TMB has emerged as a predictive biomarker of response to immune checkpoint inhibitors. CheckMate 227 demonstrated that patients with non-small cell lung cancer (NSCLC) with TMB ≥10 mutations/megabase derived enhanced benefit from first-line treatment with nivolumab + ipilimumab vs chemotherapy (Hellmann et al. NEJM 2018). Standardized approaches for the measurement and reporting of TMB are essential for the real-world implementation of TMB. This study aimed to refine a bioinformatic pipeline for mutation calling and annotation of whole exome sequencing (WES) data for TMB assessment.
In CheckMate 026, TMB was assessed by WES on formalin-fixed, paraffin-embedded tumor samples and matched blood from 312 patients with NSCLC (Carbone et al. NEJM 2017). Data from each sample were aligned to a reference human genome and somatic mutations were called by comparing matched tumor and blood samples using the TNsnv and Strelka algorithms. The somatic mutations were additionally filtered for germline variants in public databases. TMB scores were compared with data from 710 NSCLC samples in The Cancer Genome Atlas (TCGA) dataset. We examined the concordance of TMB estimates using several mutation filtering schemes, with and without matched germline controls.
TMB scores including synonymous, indel, frameshift, and nonsense mutations (all mutations) were ∼3-fold higher than matched data filtered for missense mutations only, but values were highly correlated (Spearman’s r = 0.99). Scores including missense mutations only were similar to those generated from TCGA, but those including all mutations were on average higher. Using public databases for germline subtraction showed a trend for race-dependent increases in TMB scores.
Standardization of bioinformatic analyses is critical to the clinical implementation of TMB assessment. TMB assessment is sensitive to variations in bioinformatic parameters (eg, which type of mutation to include), which may affect the identification of patients likely to respond to immunotherapy. These results show that data from different pipelines are highly correlated, suggesting that reliable assessment of TMB across different centers and platforms is achievable.
Clinical trial identification
Legal entity responsible for the study
Medical writing assistance was provided by Stuart Rulten, PhD, of Spark Medica Inc. (US), funded by Bristol-Myers Squibb.
H. Chang, S. Srinivasan, A. Sasson, R. Golhar, D. Greenawalt, J.D. Szustakowski: Employment and stock ownership: Bristol-Myers Squibb. S. Kirov: Stock ownership and compensation: Bristol-Myers Squibb.