Merck Bid for Second Pan-Cancer Keytruda Indication Would Raise Tumor Mutation Burden to CDx Status

NEW YORK – If Merck is successful in garnering US Food and Drug Administration approval for pembrolizumab (Keytruda) in advanced cancer patients with high tumor mutational burden, it will be the second pan-cancer indication for the drug.

This will allow Merck, which recorded pembrolizumab sales of $11.1 billion last year, to make the drug available to more cancer patients who are out of treatment options.

"What we are doing is providing a significant path towards pembrolizumab monotherapy in the second line [or later] setting where it didn't exist before, filling in the gaps left behind by the current indications," said Dave Fabrizio, VP of translational strategy at Foundation Medicine.

The company's FoundationOne CDx was used to establish TMB status in patients in one portion of the wide-ranging KEYNOTE-158 trial, and the data from this portion is supporting Merck's latest bid for another pan-cancer indication for pembrolizumab.

However, oncologists will still have to contend with this complex biomarker amid evolving research and alongside PD-L1 expression status, microsatellite instability (MSI), and mismatch repair deficiency (dMMR), when deciding whether to prescribe the drug to patients.

Since pembrolizumab first came to market in 2014, the drug has been approved for a number of indications, several of which stipulate assessing PD-L1 expression status to determine whether patients are likely to respond. Then, in 2017, the FDA approved the first pan-cancer indication for pembrolizumab, relying on MSI-high and dMMR to identify patients with unresectable or metastatic solid tumors who are progressing on other drugs or are out of alternatives, and for whom the immunotherapy may be an option.

In seeking a second pan-cancer indication with TMB, Merck is aiming to extend immunotherapy as an option to those with unresectable or metastatic solid tumors who are progressing on other drugs or out of alternatives, and who wouldn't be identified using MSI-high/dMMR. However, as one expert pointed out, when deciding whether to direct immunotherapy using MSI or TMB at the cutoff Merck has stipulated — 10 mutations/megabase — oncologists should keep in mind that the latter has only shown to extend response rate, and the former is now associated with a survival benefit.

While a TMB cutoff at 10 mutations/megabase is "very inclusive, the lack of a survival benefit shows it to be suboptimal," said Luis Diaz, head of the division of solid tumor oncology at Memorial Sloan Kettering. "Scientifically, it is a very interesting area to think about a mutational burden cutoff. In the end, it'll be more complex than just a cut point and involve more specifically, the nature of the types of mutations."

KEYNOTE-158

The data supporting Merck's application with the FDA in the pan-cancer TMB setting comes from KEYNOTE-158, which has many arms. Data from the same study resulted in pembrolizumab's first pan-cancer indication using MSI-high and dMMR, which Diaz led. 

The available data from this trial to date, according to Diaz, will limit the use of TMB as a pan-cancer biomarker at the 10 mutation/megabase cut point in earlier treatment settings. "This is going to be a scenario where it's primarily used as a salvage mechanism when people have failed all lines of therapy," he said.

The TMB portion of KEYNOTE-158 enrolled patients with 10 tumor types — anal, biliary, cervical, endometrial, salivary, thyroid, vulvar, neuroendocrine, small cell lung cancer, and mesothelioma. There were 755 patients who were evaluable for TMB, 120 of whom had high TMB. Around 28 percent of TMB-high patients saw their tumors shrink on pembrolizumab, compared to less than 7 percent in the TMB-low group. Median duration of response was not reached for either group.

Median progression-free survival was 2.1 months and did not differ based on TMB-high or -low status. Around 24 percent of patients in the high group lived for a year without their disease worsening compared to 14 percent in the low group. Median overall survival, meanwhile, was 11 months in the TMB-high group versus 13 months in the low group. At one year, survival rates were 48 percent and 53 percent in the high and low TMB groups, respectively. 

Researchers led by Aurélien Marabelle of Gustave Roussy presented this data last year at the European Society for Medical Oncology's annual meeting. They wrote in an abstract that "the tail of the PFS curve favored TMB-high," and concluded that the data suggest TMB is a predictive biomarker for pembrolizumab monotherapy in a variety of solid tumor types.  

In Diaz's view, the lower, universal TMB cutoff may make more patients eligible for pembrolizumab but it comes at the expense of being able to identify patients who may have a survival benefit. "We might be able to make them feel better [with pembrolizumab] because a lot of these tumors cause pain or other side effects and [the drug may make] their tumors shrink," he said. But TMB in this context is "not as robust as other biomarkers such as MSI, which actually have a robust survival benefit."

Data from KEYNOTE-158 published earlier this year showed that among 233 advanced patients with 27 types of solid tumors, who had failed prior treatments and received pembrolizumab for two years, the median progression-free survival was 4.1 months and the median overall survival was 23.5 months. 

The cutoff question

Merck, meanwhile, is hoping that the data will still sway the FDA to approve the second pan-cancer indication for pembrolizumab given the unmet need in the indication the company is pursuing: patients with advanced cancers who are out of options. Merck is focusing on the fact that TMB is helping identify patients who are likely to respond to immunotherapy who wouldn't be identified via MSI-high or dMMR.

Around 15 percent of patients across various solid tumors are MSI-high or dMMR, according to studies. Generally, TMB is a biomarker that can identify individuals whose tumors have accumulated mutations as a result of defects in mismatch repair genes (MSI-high/dMMR) or for other biological and environmental factors, such as UV exposure in melanoma or smoking in lung cancer. Therefore, TMB at 10 mutations/megabase would allow Merck to identify patients for pembrolizumab therapy who have a high mutational burden for reasons other than MSI or dMMR, but the specific size of this TMB-high/non-MSI population can differ based on the specific tumor type and how MSI and TMB are calculated.

In KEYNOTE-158, 120 out of 755 cancer patients, or around 16 percent, had high TMB at the designated cutoff. Within the TMB-high subset, 15 patients, or 13 percent, were also MSI-high. There was low correlation between PD-L1 expression and TMB status. The prevalence of high TMB in the study ranged from 1 percent and 40 percent depending on tumor type. A high tumor mutation burden has been reported in "virtually every cancer type," a Merck spokesperson said.

However, patients with certain types of cancers, such as melanoma and non-small cell lung cancer, are already eligible to receive pembrolizumab based on proven efficacy in those tumor types. "What's left over is a group of rare cancer patients who have TMB for a variety of reasons that are not MSI-high, melanoma, or lung cancer," Diaz said.

Moreover, despite an FDA-approved drug label stipulating a universal cutoff for TMB in a salvage setting, the real-world use of the biomarker in specific tumor types will continue to be challenged by a diagnostics landscape crowded with NGS panels, which are different with regard to the genes they gauge, the types of mutations they factor in, how they filter germline mutations, and the bioinformatics methods they employ.

In addition, labs report the biomarker in different ways in test reports. Some use different cutoffs to separate patients into high and low categories, which impacts who gets access to immunotherapy.

Absent FDA approval for TMB as a companion diagnostic, Foundation includes the biomarker in a section of the test report for "other alterations and biomarkers identified," and reports it in terms of mutations/megabase regardless of actionability. In the "professional services" section of the report, the company notes if the score can be used to inform immunotherapy based on cut-offs in the literature and clinical evidence. Fabrizio noted that these cutoffs range from 10 mutations/megabase to 20 mutations/megabase, depending on the particular histology.  

Another NGS cancer testing firm, Caris Lifesciences, has been using universal cutoffs for TMB categories of low, intermediate, and high in its MI Profile report. In the current report, 17 mutations/megabase and above delineates high TMB status. However, the company will soon launch a whole-exome sequencing test, and will start reporting TMB as a "percentile score of mutations/megabase on a tumor by tumor basis," which the company said will provide more specific information for an individual's cancer type.

The non-profit Friends of Cancer Research (FOCR) has been working with NGS test providers, like Foundation, Caris, MSK and others, to study the variability between how labs are calculating and reporting the biomarker, aiming to arrive at some standardization. "Our hope through this [TMB harmonization] project is to determine the amount of variation between … labs and consider ways that a common reference standard may be used to normalize different TMB scores between labs participating," FOCR CEO Jeff Allen said.

As part of this project, FOCR released a concept paper which describes efforts by labs to arrive at a minimum cutoff level for tissue-agnostic drug development. There was agreement in the working group that 10 mutations/megabase was a reasonable lower limit for enrolling patients into studies for tissue-agnostic trials. Allen emphasized, however, that while the hope is that this minimum threshold would be used to enable more patients to enroll in clinical trials, this has no bearing on the TMB cutoffs that may be ultimately validated and used by FDA-approved tests. 

FOCR also recently published a paper from the first phase of its harmonization project, in which 11 labs explored the different capabilities of their tests in determining TMB across 32 tumor types using whole-exome sequencing as a reference. "Our publication shows that a [TMB score of] 10 with one test may not be a 10 with another," Allen said.

In this phase of the project, the labs used the TMB range 0 to 40 mutations/megabase to divide 32 tumor types into three strata. Stratum 1 had eight tumor types that had good TMB distributions, including bladder urothelial cancer, colon adenocarcinoma, head and neck squamous cell cancer, lung adenocarcinoma, lung squamous cell, skin cutaneous melanoma, stomach adenocarcinoma, and uterine corpus endometrial carcinoma. Stratum 2 and 3 had tumor types that had low or very low TMB.

The authors of the paper, which included Foundation's Fabrizio, experts from other labs, and FOCR, wrote that TMB measurements are likely to be most useful in guiding decisions for cancer patients with tumor types in stratum 1, and that labs should include these cancers when validating their TMB tests. But they recognized that other histologies that are less likely to have high TMB would also benefit from reliable tests that can identify best responders. Ultimately, the authors acknowledged that the distributions of TMB values does appear to be histology-dependent, and as such, "studies aiming to evaluate the clinical utility of TMB and determine optimal TMB cutoffs for treatment decisions may need to account for specific cancer types."

Researchers from MSK looked at this very topic and published a study in Nature Genetics involving nearly 1,700 patients with a variety of tumor histologies, who were treated with different anti-CTLA4 and anti-PD-1/PD-L1 therapies, and had TMB determined based on the 468-gene MSK-IMPACT panel. In that analysis, higher TMB scores were associated with better treatment outcomes in general, but the researchers observed that tumor mutational load varied in range and median in each histology. 

Therefore, they reasoned that a universal cutoff for "high TMB" would be skewed in favor of cancers that tend to accumulate more mutations. To explore this further, they considered the top 20 percent of TMB scores for patients in each histology and reported that these patients had longer overall survival on immunotherapy than the rest of the patients, but the TMB cutoff delineating the top 20th percentile of patients varied by histology.

However, Fabrizio noted that this study wasn't designed to test whether a disease-specific TMB cutoff would be better at predicting outcomes with a specific immunotherapy than a universal cutoff for advanced patients.

"Not surprisingly, [the MSK study] found good clinical benefit when you stratify by 20th percentile on each specific disease. What they didn't necessarily do is [evaluate] what would have happened if we had chosen a single cutoff across all of these diseases and compare that to these specific TMB cutoffs," Fabrizio said. "I would really like to see that before making a conclusive statement that a disease specific cutoff for Keytruda in a second line-plus setting is going to [predict] response better than a pan-tumor [cutoff]."

Fabrizio acknowledged that doctors will have to weigh how to prescribe pembrolizumab in a pan-cancer fashion using TMB, using evidence from KEYNOTE-158 and research like MSK's. However, if the FDA approves this pan-cancer indication for pembrolizumab with TMB as a predictive marker at 10 mutations/megabase, that should give oncologists some confidence, he said, adding: "Ultimately, that's where the practice of medicine comes in and doctors will have to make the best decision."

Certainly, Luc Morris, the senior author of the MSK paper, has previously likened TMB to how PD-L1 expression status is used in the clinic. PD-L1 expression status is not a pan-cancer biomarker, but its adoption has been challenged by the variety of tests on the market and the different cut-offs associated with immunotherapy benefit across tumor types. "TMB is very similar in that it's not going to be the make-or-break factor for what your patient's next therapy is," he said last year, "but it's going to be one of many factors that you're going to consider."

'Work in progress'

Diaz also characterized TMB as still very much a "work in progress" and further questioned whether it might be a bit premature to push ahead with a pan-cancer indication based on TMB. "What I haven't come to terms with is whether an approval now will be good or bad for the field," he said. More research is needed on TMB, but he wondered whether this approval still allow for more research on this biomarker.

Several NGS tests for directing cancer therapy have FDA approval or clearance, including MSK's MSK-IMPACT and FoundationOne CDx, but neither are approved to report TMB as a CDx alongside a specific immunotherapy. NantHealth last year garnered clearance for its Omics Core whole-exome test for TMB, but cannot make any predictive claims about specific drugs.  

If Merck and Foundation are successful in garnering a pan-cancer CDx indication for TMB with pembrolizumab, Fabrizio noted that the company would continue to provide oncologists with disease-specific information on the biomarker based on ongoing research.

"We are essentially limited by the methods and infrastructure in which we can pursue companion diagnostics for immunotherapies or targeted therapies, which means largely separating a group into positive or negative or high or low [biomarker groups]," he said. The FDA approval process will designate a discrete cutoff for TMB, for differentiating high and low status, but "the science tells us that TMB is more of a continuous variable … associated with a probability of response depending on your actual score." 

The constructs of the FDA's CDx designation does "leave behind" information that can be gleaned from TMB, which can be important for oncologists to have in making treatment decisions for patients, he said. "Certainly, doctors need to have this information at their fingertips," Fabrizio noted, "so that they can use not only the label associated with the approval, but the actual score itself in the context of all the other information that's being provided to them through a Foundation Medicine report."