STK11 and/or KEAP1 alterations in KRAS-mutant NSCLC treated with immune checkpoint inhibitors or chemotherapy: An episode-based Project GENIE analysis.

Abstract

8544 Background: STK11 and KEAP1 alterations are linked to aggressive biology in KRAS-mutant NSCLC and have been proposed to confer immune resistance and inferior outcomes with immune checkpoint inhibitor (ICI) though data remains inconsistent. We evaluated whether STK11 and/or KEAP1 alterations are associated with inferior overall survival (OS) and whether outcomes differ by treatment class (ICI vs chemotherapy) in a real-world clinic-genomic cohort. Methods: Using GENIE BPC NSCLC v2.0-public data, we created an episode-based cohort of KRAS-mutant NSCLC treatment regimens categorized as ICI or chemotherapy. STK11 and/or KEAP1 alteration status (single or dual) was the primary exposure. OS was measured from regimen start and analyzed with multivariable Cox models using patient-level clustering, adjusting for available covariates (age, sex, histology, tumor mutational burden [TMB], and KRAS G12C). We tested effect modification with an interaction term (STK11/KEAP1 × ICI). Sensitivity analyses included 3-, 6-, and 12-months landmark models and inverse probability of treatment weighting (IPTW). PD-L1, ECOG PS and line of therapy were not available. Results: We identified 542 treatment episodes from 293 patients (ICI = 149; chemo = 393) including 181 episodes that were STK11/KEAP1-altered. In the main cluster-robust model, STK11/KEAP1 alterations were associated with worse OS (aHR 1.68, 95% CI 1.27 - 2.24; p < 0.001). ICI (vs chemo) was associated with higher hazard ratio (aHR 1.76, 95% CI 1.42 - 2.18; p < 0.001) likely reflecting later-line use of ICI. The STK11/KEAP1 × ICI interaction was not significant as aHR 1.13, 95% CI 0.76 - 1.67; p = 0.559). In the 3-month landmark analysis STK11/KEAP1 remained adverse (aHR 1.61, 95% CI 1.19 - 2.18; p = 0.002), and the interaction estimate shifted below 1.0 (aHR 0.86, 95% CI 0.55 - 1.35; p = 0.519) suggesting a possible trend toward relatively greater ICI benefit among altered patients who survive beyond early attrition. Results were directionally consistent across 6- and 12-month landmark and IPTW analyses. Conclusions: In this real-world KRAS-mutant NSCLC cohort, STK11 and/or KEAP1 alterations were consistently associated with worse OS across treatment classes, supporting a primarily prognostic role. We did not find evidence of ICI-specific resistance by STK11/KEAP1 status, although modest interaction effects may be underpowered. Unmeasured confounding (ECOG PS, PD-L1 and line of therapy) remains a key limitation highlighting the need for larger, highly annotated datasets. Model STK11/KEAP1 aHR (95% CI); p ICI vs Chemo aHR (95% CI); p Interaction aHR (95% CI); p Main 1.68 (1.27–2.24); <0.001 1.76 (1.42–2.18); <0.001 1.13 (0.76–1.67); 0.559 3-mo Landmark 1.61 (1.19–2.18); 0.002 1.76 (1.39–2.23); <0.001 0.86 (0.55–1.35); 0.519 IPTW 1.68 (1.27–2.24); <0.001 1.76 (1.42–2.19); <0.001 1.13 (0.75–1.70); 0.546