NEW YORK – Cancer patients with 16 or more mutations per megabase in their tumors had better outcomes on Genentech's PD-L1 inhibitor atezolizumab (Tecentriq) compared to those with 10 to 16 tumor mutations per megabase, according to data presented during the American Association for Cancer Research's annual meeting.
The data adds to the research and criticism mounting against the 10 mutation per megabase tumor mutational burden (TMB) cutoff that the US Food and Drug Administration approved as a biomarker to guide the tissue-agnostic administration of Merck's PD-1 inhibitor pembrolizumab (Keytruda). Multiple studies now have suggested that a higher cutoff might improve the biomarker's clinical utility when used to determine administration of immune checkpoint inhibitors in a tissue-agnostic fashion.
In the research presented at AACR, lead investigators Claire Friedman of Memorial Sloan Kettering and John Hainsworth of the Sarah Cannon Research Institute presented atezolizumab's activity by patients' TMB status according to different cutoffs. These patients were part of Genentech's Phase II basket trial, MyPathway, in which the drugmaker is exploring the activity of its treatments in new molecularly defined indications, and in one of these cohorts, 121 patients with 21 solid tumor types and a TMB of at least 10 mutations per megabase received atezolizumab monotherapy.
MyPathway participants could have received any CLIA-certified assay to determine their TMB status. As a primary endpoint, investigators were interested in the overall response rate on atezolizumab of patients with a TMB of 16 mutations per megabase or higher, as determined by Foundation Medicine's FoundationOne CDx. This is the same assay that the FDA approved as the companion diagnostic alongside pembrolizumab's tissue-agnostic approval last year.
Patients tested on another assay for MyPathway enrollment had to provide tissue for retrospective retesting using FoundationOne CDx. Ultimately, there were 90 patients in the primary efficacy evaluation and 120 patients considered in the overall analysis.
"This [retesting] helped limit variability in TMB measurements observed between different gene panels," Friendman explained in her presentation at AACR. Indeed, there is widespread uncertainty when it comes to inter-assay variation for TMB assessment.
Among 42 patients with TMB of least 16 mutations per megabase as determined by FoundationOne CDx, the overall response rate with atezolizumab was just above 38 percent. This was a notable improvement compared to the 2.1 percent overall rate observed among 48 patients with TMB between 10 and 16 mutations per megabase. Similarly, the disease control rate was lower and progression-free and overall survival times were shorter in this latter group compared to the group with at least 16 mutations per megabase. Specifically, the disease control rates in the 10-to-16 versus the 16-and-up mutation subgroups were 22.9 percent and 61.9 percent, respectively; the median progression-free survival was 1.8 months versus 5.7 months, respectively; and the overall survival was 11.4 months versus 13.7 months, respectively.
Extending their analysis to include all study participants who had their TMB status determined by any CLIA-certified assay, the researchers reported a 28.6 percent overall response rate among 56 patients with TMB of at least 16 mutations per megabase, versus 3.2 percent among 62 patients with TMB between 10 and 16 mutations per megabase.
Among patients in the higher TMB subgroup who initially had testing with a test other than Foundation's, the earlier test and FoundationOne CDx results agreed 74.4 percent of the time in determining whether these patients had a TMB of at least 16 mutations per megabase.
Accounting for microsatellite instability
Before the FDA approved pembrolizumab in a tissue-agnostic setting using TMB as the biomarker, the agency in 2017 granted tissue-agnostic approval to the immunotherapy relying on microsatellite instability. Patients with high MSI also have an exceptionally high tumor mutation burden, which has led experts to question the predictive value of TMB in those who are microsatellite stable (MSS).
The MyPathway researchers also explored whether atezolizumab had activity in patients who had high TMB not attributed to MSI. To address this question, the researchers further stratified the study cohort into subgroups by both TMB and MSI status. Among patients with TMB of at least 16 mutations per megabase by FoundationOne CDx, the response rate was 56.7 percent among MSI-high patients, compared to 30 percent among those with MSS tumors.
"This suggests that atezolizumab has meaningful activity in patients with tumors characterized by a TMB of at least 16 mutations per megabase, regardless of MSI status," Friedman said, adding later that this suggests "there are genomic mechanisms other than MSI that drive response to atezolizumab in this population."
Next, researchers considered the impact of PD-L1 expression and noted meaningful activity in all PD-L1-expression subgroups when patients' TMB was at least 16 mutations per megabase. However, patients' response rates numerically increased with higher PD-L1-expression scores.
Finally, the researchers wanted to explore atezolizumab responses by TMB across different cancer types. The biggest subgroups by tumor type were breast and colorectal cancer patients.
Among 10 colorectal cancer patients with TMB of at least 16 mutations per megabase, the overall response rate was 70 percent; only three out of the seven responders had MSI-high tumors.
In contrast, only one out of seven TMB-high breast cancer patients with MSS tumors responded to atezolizumab. Two biliary tract cancer patients with TMB of at least 16 mutations per megabase experienced complete responses to atezolizumab.
Although the MyPathway study considered the activity of atezolizumab as opposed to pembrolizumab, the findings further the field's understanding of TMB as a tissue-agnostic predictive marker, suggested Sarah Cannon's Hainsworth.
"There's no doubt that the response goes up as the TMB goes up," he said. "In our study, it looks like the response rate is pretty low below 16." In the Keynote-158 study, the basis of pembrolizumab's tissue-agnostic approval, he noted that the roughly 12 percent response rate observed in patients with TMB between 10 and 13 mutations per megabase "wasn't very good either."
"We have an issue at the low end of the spectrum [of the TMB cutoff]," Hainsworth said, pointing out as well that different cancer types can have markedly different responses at the same cutoff.
Among breast cancer patients, for instance, who mostly had estrogen receptor-positive tumors, there was limited response to atezolizumab. There were 17 breast cancer patients, according to Hainsworth, whose TMB fell between 10 and 16 mutations per megabase, and none responded to the checkpoint inhibitor. The response rates among colorectal cancer patients with MSS tumors, on the other hand, were what Hainsworth called "a little surprising."
"Those [colorectal cancer patients] are typically considered not likely to respond and weren't in the group that was tested in the Keynote-158 study," he continued, pointing out that breast or prostate cancer patients were also not included in that study. "That study was based on selecting patients in 10 different tumor types, all of which have shown activity previously with checkpoint inhibitors," he said.
Considering the latest MyPathway results, coupled with the missing cancer types in Keynote-158, Hainsworth concluded that he "would not be very likely to use atezolizumab in patients with less than 16 TMB, and I'm not sure I'd be excited about using pembrolizumab either."