NEW YORK – The lack of diversity in public genomic databases may be inflating tumor mutational burden (TMB) measurements and leading to inappropriate use of immunotherapy, especially among Black cancer patients, recent research suggests.
Researchers led by Aaron Mansfield, a translational scientist at the Mayo Clinic, pointed out in a paper published in NPJ Precision Oncology last month that paired tumor-normal sequencing is the gold standard for filtering out germline variants when measuring TMB, but is infrequently done compared to tumor-only sequencing.
Instead, genetic testing labs often rely on public germline variant databases to subtract out germline mutations in TMB counts. This becomes problematic in the era of immunotherapy, particularly given the growing popularity of TMB as a predictive biomarker for identifying best responders to these costly treatments.
The problem with relying on germline variant databases, Mansfield and colleagues wrote, is that the people who have been sequenced and whose data are in these large databases do not necessarily reflect the broader population in terms of ancestral backgrounds. Numerous analyses — including two that the researchers cited in their NPJ Precision Oncology paper — have shown previously that the majority of those sequenced and genotyped in large public databases are white people of European descent. This becomes a source of bias when it comes to measuring TMB in minority patients, such as those of African ancestry, wrote the researchers.
To demonstrate this phenomenon and quantify the inflation, Mansfield and colleagues used a deidentified cohort of 701 patients with newly diagnosed multiple myeloma, of whom 575 self-identified as white and 126 self-identified as Black. Of note, the researchers acknowledged a limitation to their study, which is that self-identified race is not an ideal surrogate for genetic ancestry. Researchers relied on data, particularly paired tumor-normal exome sequencing data, from patients enrolled in the Multiple Myeloma Research Foundation's CoMMpass study.
Although this analysis focused on multiple myeloma, Mansfield said in an interview that he and his colleagues expect to see this race-based disparity in TMB measurements in tumor types other than multiple myeloma. They have, accordingly, requested datasets in solid tumors to validate this in future studies.
To assess the accuracy of TMB measurements, Mansfield and colleagues first looked at exome sequencing data available within CoMMpass and compared patients' paired tumor-germline data. Using this approach, the average TMB measurement — defined as the number of nonsynonymous somatic mutations per megabase — was similar between Black and white patients, averaging 6.09 and 5.47, respectively.
The researchers then considered TMB measurements according to race with only tumor-only sequencing, using the 1000 Genomes Project and Exome Aggregation Consortium (ExAC) databases to filter out non-somatic variants. This approach, they reported, inflated TMB measurements for both Black and white patients, but to a greater extent for Black patients.
First, when they subtracted out all germline mutations from the databases, the TMB estimates were 7.9 and 7.4 for Black and white patients, respectively. However, when the researchers filtered out variants with minor allele frequencies of at least 0.1 percent from the databases, the difference in TMB estimates between Black and white patients widened to 12.1 and 10.1, respectively. Finally, they did the same analysis but excluding variants with minor allele frequency of at least one percent and reported TMB estimates of 22.4 and 13.5 for Black and white patients, respectively.
When Mansfield and colleagues repeated their TMB analyses using 1,059 cancer-specific genes from OncoKB, they observed a similar disproportionate rate of TMB inflation, which was again more extreme among patients who identified as Black.
Treatment implications
Inflated TMB measurements can significantly influence patients' access to immunotherapies, especially given the US Food and Drug Administration's broad tissue-agnostic approval last year for pembrolizumab (Merck's Keytruda) for refractory patients with TMB-high solid tumors. In granting accelerated approval to this indication, the FDA accepted a TMB-high cutoff of 10 somatic mutations per megabase assessed using Foundation Medicine's FoundationOneCDx tumor-only NGS panel.
While Foundation's test has been extensively validated for concordance against tumor-normal WES measurements, it is nonetheless a tumor-only test that uses databases to filter out germline mutations. And while it is the only FDA-designated companion diagnostic for guiding TMB-based tissue-agnostic prescribing decisions with pembrolizumab, in practice, FoundationOne CDx is not the only method being used to assess TMB, and stakeholders have been working on harmonizing TMB measurements across different testing methods and platforms.
In their paper, Mansfield and colleagues explained that the inflated TMB measurements identified in their study suggest that relying on the 10 mutations per megabase TMB cut-off is likely to lead to more Black patients receiving immunotherapy, even though they are unlikely to benefit from treatment. Â
Additionally, while the present study stratified patients into just two self-reported race categories, the researchers hypothesized that the disproportionate TMB inflation would likely affect other minority patient populations underrepresented in large databases. "The inflated TMBs are likely relevant for other ethnic groups including Asians, Pacific Islanders, and other underrepresented groups," they wrote.
Mansfield and colleagues' research identifies but does not fix the TMB inflation issue among minority groups. That will take time and a concerted effort from stakeholders in the field to improve minority representation in research and databases, though Mansfield hopes this study will "encourage oncologists to look under the hood of how TMB is determined with the platform they use and question whether some results may be spurious."
This study is part of mounting research raising concerns about TMB as a tissue-agnostic predictive biomarker. In recent months, oncologists have acknowledged that the biomarker can be extremely useful for predicting treatment responses in certain patient populations and cancer types, but far less useful — if not detrimental — in others. Mansfield echoed one potential solution that oncologists have consistently floated alongside their concerns: using multi-biomarker approaches.
"Ultimately, I hope we can move beyond a single biomarker and use a more comprehensive immunoscore that has greater predictive capabilities than the current state of the art," he said. "As a clinician, I'd like to spare patients the toxicities of drugs that won't help them and offer something that might provide benefit instead."