Importance: Tumor mutational burden (TMB) greater than or equal to 10 mutations per megabase (mut/Mb) has received agnostic FDA approval for pembrolizumab. However, this TMB cut-off alone is not a complete predictor of overall survival (OS) in patients treated with immune checkpoint inhibitors (ICIs).
Objective. To analyze the influence of the molecular profile in patients with TMB ≥ 10mut/Mb treated with ICIs.
Design, Setting, and Participants. This post-hoc analysis evaluated the clinical and molecular features of tumor-normal pairs from 1,661 patients with solid tumors sequenced using the MSK-IMPACT assay and treated with ICIs.
Main Outcomes and Measures. We performed OS analysis and compared the results for TMB thresholds of ≥ 10, ≥ 20, and < 10 mut/Mb. For a TMB ≥ 10mut/Mb, we assessed OS according to mutational status. For all genes exhibiting a correlation with OS (P < 0.05), we conducted a Cox multivariate analysis stratified by median TMB, sex, median age, microsatellite instability (MSI) status, and histology.
Results. After a maximum follow-up of 80 months, a total of 1,661 patients were investigated, and survival to ICIs increased with higher TMB cut-offs. The median OS was 42 months for TMB ≥ 10 or 20mut/Mb, and 15 months for TMB < 10 mut/Mb (P < 0.005). Patients harboring a TMB ≥ 10 (N=488, 29%) were further stratified by their somatic mutation profile in key cancer genes. When only genes with n ≥ 5 mutations were considered, mutations in E2F3 or STK11 were correlated with worse OS, and those with mutations in NTRK3, PTPRD, RNF43, TENT5C, TET1 or ZFHX3 were correlated with better OS in TMB-high patients receiving ICIs compared to wild-type patients. These associations were confirmed by univariate and multivariate analyses (P < 0.05). MSI status and clinical features, including age, sex, and histology (except for melanoma), failed to predict outcomes to ICIs in patients with high TMB.
Conclusion and relevance. The findings suggest that combining TMB information and mutation profiles in key cancer genes can be used to better qualify patients for ICI treatment and predict their OS.