NEW YORK – A Dana-Farber Cancer Institute- and Harvard-led team has demonstrated that tumor mutational burden (TMB) tends to be overestimated in non-European cancer patients tested with tumor-only sequencing, which can impact outcomes on immune checkpoint inhibitor immunotherapies that use high TMB as a biomarker.
To address the effect, the investigators called for more extensive biomarker research in diverse human populations, and proposed an ancestry-informed calibration coefficient approach to bolster the accuracy of current TMB estimates.
"If this holds in the general patient population, it will have a substantial impact on treatment selection for many patients," co-senior study author Alexander Gusev, a data science, population sciences, and genetics researcher affiliated with Dana-Farber Cancer Institute, Brigham and Women's Hospital, and Harvard Medical School, said in a statement.
For a study appearing in Cancer Cell on Thursday, the researchers began by looking at the potential for spurious TMB measurements in individuals of non-European ancestry, who have traditionally been neglected by genetic studies. Such measurements are crucial for cancer cases where TMB-high status is used to guide treatment with checkpoint blockade immunotherapy drugs such as Keytruda (pembrolizumab), they explained.
Bringing together data for more than 3,600 cancer patients profiled for the Cancer Genome Atlas project, the team considered the TMB scores that would be given for the cases using tumor-only exome sequencing versus tumor and matched normal sequencing, focusing on hundreds of genes targeted by Oncopanel somatic mutation tests.
In that dataset and in a validation group that included almost 2,400 non-small cell lung cancer (NSCLC) patients profiled on MSK-IMPACT platforms, the researchers found that TMB-high cases were overestimated using tumor-only sequence data from non-European individuals, in part due to germline variation being misclassified as somatic mutations.
In addition, the team reported that overall survival outcomes did not improve for the subset of 121 Asian or African American NSCLC patients with seemingly high TMB who received immune checkpoint immunotherapy treatment, consistent with the notion that at least some of the TMB classifications made with tumor-only sequence data were higher than they would have been if matched normal tissue sequences were taken into account.
In contrast, the researchers found that high TMB coincided with better-than-usual survival outcomes in more than 1,500 immune checkpoint inhibitor-treated European NSCLC patients with high TMB.
"This does not mean TMB should not be used as a biomarker for non-European ancestry individuals, but it does mean that much more study and data collection is needed to determine the precise thresholds to be used to optimize the treatment outcome,” Gusev noted.
On the contrary, he suggested, those involved in the study "hope this emphasizes the broader importance of evaluating biomarker accuracy in diverse populations."
The investigators saw ancestry-associated biases when they focused on alterations in individual gene markers for immune checkpoint immunotherapy response. After sifting through nearly three dozen genes with ancestry-related variant patterns, they dug into data for MGA gene alterations, which have been implicated in longer survival and stretched out time to treatment failure in immune checkpoint inhibitor-treated NSCLC patients of European descent.
With tumor and normal data for nearly 1,900 immunotherapy-treated NSCLC patients, the team confirmed ties between enhanced survival and MGA alterations in the European patients. But such changes did not track with outcomes in the NSCLC patients with African or Asian ancestry.
"We observed ancestry-specific effects at some individual driver genes, suggesting that this phenomenon may not be unique to TMB, but also hold for individual mutations and have implications for other targeted therapies," Gusev elaborated, adding that it will be "critical to accelerate the collection of data and outcomes from these populations so that we can better understand how effectively TMB and other biomarkers generalize to all patients."
To tackle the current TMB overestimate concern, the team outlined a recalibration strategy that takes ancestry into account and overcomes what Gusev described as "purely technical bias" introduced when estimating TMB from tumor-only sequence data.
After training the calibration coefficient approach with TCGA data, the group used the method to come up with ancestry-corrected TMB measurements for more than 1,800 patients from Dana-Farber, who together received checkpoint immunotherapy for seven cancer types, and for 234 checkpoint immunotherapy-treated NSCLC patients from Memorial Sloan Kettering Cancer Center.
The corrected TMB classifications better tracked with those drawn from paired tumor and normal sequence data, the investigators reported, separating "true" from "false" TMB groups with distinct overall survival and treatment response curves.
"Differential estimation between ancestries is driven by underrepresentation of non-European populations in germline databases and is directly mediated by patient ancestry rather than race/ethnicity," the authors emphasized. "We thus suggest revising TMB calculations for tumor-only samples using the tumor-based ancestry inference employed herein, when paired germline-tumor samples or large reference panels are unavailable."