Clinicogenomic determinants of early progression on first-line osimertinib in <i>EGFR</i> -mutant NSCLC.

Abstract

8634 Background: With the recent approvals of the FLAURA2 and MARIPOSA regimens in EGFR -mutant non-small cell lung cancer (NSCLC), identifying clinicogenomic features predictive of suboptimal outcomes with first-line osimertinib monotherapy is critical in guiding upfront treatment selections. Methods: This is a multicenter retrospective study enrolling patients (pts) with advanced NSCLC harboring common EGFR mutations (exon 19 deletions [ex19del], L858R) treated with first-line osimertinib monotherapy across 13 centers (Clinical cohort). Comprehensive baseline genomic data on tumor tissue were available for DFCI and MSKCC (Genomic cohort). The association of clinicogenomic features with real-world progression-free survival (rwPFS) was investigated. Time-dependent discrimination for very-early progression (rwPFS &lt; 6 months) was evaluated using Receiver Operating Characteristic (ROC) curve-based Area Under the Curve (AUC). Variable importance was assessed by delta (Δ) AUC after covariate exclusion. Moreover, very-early progression was analyzed with a Cox model censored at 6 months, reporting adjusted hazard ratios (aHR). Results: A total of 1488 pts were enrolled in the clinical cohort; at median follow-up of 43.3 months, median rwPFS was 16.3 months (95%CI 15.3-17.4), median overall survival was 36.7 months (95%CI 34.8-39.4). The strongest predictor of very-early rwPFS in a multivariable model, including, age, sex, EGFR mutations, PD-L1 tumor proportion score (TPS), concurrent TP53 mutations, metastatic sites (brain, bone, liver), performance status, was PD-L1 TPS (aHR 2.75 for ≥50% versus 0%, p &lt; 0.0001), with the largest decrease in the AUC when excluded from the model (ΔAUC = 0.17). Looking at genomic features in the genomic cohort, at median follow up of 38.2 months, loss-of-function mutations in KMT2D (HR 7.1, p &lt; 0.001), TP53 (HR 1.4, p = 0.01), RB1 (HR 1.7, p = 0.01), RBM10 (HR 1.6, p = 0.01), CREBBP (HR 2.7, p = 0.02), ATM (HR 2.1, p = 0.04 ), predicted shorter rwPFS. Next, we investigated the role of concurrent tumor suppressor gene ( TSG ) alterations beyond TP53 , including in this cathegory those with a p &lt; 0.1 for rwPFS and mutated in at least 5 cases (RBM10, CREBBP, ATM, RB1, KMT2D, TSC2, PTEN). TSG MUT had shorter rwPFS compared to TSG WT (aHR 1.5, p = 0.01), specifically in L858R subgroup (aHR 1.7, p = 0.02), while TP53 MUT had shorter rwPFS compared to TP53 WT in ex19del only (aHR 1.7, p = 0.01). Combining TP53 with TSG, pts with TP53 MUT plus at least one TSG MUT and pts with TP53 WT -TSG MUT had shorter rwPFS compared to TP53 WT -TSG WT and TP53 MUT -TSG WT (10.8, 13.1, 21.9, 18.1 months, respectively, p = 0.004). In the genomic cohort, PD-L1 TPS was the strongest predictor of very-early rwPFS (ΔAUC = 0.16), followed by TP53 combined with TSG (ΔAUC = 0.08). Conclusions: In this multicenter real-world cohort, baseline PD-L1 and TP53 combined with TSG predict very-early progression to first-line Osimertinib.