Broader gene representation by whole-exome sequencing improves accuracy of tumor mutational burden assessment for selection of pembrolizumab immunotherapy

Although the tissue-agnostic FDA approval of pembrolizumab for tumor mutational burden (TMB)-high tumors has provided meaningful clinical benefit to patients, there remains a need to optimize TMB assessment. In this cohort study, we investigated the discordance between whole-exome sequencing (WES) and panel-based methods and evaluated their relative clinical utility in identifying patients likely to benefit from pembrolizumab. Molecularly-profiled tumors from patients treated with pembrolizumab were analyzed (N = 26,756). TMB was calculated using WES data or panels of genes (324, 523, and 648) from commercially available assays (TMB-High ≥ 10 mutations/Mb). Pembrolizumab-specific overall survival (OS) was calculated from insurance claims data (treatment start to last contact). Targeted gene panels tended to overestimate TMB and demonstrated ~ 10-15% discordance in binary TMB calls from WES, which was most pronounced for the smallest panel. Discordant cases for all three panels clustered near the TMB clinical calling threshold. WES offered improved stratification of patients for pembrolizumab treatment: in a subset of "TMB-reliant" tumor types lacking disease-specific immune checkpoint inhibitor (ICI) indications (n = 3,981), median OS was approximately 5 months longer for cases identified as WES-TMB-High but panel-TMB-Low compared to those identified as WES-TMB-Low but panel-TMB-High (p < 0.05 for 324-gene and 648-gene panels). Approximately, 10-11% of patients were potentially misclassified by panels. These findings emphasize the importance of broader gene representation for accurate TMB determination near the clinical threshold-especially in tumor types lacking disease-specific ICI indications, where tissue-agnostic MSI-High/dMMR or TMB-High labeling represents the principal on-label route to ICI therapy.