Michaël Duruisseaux1, Perrine Masson2, Arnaud Nativel2, Nicolas Gerard3, Jacques Cadranel4, Aurélie Swalduz5, Matthieu Coudron2, Guillaume Bouchard2, Riad Kahoul2, Evgueni Jacob2, Jim Bosley2, Jean-Louis Palgen2, Adèle L'Hostis2, Claudio Monteiro2
1Respiratory Department and Early Phase, Louis Pradel Hospital, Hospices Civils de Lyon Cancer Institute, 2Nova Insilico, 3Institut Curie, 4Université Paris Sorbonne, Service de Pneumologie et oncologie thoracique, Hôpital Tenon, APHP. Sorbonne Université, 5Department of Medical Oncology, Centre Léon Bérard
Introduction Targeted therapies have significantly improved outcomes for patients with EGFR-mutated non-small cell lung cancer (NSCLC), yet acquired resistance limits their long-term efficacy. Combination treatments offer a potential solution, but determining the most effective regimen remains a challenge due to interpatient variability and to the dynamic nature of resistance mechanisms. In silico clinical trials, based on mechanistic computational models, provide a promising alternative for evaluating treatment efficacy before real-world trials. Objective To demonstrate the predictive potential of a robust mechanistic model for EGFR-mutated advanced lung adenocarcinoma (aLUAD) in estimating the efficacy of treatments and treatment combinations before real-world clinical trials. Methods We extended a previously published computational mechanistic model of EGFR-mutated aLUAD [1] by integrating physiologically based pharmacokinetic (PBPK) models for platinum-based chemotherapy doublet (cisplatin and pemetrexed), enabling a comprehensive evaluation of combination therapies. This mechanistic framework captures causal relationships between biological pathways, drug mechanisms, and tumor progression, ensuring accurate modeling of pharmacodynamic effects. We calibrated the model with preclinical experimental data, including in vitro survivability assays and in vivo xenograft tumor growth inhibition studies. We accounted for interpatient variability by generating virtual patient cohorts, parameterized based on distributions informed by existing clinical trials. Using this set-up, we independently and prospectively simulated the outcomes of the FLAURA2 randomized phase III clinical trial [2], which evaluated osimertinib against its combination with platinum-based chemotherapy. For this, we generated virtual twin cohorts by replicating trial protocols and leveraging publicly available patient data. Results The in silico trial predictions were generated and publicly shared [3] before the trial sponsor released the FLAURA2 results. Simulated outcomes closely matched clinical trial data, with overlapping confidence intervals, similar hazard ratios (FLAURA2: 0.62 [0.49, 0.79] vs simulations: 0.602 [0.49, 0.736]), and median survival times between real and simulated cohorts (combination therapy arm: 25.5 [24.7,NC] for FLAURA2 vs 25.9 [25.1, 27.1] for simulations). The Kaplan-Meier curves of the simulated and actual trials exhibited comparable trajectories. To assess statistical concordance, we conducted bootstrapped weighted log-rank tests, comparing simulated and observed survival curves for the FLAURA2 trial arms. Between 94–98% of these tests were statistically non-significant (a=0.05), confirming that the in silico predictions were highly consistent with real-world trial outcomes. Conclusions This study provides one of the first validated examples of a mechanistic in silico clinical trial accurately predicting real-world trial results. These findings suggest that mechanistic computational models could serve as powerful tools for clinical trial optimization, hypothesis generation, and treatment strategy evaluation. In silico trials may help overcome challenges associated with comparator arms in single-arm trials and could become a new standard for defining statistical hypotheses in drug development, particularly for EGFR-mutated aLUAD and other oncology indications.
[1} Darré, H. et al. (2024). Comparing the Efficacy of Two Generations of EGFR-TKIs: An Integrated Drug–Disease Mechanistic Model Approach in EGFR-Mutated Lung Adenocarcinoma. Biomedicines. DOI: https://doi.org/10.3390/biomedicines12030704 [2] Planchard, D. et al. (2023). Osimertinib with or without Chemotherapy in EGFR-Mutated Advanced NSCLC. The New England Journal of Medicine. DOI: 10.1056/NEJMoa2306434 [3] Duruisseaux, M.[@MDuruisseaux]. (2023, September 8). FLAURA2 will be presented on Monday,We share an in silico prediction of FLAURA2 outcomes, using jinko[…]. X (Ex twitter). https://x.com/MDuruisseaux/status/1700257878551228503?s=20
Reference: PAGE 33 (2025) Abstr 11588 [www.page-meeting.org/?abstract=11588]
Poster: Drug/Disease Modelling - Oncology