Tumor mutational burden (TMB) is currently being evaluated as a biomarker to predict response to immuno-oncology therapies in several studies. However, there is no agreed-upon standard for measuring TMB, with multiple companies and institutions using their own assays, which are mostly based on targeted next-generation sequencing.
Two independent initiatives — one by Friends of Cancer Research in the US, the other by the Quality Assurance Initiative Pathology (QuIP) in Germany — are now seeking to harmonize and standardize TMB testing, comparing existing assays along the way.
Earlier this month, Friends of Cancer Research, a non-profit organization based in Washington, DC, held a meeting of its TMB Harmonization Working Group, at which the two initiatives outlined their plans. Both organizations say their efforts complement each other, and they plan to jointly present first results at the European Society for Medical Oncology (ESMO) annual congress in Munich this fall.
Going forward, TMB will likely be used to predict whether a patient will respond to certain immunotherapies. “But what has not been seen is, is that threshold determined in a similar manner by different tests that have the same intended use?” said Jeff Allen, CEO of Friends of Cancer Research. “Hopefully, we will be able to provide some alignment, or at least an understanding of how the different tests relate to one another.”
“Both initiatives see the very same issues when it comes to testing of mutational load” and plan to provide recommendations and standardize current approaches, said Albrecht Stenzinger, head of the Institute of Pathology Heidelberg Center for Molecular Pathology at Heidelberg University Hospital and one of the organizers of the QuIP project.
A third initiative, by the European Society of Pathology in collaboration with the International Quality Network for Pathology (IQN Path), is currently getting underway but is not as far along yet, according to QuIP members.
TMB essentially measures how many somatic mutations a tumor cell has accumulated. The idea is that these mutations reflect the neoantigens — altered proteins a tumor cell expresses on its surface — that attract T cells, resulting in an immune response. The more mutations, and therefore neoantigens, the thinking goes, the better the response to an immune checkpoint blocker.
Several retrospective studies, particularly in non-small cell lung cancer but also in urothelial carcinoma and head and neck squamous cell cancer, have borne this out, confirming that TMB has clinical utility. However, TMB levels appear to differ between tumor types, and different tumors may require different TMB cut-offs.
Whole-exome sequencing currently serves as the “gold standard” for determining TMB, expressing mutational load as an absolute number, but is too expensive and time consuming to conduct routinely. Several companies and institutions have therefore developed panel-based NGS assays to measure TMB, which usually report the number of mutations per megabase. However, it is unclear if these assays reflect genome-wide mutational load in a similar manner.
While TMB is currently not approved as a companion diagnostic, pharmaceutical companies have started to incorporate TMB assays into their clinical trials. Bristol-Myers Squibb, for example, is currently utilizing Foundation Medicine’s FoundationOne CDx assay, as well as WES, to evaluate TMB in its trials. For example, the company used Foundation’s assay to assess TMB as a predictive biomarker in its phase 3 CheckMate-227 trial, which evaluated Opdivo (nivolumab)-based therapy combinations in non-small cell lung cancer.
Steven Averbuch, head of precision medicine at Bristol-Myers, said in an email that the two companies will continue to collaborate and, if Opdivo plus low-dose Yervoy (ipilimumab) is approved as a first-line treatment for NSCLC, “will advance the partnership to seek regulatory approval for F1CDx to assess TMB.”
Bristol-Myers is participating in both TMB harmonization initiatives, including providing an unspecified amount of financial support. “We look forward to the results of these efforts describing standards for assessing TMB as well as the potential demonstration of the comparability of different TMB assays,” he said.
Friends of Cancer Research’s TMB testing project builds on the field’s experience with PD-L1 testing, which also relies on multiple assays that were only recently compared against each other. “TMB has the potential to go in a similar direction if it’s not harmonized early on,” said Allen, “so we’ve tried to bring together a number of different laboratories and test manufacturers to explore the potential of alignment between their assays.”
Specifically, Friends of Cancer Research is examining TMB assays from Foundation Medicine, Personal Genome Diagnostics, Thermo Fisher Scientific, Illumina, Guardant Health, NeoGenomics Laboratories, Qiagen, and Memorial Sloan Kettering Cancer Center, and those companies and labs are participating in the project. Other interested institutions are still welcome to join, Allen said.
Several pharmaceutical companies, including Bristol-Myers, Genentech, Merck, AstraZeneca, Pfizer, and EMD Serono, as well as clinical key opinion leaders in the field have also had input to the project.
The initiative is divided into three phases, which Allen said may overlap somewhat in their timing. For the first phase, an in silico project, the seven diagnostic partners are applying their assays to exome sequencing data generated by The Cancer Genome Atlas (TCGA), using a standardized protocol and their own bioinformatics pipeline. The hope is that “we will obtain at least a baseline understanding of how the different tests compare to one another in terms of the conclusions they reach around the number of mutations per megabase,” Allen said. The initial analysis for this phase has already been completed and the participants plan to present their results in the near future, he said, though not the upcoming American Society of Clinical Oncology (ASCO) meeting.
The next phase of the project will apply the TMB assays to standardized human cancer cell lines and compare the results to exome sequence data. Again, the test manufacturers will generate the data for their own assays, using an agreed-upon protocol, and apply their own informatics pipeline.
This phase will require cell lines to be characterized by exome sequencing, so they can serve as a reference standard, not only for this project but also for future ones. Allen said the group is currently identifying suitable cell lines and looking for an independent laboratory to provide the exome sequence data.
While cell lines are less complex than tumor samples, Allen said, they have the advantage that their TMB and tumor purity is defined. The plan is to select already-established cancer cell lines that represent a variety of tumor types and levels of mutational burden. For example, it may be that two tests yield similar results at lower levels of mutational burden but different results at higher TMB levels.
“Those are important things that the group is hoping to sort out,” Allen said, adding that the goal is to complete this phase of the project by the fall, hopefully before the first drug gets approved with TMB as a companion diagnostic. However, “this project would still be informative if that were to happen, because we’re well on our way to understanding how different tests relate to one another,” he said.
For the third part of the project, the initiative will analyze clinical samples retrospectively, along with patient outcome data, in order to assess how clinically meaningful cutoff values vary between tests. The exact approach for this phase has yet to be determined, Allen said, and will depend on questions such as what TMB thresholds will be used in different cancer types.
The overall goal is to compare results from different TMB tests to one another, Allen said. For example, if a patient received a result from Foundation Medicine indicating a level of 12 mutations per megabase, he or she should be able to convert that to a number they would have received from a test performed by PGDx, he explained.
The QuIP initiative, which involves 11 academic pathology institutes in Germany and Switzerland, as well as TMB test manufacturers and pharmaceutical companies, including Bristol-Myers, is pursuing similar goals as the Friends of Cancer Research working group, but is taking a slightly different approach. The two initiatives found out about each other in 2017 and have met twice so far this year to coordinate their efforts.
QuIP is currently preparing to conduct a technical comparison study of TMB assays from Foundation Medicine; Thermo Fisher Scientific; Qiagen; Illumina; Neo New Oncology, which is part of Siemens Healthineers; and several academic laboratories. The goal of the study is “to understand what the various panels actually measure and how this compares to the reference standard, which is whole-exome sequence data,” said Stenzinger.
Testing for the project will be conducted by the participating pathology labs. Each lab will receive the same set of formalin-fixed paraffin-embedded tumor samples, for which mutational load was previously determined by exome sequencing, and will perform at least two types of TMB assays on them. In addition, the assay manufacturers will receive the same samples to run their own test. The aim is to complete the study by the fall, and to present results at the ESMO congress.
Using FFPE samples instead of cell lines, and having pathology labs perform most of the testing, has the advantage that it reflects what labs will encounter in routine diagnostics. For example, Stenzinger said, it is important that the labs in the study do everything from extracting DNA to creating the sequencing libraries, running the assay, and performing the bioinformatic analysis on their own. This can also point to differences resulting from a lab’s performance rather than from the assay. “The QuIP study is taking a very straightforward approach; we go straight to real-life samples and see what’s going on with the various panels that are currently available,” he said.
In addition to the planned wet lab comparison, one lab involved in the QuIP project has conducted an in silico analysis, similar to the one by Friends of Cancer Research, for which it compared TMB from TCGA exome data with TMB generated from TCGA data using currently available TMB panels. The results, which the group has submitted to a journal for publication, were similar to those of the US group, which Stenzinger said was encouraging.
For example, he said, the size and composition of the panel appears to matter, and whether the analysis includes just clonal or also subclonal events. It also makes a difference whether the panel is enriched for oncogenes or tumor suppressors, and whether it includes DNA repair genes.
“A panel of 0.4 to 0.5 megabases size will probably not work,” he said. “Based on the analysis we did in the German initiative, we think it’s really 1-megabase-plus that you need, at least.”
In the end, neither initiative will likely recommend a single assay, but they can probably provide a way to compare results from different tests, Allen said. “I don’t think we know enough about the contribution of different genomic alterations in order to say ‘this is the best way to call TMB’ ,” he said, but “even as this evidence emerges, I think it will be important to know how different tests relate to one another. “
Stenzinger agreed. “These two studies are about understanding the performance of the various assays, and to get a deeper understanding of the parameters that influence those results,” he said. “While there will not be a single panel that is the one and only one, it is more about what the core characteristics are that are needed for a robust measurement of mutational load.”