Background: Immune checkpoint inhibitors (ICIs) targeting anti-PD-1/L1 have expanded the treatment landscape against cancers, but are only effective in a subset of patients across all cancer types. Tumor mutation burden (TMB) is postulated to be a generic determinant of ICI dependent-tumor rejection. Here, we clarify the association between TMB and survival outcomes among microsatellite-stable cancers in a real-world clinicogenomic cohort including 70,698 patients distributed across 27 histologies.Methods: Patients were segregated into mutually exclusive anti-PD-1/L1 (N=14,736, immune checkpoint inhibitor, ICI) and non-ICI cohorts (N=55,962) if they had received pembrolizumab, nivolumab, atezolizumab, durvalumab, avelumab, or cemiplimab, and non-ICI therapies, respectively, during the course of treatment for individual cancer types. Patients treated with chemo-immunotherapy combination regimens were included in the ICI cohorts, but anti-CTLA-4 agents were excluded from both cohorts. For both ICI and non-ICI cohorts, patients were reiteratively categorized as TMB-high or -low using increasing TMB thresholds up to the 90th percentile TMB for individual cancer types. Outcomes were then assessed using Cox regression analyses of overall survival (OS).Results: Across 27 histologies, 12 cancer types had at least one TMB cutoff associated with survival benefit upon ICIs (melanoma, NSCLC, SCLC, bladder, cervical, colorectal, gastric, ovarian, head and neck, vulvar cancers, as well as carcinoma of unknown primary, and sarcoma), and these associations were generally conserved across sex and age groups, although not necessarily significant in all cases. In two cancer types (cholangiocarcinoma and neuroendocrine tumors), TMB cutoffs were associated with worse OS with ICIs. The lowest TMB cutoff in which a survival benefit could be detected with ICIs ranged from 2 mut/Mb in melanoma to 9 mut/Mb in colorectal cancer. TMB thresholds with the greatest predictive value (defined as the lowest survival hazard ratio) were frequently near the highest TMB examined for individual cancer types, suggesting that TMB may scale with ICI benefit.
Conclusion: TMB was associated with survival benefit or detriment depending on tissue and treatment context. Detectable survival benefits were noted over a broad range of TMB cutoffs across cancer types, and a dose-dependent relationship between TMB and outcomes was observed in a subset of cancers. These results may have implications for cancer-agnostic and universal TMB cutoffs to guide the use of anti-PD-1/L1 therapies, and underline the importance of tissue context in the clinical development of ICI biomarkers.
