In silico validation of MARIPOSA trial outcomes in EGFR mutated NSCLC using a QSP based disease model integrated with a newly implemented lazertinib drug model

Objectives: The MARIPOSA trial compared lazertinib plus amivantamab to osimertinib in EGFR-mutated advanced NSCLC. In silico clinical trials using quantitative systems pharmacology (QSP)-based disease models can help predict trial outcomes early, potentially guiding clinical strategies. Our previously validated NSCLC QSP model, which accurately forecasted FLAURA2 results, was extended by incorporating a new lazertinib PK model developed from public data [ 1, 2], to replicate MARIPOSA outcomes. Methods: A mechanistic QSP NSCLC model was integrated with PK profiles for amivantamab, lazertinib, and osimertinib reflecting inhibitory constants and mechanistic drug–tumor interactions. Virtual cohorts replicating MARIPOSA’s design were simulated. We compared predicted and observed progression-free survival (PFS), median PFS, and hazard ratios (HRs). Kaplan–Meier curves and bootstrapped weighted log-rank tests evaluated alignment between simulated and actual data. Results: Predicted PFS for osimertinib was 19.0 months (95 % CI 17.6–20.3) versus 16.62 months observed (15.02–18.56), and for lazertinib + amivantamab was 26.7 months (23.3–30.4) versus 22.2 months observed (20.2–27.6) [ 3]. The predicted HR (0.62, 95 % CI 0.53–0.71) closely approximated the reported HR (0.70, 0.58–0.85), with overlapping confidence intervals. More than 80 % of bootstrapped tests were non-significant, indicating robust concordance with MARIPOSA outcomes. Conclusions: By accurately reproducing MARIPOSA results, this QSP-based in silico approach demonstrates its maturity and potential to inform trial design in EGFR-mutated NSCLC. The use of such a model can refine statistical hypotheses and improve the selection of target hazard ratios, optimizing patient-centered trial planning. Future incorporation of physiologically based pharmacokinetic (PBPK) modeling for lazertinib will be used to enhance predictive accuracy further, accounting for site-specific drug concentrations.