In an effort to determine best practices and ensure consistent clinical interpretation of tumor mutational burden ( TMB) assessment for cancer patients, a group of diagnostic test partners conducted an in silico analysis and found that panel-derived TMB strongly correlated with whole-exome sequencing (WES) data provided from The Cancer Genome Atlas (TCGA). The group’s findings were presented by David Fabrizio, PhD, Cancer Immunotherapy Leader at Foundation Medicine during the 33rd Annual Meeting of the Society for Immunotherapy of Cancer (SITC 2018).
TMB is a measure of the number of somatic mutations in the genome of a patient’s specific cancer, and may serve as a predictive biomarker of response to immune checkpoint inhibitors (ICI) across several cancers. TMB can be estimated using next-generation sequencing (NGS), but quantifications can vary across different platforms, testing panel size and composition, and bioinformatic algorithms. TMB is also a candidate for pan-tumor biomarker of ICI sensitivity.
Eleven members of the Friends of Cancer Research (Friends) group analyzed WES samples across 32 cancer types. The group identified sources of TMB variability and worked to harmonize TMB estimation to ensure consistent clinical interpretation in the future. The rationale for conducting the study is that theoretical variation in TMB quantification across panel-based diagnostic platforms exists and warrants empirical alignment with a reference standard.
“High levels of TMB are observed in numerous tumor types and the correlation with checkpoint inhibitor therapy has been observed across nu-merous individual tumor types,” said Dr. Fabrizio. “However, TMB is a complex biomarker. It’s a continuous variable and it requires standard-ization and consistency across technologies and platforms.”
The group calculated the TMB measurements from the WES data using a predefined method, explained Dr. Fabrizio. Each partner then conducted a subsequent calculation of TMB by down sampling the WES data, and restricted the calculation only to the genes that were codified in their technology platform. “This allowed for the comparison of the down sampling data back to the WES data,” he said.
Each diagnostic partner calculated TMB from the subset of the exome restricted to the genes covered by their targeted panel and used their own bioinformatics pipeline, the results of which were called ‘panel-derived TMB’. A gold-standard TMB estimate was calculated from the entire exome using a uniform bioinformatics pipeline that all members agreed upon, the results of which were named ‘WES-derived TMB’. Linear regression analyses were performed to investigate the relationship between WES-derived TMB and panel-derived TMB. Exploratory analyses by cancer type were also performed, and bias and variability in TMB estimates across panel-derived TMB values were assessed.
The investigators found that in silico quantification of TMB is relatively consistent a wide range of TMB values (0-40 mut/Mb) between the panels. The investigators noted that some variation in TMB quantification across panel-based diagnostic platforms exists. Identifying factors that contribute to variation will help enable harmonization and ensure appropriate use and implementation of TMB assessment results in the clinic. “What we have seen across all these platforms is some level of agreement, but also, some level of variability does exist,” Dr. Fabrizio said. “We need to understand what contributes principally to this variability conducting further investigation.”
Beyond this, the group also investigated whether tumor types contributed to the differences in the correlation between whole-exome–derived TMB value and the panel-derived TMB value. They found that tumor type could affect the correlation. Dr. Fabrizio said that when focusing on some of the more heavily investigated tumors—such as lung, bladder, urothelial, and melanoma—that are treated with checkpoint inhibitors, “a similar level of variation is exhibited, but it’s not well understood what is contributing to this variation.”
Subsequent research should assess the effect of biologic factors (e.g. specimen type, cancer type, treatment setting), the impact of variation on clinical outcomes, align standards, and define best practices for quantification of TMB, he recommended.
“We need to define and publish what we consider to be best practices for the analytic validation of TMB, which should include accuracy, precision, sensitivity, and alignment against what we consider to be a universal set of standards.”