Background Pembrolizumab, with or without chemotherapy, is NCCN guideline-recommended treatment for NSCLC cancer patients depending on tumor PD-L1 status by IHC. PD-L1 IHC provides guidance for treatment selection for response, but does not accurately predict survival benefit from pembrolizumab. Emerging evidence suggests TMB and other genomic markers (KRAS, STK11, TP53 mutations), may have clinical utility for predicting survival benefit.

Methods We identified a cohort of metastatic EGFR/ALK wild type NSCLC patients (n=116) whose tumors underwent comprehensive profiling (June 2017-March 2019) for genomic variants, TMB and PD-L1 IHC 22C3 prior to selection of pembrolizumab (n=43), pembrolizumab + chemotherapy (n=41), or chemotherapy only (excluding subsequent targeted therapy or immunotherapy) (n=32) at Roswell Park Comprehensive Cancer Center, with at least one year of follow up. TMB was assessed using a 1.75 Mb capture of 409 oncogenes with full exon coverage (DNA-Seq), with ‘high’ TMB interpreted as ≥10 mutations/Mb. Electronic pharmacy records were curated to create pre and post-test treatment histories for each patient. Cox regression analysis evaluated OS with pembrolizumab monotherapy or pembrolizumab + chemotherapy vs chemotherapy only, adjusting for covariates including oncogenic driver mutations, TMB, PD-L1 IHC demographics, clinicopathologic characteristics, prior treatment, and performance status. Using the same model, we then assessed overall survival for each treatment group by TMB, KRAS, STK11, and TP53 mutant status.

Results Overall, 47% of tumors were PD-L1 high, 47% TMB high, 34% KRAS mutant (codons 12, 13, 60, 61), 52% TP53 mutant and 16% STK11 mutant. As expected, pembrolizumab with or without chemotherapy significantly improved overall survival (OS) compared to chemotherapy alone; with TMB, smoking, and ECOG status identified as significant covariates. PD-L1 IHC status was not associated with OS for any treatment. TMB high status was significant for OS benefit with pembrolizumab either as monotherapy [HR=0.02; CI=0.01–0.40; p=0.01] or in combination with chemotherapy [HR=0.20; CI=0.04–0.95; p=0.04]. KRAS mutant status was independently significant for OS benefit from pembrolizumab + chemotherapy [HR=0.01; CI=0.01–0.79; p=0.04] but not for pembrolizumab monotherapy or chemotherapy alone. Among patients who received pembrolizumab monotherapy, there was a trend toward increased risk of death in those with STK11 mutations [HR=17.54; CI=0.35–1,000; p=0.15], whereas TP53 mutant status trended toward survival benefit [HR=0.18; CI=0.02–1.53; p=0.11].

Conclusions Data comparing pembrolizumab treatments with chemotherapy and independent marker associations suggest TMB has predictive power for determining overall survival benefit from pembrolizumab, while KRAS, STK11, and TP53 mutational status demonstrated potential prognostic relevance for NSCLC.