrated model against actual clinical outcomes. Methods We constructed a multi-compartment PBPK model with specific compartments for plasma, lung, liver, bone, and brain to account for cisplatin’s tissue distribution, protein binding, and interconversion between active and inactive species. Literature data informed key parameters (tissue partition coefficients, plasma protein-binding kinetics, clearance rates), and calibration was performed against published plasma/tumor platinum concentrations from multiple clinical studies [1,2]. We then integrated this PBPK model into a QSP framework for EGFR-mutated NSCLC (ISELA-V2)[3] implemented on the Jinko platform [4]. The QSP model simulates tumor growth, metastasis evolution, and drug-specific mechanisms (TKI vs. cytotoxic), allowing exploration of treatment effects over multi-year time horizons. The resulting PBPK–QSP model was used to run a virtual trial mirroring FLAURA2 design, comparing osimertinib alone to osimertinib plus cisplatin/pemetrexed, in a large cohort of simulated patients with heterogenous baseline characteristics. Progression-free survival (PFS) was the primary endpoint. Statistical consistency (via bootstrapped log-rank tests) was assessed by comparing simulated and observed Kaplan–Meier curves. Results The cisplatin PBPK model accurately predicted both plasma and intratumoral concentrations, with deviations below 17% and 8% for plasma and tumor levels, respectively, at 30 minutes post-infusion. Pharmacokinetic metrics (Cmax, AUC) were captured within 12% and 38% of published reference values. Integrating PBPK predictions into the QSP model enabled robust simulation of combination therapy in metastatic settings. In a FLAURA2-like in silico trial, predicted PFS and response rates for the cisplatin-based arms aligned closely with real-world results. Notably, bootstrapped log-rank tests revealed no significant differences (at a=0.05) in =94% of replicates when comparing simulated and actual PFS [5]. These findings indicate a strong predictive capacity of the PBPK–QSP approach for both drug exposure and clinical benefit in EGFR-mutated NSCLC. Conclusion We successfully developed and prospectively validated a comprehensive PBPK–QSP model of cisplatin in metastatic EGFR-mutated NSCLC using the Jinko platform. By replicating the FLAURA2 trial design, the model demonstrated its ability to faithfully predict improved PFS with the addition of cisplatin/pemetrexed to osimertinib. This first-of-its-kind prospective validation highlights the potential of integrating physiologically based PK with disease-level mechanistic modeling to inform trial design and therapeutic decision-making. The validated model can be further applied to evaluate novel drug combinations, optimize dosing strategies, and generate synthetic controls, helping accelerate clinical development while minimizing patient risk.
Reference: PAGE 33 (2025) Abstr 11536 [www.page-meeting.org/?abstract=11536]
Poster: Drug/Disease Modelling - Oncology