NEW YORK – A new publication led by investigators from Roche subsidiary Foundation Medicine has confirmed in real-world treatment data that Foundation's tumor mutational burden assay, which is run as part of its comprehensive genomic profiling test FoundationOne CDx, does predict the effectiveness of Merck's Keytruda (pembrolizumab) in patients with a variety of advanced solid tumors.
Foundation won companion diagnostic approval for this application from the US Food and Drug Administration in 2020, but at the time there were lingering questions about its effectiveness including whether the tumor shrinkage measure used in the pivotal trial of the test could be assumed to translate into better overall survival.
In a commentary at the time, University of California, San Francisco epidemiologist Vinay Prasad highlighted that question among others as reasons that the FDA approval might have been an "unwise decision."
Ryon Graf, head of real-world outcomes research at Foundation Medicine, said that his team believed it could address some of these criticisms to analyze a broad dataset of patients treated with immune checkpoint inhibitors. Along with collaborators from Roche's Flatiron Health and several academic cancer centers, the Foundation Medicine researchers published their results last week in the Journal for Immunotherapy of Cancer.
"We knew that each individual indication was not going to be investigated in prospective randomized controlled trials, but we felt the next best thing we can do is to use this really high-quality dataset and rigorous methods to help close some of these gaps," Graf said.
Working with Flatiron, the researchers were able to review outcomes data on more than 8,000 patients with 24 types of cancer treated across 280 facilities. Graf stressed that the analysis used an overall survival endpoint validated against the National Death Index and a statistical analysis plan that prespecified certain levels of power that would be considered valid.
The result, among other findings, was that the observed tumor shrinkage seen in the pivotal prospective trial of Keytruda did appear to carry through to overall survival.
Although it might not be the most exciting finding, Graf said, this type of revalidation is something that the company sees as important. "It's something that I personally feel is a little bit underappreciated in the literature — not just the discovery, not the initial validation, but the actual follow-up," he said.
Real-world data also offers the opportunity to confirm that test performance seen in the controlled environment of a trial can hold up when applied in the more variable setting of the clinic, Graf said.
"Showing that it actually comes through over years and years for a really highly complex assay is nice."
In Graf's view, the study results don't represent an unequivocal yes as to the value of Foundation's TMB test in the pan-cancer advanced disease setting, but it does address caveats raised at the time of its FDA approval.
"There are people out there who are going to say everything should be supported by a Phase III prospective randomized controlled trial, which is completely impractical," he said. "But then you also have on the other end people [saying], 'Let's be practical here.'"
"I do understand the criticisms that some people have about remaining evidence gaps or lack of prospective randomized validation of some of these disease areas. But I do feel that this is a good step in the right direction," he added.
Aside from confirming overall survival predictiveness, the analysis also allowed investigators to address another question that was raised during the test's pan-cancer approval, the appropriateness of FoundationOne's cutoff for TMB of 10 mutations per megabase (mut/Mb) across different tumor types.
Graf said opposition to a universal cutoff is valid, as data has suggested that optimizing cutoffs to each individual tumor type may provide better discrimination. However, this is somewhat confounded by the fact that studies have used different technologies that might not be directly comparable.
Based on their analysis of the 8,000-patient dataset, the authors of the new study concluded that 10 mut/Mb remained "a reasonable single cut point" for pan-tumor assessment using Foundation's test.
That said, the team wrote, this should be taken in the context of Foundation's CDx approval, which is specific to patients who have exhausted standard-of-care options and is something that people "gloss over sometimes," Graf said.
For newer indications, especially where immunotherapies are being explored in the first-line setting, the field is going to have to work hard to optimize cutoff points for TMB that are appropriate in various specific scenarios, Graf said.
In their database analysis, the investigators were able to also look at cases where immunotherapy was prescribed alongside chemo. In this dual-treatment context, there was a much smaller difference in outcomes using the 10-mutation cutoff, with real enrichment of benefit resting at 20 mutations or more.
According to Graf, this analysis was driven by UC Davis oncologist David Gandara, the study's first author.
"We had drafted this analysis around the question of single-agent use, but he convinced us that this was something we could still explore in the data, and interestingly, we did see that the degree of stratification for benefit occurs at that higher level," Graf said.
Numerous immunotherapy trials have now been completed and proven successful in first-line settings, commonly in combination with chemotherapy, Graf added.
As such, there is an opportunity for investigators, as long as they have the appropriate samples, to go back and explore whether a different TMB cutoff might offer an opportunity to better personalize treatment.
The same could be true for failed first-line trials, where immunotherapy fell short in all-comers, but might offer benefit in a TMB-defined subset of patients.