Skip to main content

NCI-MATCH Data Shows Value of Platform Trials for Patients With Rare Genomically Driven Cancers

Premium

NEW YORK – Around 12 percent of approximately 5,500 sequenced cancer patients received treatment based on molecular tumor markers within the National Cancer Institute's Molecular Analysis for Therapy Choice (MATCH) trial, according to recently published data.

The findings from NCI-MATCH, one of the first precision oncology-focused umbrella trials, published on Tuesday in the Journal of Clinical Oncology, demonstrate the value of multi-cohort platform trials as an option for refractory cancer patients who are out of standard treatment options and who have particularly rare genomic tumor alterations that would not be explored in a traditional drug trial. In the study, actionable markers tended to be those that occur in 3 percent or fewer of cancer patients. However, the relatively low match rate also points to the limitations of current scientific knowledge on the potential targetability of oncogenic biomarkers, as well as the need for more treatment options and combination approaches that can interrogate molecularly driven cancers.

In the JCO paper, NCI-MATCH investigators, led by Keith Flaherty from Massachusetts General Hospital, described the characteristics of the more than 6,000 cancer patients who signed up to partake in the study. NCI-MATCH, launched in 2015, employs an "umbrella design" that allows investigators to explore the activity of targeted drugs in patients with different tumor types driven by the same molecular aberration.

Although cancer patients were enthusiastic about partaking in the study, there were signs that the rarity of many targetable molecular drivers and the lack of drugs for certain tumor markers could make it challenging to match the majority of patients to therapeutic arms. The latest data published in JCO reflects some of these challenges, where 6,391 patients registered to partake in the trial, 5,954 patients submitted a tumor sample, and 5,540 (93 percent) had their samples successfully sequenced.

Based purely on the molecular alterations identified, 38 percent of 5,540 patients had a marker that was being investigated in one of NCI-MATCH's treatment arms. After accounting for exclusion criteria, 26 percent of patients were eligible for enrollment in a study arm. Some of these patients ultimately couldn't get on a treatment arm, however, because the subprotocol was fully accrued or had reached its limit for a specific tumor type, which reduced the eligible population to 985 patients, or 18 percent of the sequenced cohort. Ultimately, only 12 percent, or 686 patients, actually received a treatment within the study.

Flaherty explained in an interview that it's never the case that 100 percent of patients eligible for a study enroll in the trial. After accounting for exclusion criteria, he highlighted the fact that 686 out of 985 eligible patients, or 70 percent, actually got on a treatment arm in NCI-MATCH and noted that this is in the range with what's typically seen in other cooperative group treatment studies.

"These are metastatic patients, who have exhausted standard therapies, and their tumors are progressing," said Flaherty, explaining one of the main reasons for the treatment assignment rate in the study. "Some of these patients can become ineligible for enrollment within days to weeks" as they're being evaluated for study eligibility.

Because advanced, heavily pre-treated, refractory cancer patients are usually enrolled in platform trials like these, the 12 percent treatment assignment rate is also in line with what's been seen in other precision oncology umbrella and basket studies. For example, an MD Anderson Cancer Center study described in a 2015 JCO paper had a match rate of 11 percent. 

In NCI-MATCH, researchers had a certain clinical evidence bar that molecular markers had to meet in order to be consider "actionable." It wasn't enough if the association between a tumor marker and drug response was suggested in preclinical evidence, but there also had to be "some modicum of clinical evidence," Flaherty said. In establishing the 30 treatment subprotocols, investigators embarked on an "exhaustive search" for clinical data that provided at least a preliminary indication that patients with certain tumor alterations might respond to a particular drug. Unfortunately, Flaherty acknowledged, most patients just don't have targetable tumor markers that met NCI-MATCH's evidence threshold.

Some of the most common mutations identified in patients occurred in TP53 (47 percent), KRAS (21 percent), and APC (12 percent), with KRAS and TP53 mutations most frequently occurring together in 12 percent of patients. However, the most common mutations reported in NCI-MATCH participants were not the most actionable.

For example, when Flaherty and colleagues were looking for clinical evidence for drugs that could potentially target KRAS-mutant cancers, there weren't any. Drugs targeting KRAS G12C-mutant cancers, which comprise a small segment of the larger KRAS-mutated cancer population, hadn't entered clinical trials yet. As such, NCI-MATCH investigators couldn't launch any treatment subprotocols in KRAS-mutant cancers, even though up to 25 percent of cancer patients harbor such mutations.

"In the NCI-MATCH study, we saw how common KRAS mutant cancers are, and if we had great ideas for KRAS-mutant cancers, we for sure would have included them in the study," Flaherty said. "There are very well-known molecular subsets in cancer for which it's not a total black box, but it's that we didn't, and to a large degree don't yet, have therapies to offer." 

The most actionable alterations occurred in PIK3CA (12 percent) and PTEN (6 percent), though for the most part, actionable mutations occurred in 3 percent or fewer participants. None of the 18 treatment arms that involved tumor alterations with a prevalence of less than 1.5 percent reached the accrual goal of at least 31 patients.

For example, patients with NTRK fusions, which occur in less than 1 percent of solid tumors, had "striking responses" to drugs targeting these alterations across histologies. The US Food and Drug Administration has to date approved two histology-agnostic drugs targeting NTRK fusions, larotrectinib (Bayer's Vitrakvi) and entrectinib (Genentech's Rozlytrek).

It would have been impossible to study these indications using a traditional trial design in specific tumor types because of the rarity of the biomarkers, but multi-cohort studies, such as the NCI-MATCH, have allowed their exploration in a tissue-agnostic fashion. "The investigation of novel therapies in less common molecular subgroups is made more efficient by investigating many therapies in parallel in an NGS-guided platform trial," Flaherty and colleagues wrote in JCO.

However, the available evidence on co-occurring mutations that could potentially confer resistance also presented a challenge for treatment assignment. Around 34 percent of patients had one alteration that could be targeted by a study drug and no other co-occurring markers.  However, 723 out of 2,083 patients with some of the most common actionable mutations could not get a treatment in NCI-MATCH because they also had other mutations that could make them resistant to the study drug.

For example, around a third of patients with targetable PIK3CA mutations were excluded because they also harbored RAS or PTEN resistance mutations. However, 42 percent of patients with PIK3CA mutations had co-occurring mutations in TP53, and although this may have conferred resistance, researchers didn't limit enrollment on this basis due to the lack of evidence. In the study, more than half the patients with actional alterations and co-occurring resistance mutations in tumor suppressor genes were still able to get on a treatment arm.

Some platform studies have yielded lackluster results in terms of outcomes, prompting some in the field to criticize precision oncology-focused cancer centers and oncologists for buying into hype. However, proponents of precision oncology have countered that the outcomes of these trials reflect the fact that cancer patients are often too sick at the time of enrollment and have suggested that the results might improve if patients were enrolled earlier in their cancer journey and given drug combinations that target multiple pathways and alterations driving their tumors.

In the iPREDICT study, for example, which is a small study but had a relatively high match rate for a precision oncology trial, researchers from the University of California, San Diego, used a variety of genomic analyses and a matching score to identify treatment options and combination approaches for patients when possible. 

Within NCI-MATCH, again, the safety and efficacy evidence was simply not there to enable combination treatment strategies for patients. Out of 30 subprotocols, only one explored a combination treatment. "Cancer is complicated and resistance should always be anticipated. ... We think that combination therapies are important," Flaherty said. "The reality is that there just are not dozens and dozens of these combinations that have been validated to a degree in a single setting that now could be explored more broadly."

However, in recent years, NCI-MATCH has evolved from centralized testing to decentralized genetic evaluation for enrollment, which may bolster the ability of the trial to find patients who match to treatment arms.

In the study, researchers used an NGS test adapted from Thermo Fisher Scientific's Oncomine AmpliSeq panel that gauges 143 genes. There was a pre-specified plan to enroll patients initially via centralized testing conducted at four labs. The JCO paper reflects analysis on patients who were tested within labs at MD Anderson Cancer Center, Massachusetts General Hospital, Yale University, as well as at the Frederick National Laboratory for Cancer Research.

However, a few years ago, after the prespecified enrollment goal was reached, the study organizers decided to launch a qualification process that would allow other labs to test patients for enrollment. "We knew that the vast majority of cancer patients were not being tested at those four academic laboratories," Flaherty said. "The big commercial testers ... and many other commercial and academic laboratories stepped forward to enter this credentialing process." 

In order to be qualified, labs had to, among other things, present data on their testing capacity and demonstrate that they could test for all types of genetic alterations of interest in the study. When qualified labs detect a marker being studied in an NCI-MATCH treatment arm, they include that information in the patient's test report, allowing the ordering physician to consider that as an option.

"What's nice about platform trials is the efficiency for doctors and patients, because you do a single test and it informs eligibility for multiple potential treatment options," Flaherty said, adding that platform trials are a great mechanism for investigating tumor types that haven't been studied as much. "The accumulation of rare cancer types that we saw in NCI-MATCH are a great testament to this point," he added.  

This lab qualification process developed within NCI-MATCH now has been adopted within all the NCI-sponsored platform trials, Flaherty said. Researchers are also conducting further studies based on exome sequencing of tumor samples from patients in treatment arms.

Lastly, researchers also compared the genomic profile of heavily pretreated patients in NCI-MATCH against treatment-naïve patients in The Cancer Genome Atlas Program hypothesizing that the more treatment a patient received the more molecularly altered their tumors would be in response. However, treatment-naïve and previously treated patients in these two studies were not that different in terms of their tumor genomic characteristics. In seven tumor types, the frequencies of 10 genes that tended to harbor mutations were "broadly similar" and the rate of co-occurring mutations were only somewhat higher in NCI-MATCH patients compared to TCGA patients, Flaherty and colleagues found. 

There was evidence that NCI-MATCH patients acquired resistance mutations following treatments. For example, nearly half of the patients with EGFR-mutated non-small cell lung cancer and previously on EGFR inhibitors had EGFR T790M resistance mutations.

"Basically, it suggests that if we did a detailed accounting of all the prior treatments … that the patients who emerge as resistant to those therapies, they basically have that built into their tumors, at least at a genetic level. … They kind of have what they need to be able to survive, persist, and emerge as resistant tumors to available treatments at the outset," Flaherty said. "That was surprising."