PD-L1 expression across molecular driver subsets in non-small cell lung carcinoma: a 299-case real-world thoracic oncology cohort

Abstract Introduction PD-L1 immunohistochemistry and molecular driver testing are complementary components of non-small cell lung carcinoma (NSCLC) evaluation. We examined the distribution of PD-L1 tumor proportion score (TPS) across routinely detected molecular driver subsets in a consecutive real-world thoracic oncology cohort. Methods We retrospectively reviewed 299 consecutive NSCLC cases that underwent clinical targeted next-generation sequencing (NGS) between 2023 and 2025. Archived VENTANA SP263-stained PD-L1 slides were re-reviewed by two pathologists and categorized as TPS < 1%, 1–49%, or > = 50%. Driver-association analyses were restricted to PD-L1-tested tumors with an evaluable, mutually exclusive primary driver category; cases with indeterminate DNA-driver status were excluded. Prespecified KRAS and EGFR contrasts used reproducible 2 × 2 methods, with univariable Firth regression as a sparse-cell sensitivity analysis. Results PD-L1 results were available for 237 tumors. TPS was < 1% in 63 cases (26.6%), 1–49% in 108 cases (45.6%), and > = 50% in 66 cases (27.8%). Among 223 PD-L1-tested tumors with an evaluable, mutually exclusive primary driver category, KRAS-mutant tumors were more frequently PD-L1 high than the other evaluable categories (24/54 [44.4%] vs. 35/169 [20.7%]; OR, 3.06; 95% CI, 1.59–5.89; p = 0.001). EGFR-mutant tumors showed a lower PD-L1-high frequency, but this observation was based on only two PD-L1-high events (2/26 [7.7%] vs. 57/197 [28.9%]; OR, 0.20; 95% CI, 0.05–0.89; p = 0.019). Eleven cases with indeterminate DNA-driver status were excluded from driver-frequency and association analyses. Conclusion PD-L1 high expression was enriched among KRAS-mutant NSCLC in this diagnostic cohort, whereas EGFR-mutant tumors showed a directionally lower but sparse-data-limited PD-L1-high pattern. These findings support integrated interpretation of PD-L1 and molecular driver results in routine practice, but they should be viewed as biomarker-distribution evidence rather than proof of immunotherapy benefit or prognosis.