Tumor purity as confounder for the determination of tumor mutational burden (TMB).

Abstract

e14656 Background: TMB determination is a prerequisite to stratify advanced cancer patients for pembrolizumab based on the histology-agnostic FDA approval. The accuracy of TMB measurement is challenged by low tumor purity resulting in a lower number of detected somatic variants for a fixed threshold of variant allele frequency (VAF) and a lower measured TMB level as a consequence. Methods: We derived a mathematical formula describing the influence of tumor purity on the VAF of somatic variants, in which the DNA copy number of the corresponding region of the tumor genome is a parameter. We simulated the effect of tumor purity in a cohort of 4,900 tumors from the TCGA project with tumor purity of at least 80%. TMB was calculated as the number of missense mutations with VAF greater or equal than 5%. Tumors were classified as TMB-high when harboring at least 199 missense mutations. This cutpoint corresponds to 10 mut/Mb as derived in the TMB bridging study. We compared tumors with a simulated tumor purity of 70%, 60%, ..., and 10% with tumors with a tumor purity of 80%, which served as reference. Results: The median of TMB in the pan-cancer cohort was reduced by 2%, 10%, and 32% for tumor purity of 60%, 40%, and 20% compared to the reference of 80%. For the 10% tumors with the worst estimation, TMB was reduced even by 12%, 35%, and 77% or more, respectively. In the pan-cancer cohort, the sensitivity to detect high TMB was 98%, 90%, and 62% for tumor purity of 60%, 40%, and 20%, respectively. For head and neck cancer, lung adenocarcinoma, lung squamous cell carcinoma, and cutaneous melanoma, the sensitivity to detect high TMB was 80%, 89%, 92%, and 91%, respectively, for tumors of tumor purity 40%. Conclusions: High tumor purity is critical for accurate determination of the TMB and to reach a high sensitivity concerning the selection of patients for immunotherapy. As shown by the simulations, the sensitivity to detect high TMB is close to 100% for tumor purity of at least 60%, while a substantial proportion of patients that could receive immunotherapy is missed for tumors with purity of 40% or less. Potential ways out of the TMB underestimation include increasing the sequencing coverage and lowering the VAF threshold for mutation calling or using clonal TMB with an adapted cutpoint instead of total TMB as biomarker for immunotherapy guidance.