Clinicogenomic predictors of first-line immune checkpoint inhibitor outcomes in non-small cell lung cancer: A nationwide C-CAT cohort from Japan

Background: The efficacy of immune checkpoint inhibitors (ICIs) in non-small cell lung cancer (NSCLC) varies, underscoring the need for predictive biomarkers. Programmed death-ligand 1 (PD-L1) and tumor mutational burden (TMB) are widely used but have limited sensitivity and specificity. Increasing evidence suggests that genomic alterations shape immune responses and influence ICI outcomes. However, data from large-scale real-world cohorts, particularly in East Asian patients, remain limited. Methods: We retrospectively analyzed the Center for Cancer Genomics and Advanced Therapeutics (C-CAT) database, integrating comprehensive genomic profiling (CGP) results with clinical information from cancer patients across Japan, encompassing 99.7% of all CGP tests under the national insurance program. A total of 1,629 patients with stage IV NSCLC received first-line ICI-based therapy between 2019 and 2025. The primary endpoint was time to treatment failure (TTF); the secondary endpoint was overall survival (OS). We used age-stratified Cox models adjusted for covariates. Genomic covariates included KRAS, KEAP1 , and STK11 , with additional genes from an exploratory screening. To clarify interaction effects, we further expressed results as wild-type-referenced hazard ratios for KRAS -only, KEAP1 -only, STK11 -only, and their co-mutations ( KRAS - KEAP1 and KRAS - STK11 ). Sensitivity analyses included a formalin-fixed paraffin-embedded (FFPE)-only subset with TMB available (n = 1,095, categorized as <10 vs ≥10 mut/Mb), and models also adjusting for ECOG performance status (0, 1, ≥ 2). Results: BRAF mutation was independently associated with longer TTF (HR 0.77, 95% CI 0.59-1.00; p = 0.049), whereas KRAS , KEAP1 , and STK11 alone showed no significant associations (HRs 1.21, 1.08, and 1.09; all p > 0.21). A significant KRAS - KEAP1 interaction was associated with shorter TTF (HR 1.89, 95% CI 1.06-3.36; p = 0.032). Compared with PD-L1 ≥ 50%, PD-L1 < 1% was associated with shorter TTF (HR 1.17; p = 0.041). In a four-level comparison versus wild-type, KRAS -only and KEAP1 -only were not significant (HRs 1.09 and 1.21), whereas the KRAS - KEAP1 co-mutation was associated with higher risk of treatment failure (HR 2.48, 95% CI 1.50-4.10; p < 0.001). Results were consistent in FFPE-only subsets with TMB data. KEAP1 ( p < 0.001) and STK11 ( p = 0.001) were associated with worse OS, while BRAF showed no significant difference ( p = 0.840). Conclusions: In first-line ICI-treated stage IV NSCLC, KRAS - KEAP1 co-mutation, not KRAS or KEAP1 alone, identifies patients at high risk of early failure, whereas BRAF was associated with longer TTF. These findings highlight the importance of co-mutation profiling in guiding treatment selection and warrant prospective validation with integrative models incorporating PD-L1 and TMB.