Keith T. Flaherty, MD, discusses the rationale to conduct the NCI-MATCH trial, key findings from the recent publication, and potential permutations of the trial design that could further the role of precision medicine in oncology.
Among a growing list of available anticancer therapies, the need to effectively utilize approved agents and enroll patients on active clinical trials of promising investigational therapies is more relevant than in years past, said Keith T. Flaherty, MD, who added that identifying new ways to quantify data may be a key piece of this puzzle.
In October 2020, data from the National Cancer Institute Molecular Analysis for Therapy Choice (NCI-MATCH) precision medicine trial revealed that 37.6% of the nearly 6000 patients with refractory malignancies harbored an actionable alteration.
The trial evaluated the antitumor activity of targeted therapy among patients with underexplored cancer types. The trial demonstrated the feasibility of screening large patient populations across 1117 accrual sites.
“These data are enormously helpful in clinical practice in terms of understanding how to use next-generation sequencing [NGS] to triage patients to known available therapies and think about a strategy to triage patients to clinical trials,” said Flaherty. “It’s relevant for academic medical centers and small community-based practices. [These data] provide us with a kind of blueprint about…how we can best serve our patients.”
In an interview with OncLive®, Flaherty, professor of medicine at Harvard Medical School, director of the Henri and Belinda Termeer Center for Targeted Therapy, and director of clinical research at the Cancer Center of Massachusetts General Hospital, discussed the rationale to conduct the NCI-MATCH trial, key findings from the recent publication, and potential permutations of the trial design that could further the role of precision medicine in oncology.
OncLive®:What makes the NCI-MATCH trial such an important collaboration in oncology?
Flaherty: NCI-MATCH is an important milestone in the broad area of public and private collaboration. NCI-MATCH was jointly led by ECOG/ACRIN and the NCI. Therefore, the leadership of the trial was dually constituted of leaders across those 2 groups.
However, it was critical to the success of the trial that we had universal engagement from pharmaceutical companies. We undertook a design that surveyed the entire landscape of oncology drug development for every example of a drug that was ready for phase 2 testing. [We identified drugs that had] a strong scientific basis that there was a molecular feature—be it a genetic alteration or some other molecular feature—that would identify patients who could respond to therapy. [This was done] regardless of cancer type; we were broadly investigating [drugs] across all cancers.
We were exhaustive in this approach and went out to every company that had a drug. In some cases, 4 or 5 companies had a drug against the same target. We asked those companies if they would be willing to share their updated confidential data because there was mutual interest to make their drugs available for us to use in the trial.
We were universally successful in that outreach. The response from companies in terms of access to their drugs [represents] a broad partnership in undertaking this trial. It has ramifications in terms of the data output of the trial.
Could you highlight some of the key points from the data that were published in the Journal of Clinical Oncology?
The data that were just published in the Journal of Clinical Oncology [represents] 1 [set] among many data outputs that will come from this trial. We felt that it was important at the earliest possible point to project the molecular genomic landscape of the population of patients enrolled in this trial.
We knew going into this that there would be wide variation in the prevalence or frequency of certain genetic alterations for which we were trying to assign patients to individual treatments. We didn’t know exactly what those numbers would be because we didn’t have a [knowledge] baseline. Partly, that is due to the fact that a trial like this has never been undertaken, but also because there are important differences in terms of this clinical trial population. All patients had to have cancers for which known effective treatments were already pursued and found not to be effective. Therefore, these patients were looking for another therapy or had a type of cancer for which no effective treatments [exist].
How could we fairly estimate the frequency in which we would find certain alterations within a population of all cancer types that have never been enlisted in a clinical trial? This is particularly true for rare cancer types that make up 20% to 30% of all patients with cancer. We just lacked information to guide us.
Publishing the genomic landscape analysis represented a first-of-its-kind clinical trial population.
The next important piece is: How do these data reflect on this approach? What is the efficacy of trying to create a one-stop shop opportunity for patients to participate in a genomically guided clinical trial regardless of cancer type? Does it work to conduct a study broadly across cancer types rather than in breast, lung, or colon cancer? Is it ultimately inefficient? In some ways, those were some of the biggest questions this trial aimed to address.
The trial indicated that nearly 38% of patients harbored an actionable alteration. How was actionability defined and how did you match those patients to treatment?
Actionability is a term that wasn’t used very much many years ago regarding any kind of cancer diagnostic testing. It related to performing NGS, which can now be done broadly in the oncology field through commercial laboratories or [particular] centers.
Generally speaking, what we meant by actionability was that the frequency with which we could find alterations that pointed to known effective therapies. That wasn’t what we were studying in NCI-MATCH, but above that, is the fraction of patients for which we could assign to a treatment arm or one of a dozen clinical trials being conducted in parallel.
Therefore, 38% of patients had a genetic alteration or molecular feature that pointed to an approved therapy for the patient’s specific cancer type or an investigational arm [of a clinical trial].
For example, at the time that we launched this trial, BRAF-targeted therapies were only effective in melanoma. If we found a melanoma patient who had a BRAF mutation, we weren’t going to put them on NCI-MATCH because [BRAF/MEK inhibition] is already a known pairing.
I should also mention that there are a handful of examples where drugs were in late-stage development for a given cancer type. For example, the commitments for PI3K inhibitors in breast cancer harboring PIK3CA mutations were already made with multiple ongoing phase 3 trials. We didn’t need to do a phase 2 trial of something that had already gone through that level of testing and [would be] launching into multiple phase 3 trials. Even though those drugs were not FDA approved at the time, we still considered that ship to have sailed [to test them in this setting]. Therefore, we didn’t reinvestigate that concept.
[Regarding the] aggregate number of 38%, consider that we are talking about a group of patients for whom a known effective and approved treatment [was available] for patients with that cancer type and molecular feature. Central to the conduct of this study, the proportion of patients who had alterations that pointed to one of our treatment arms.
What were some of the most common genetic abnormalities that were identified and in what tumor types were they discovered?
We did centralized testing where we used 1 testing platform across 4 laboratories that were operating in a coordinated fashion. In conducting NGS in nearly 6000 patients, we did expect that we would “rediscover” the most common genetic alterations. What about the less common ones? What about the rare ones? By tumor type, we had relatively little guiding information. However, it wasn’t a surprise to us that we found, for example, p53 [mutations], which were estimated to be present in roughly 50% of cancers. In our population, we did in fact find p53 mutations in roughly 50% of patients.
Next on the list was KRAS, which has garnered a fair amount of attention recently. We’ve known that KRAS mutations are present in [up to] 25% of patients with cancer. KRAS is common, and it has been a point of major frustration. It is an activated oncogene unlike p53, which is an inactivated tumor suppressor gene. We wish that we could drug KRAS. Just recently, a bit of a celebration happened around some early clinical trial results for the subset of KRAS G12C–mutant cancers. Unfortunately, that is a minority of KRAS mutations, so the majority of these alterations cannot be targeted. I also need to mention that we designed this trial before those drugs were even in clinical testing, so unfortunately, we couldn’t include KRAS G12C-targeted drugs.
Then came PIK3CA mutations, which are the third most common genetic alteration we found. We did have treatment for patients who had that mutation present. In our analysis, this was about 14% of patients. PIK3CA mutations are very common in patients with breast cancer, but mind you, we didn’t take the patients from ongoing phase 3 clinical trials and enroll them in the NCI-MATCH trial. However, we took all other cancer types and found that PIK3CA mutations are quite common in head and neck cancer, endometrial cancer, colon cancer, and others.
[We also identified] a couple more tumor suppressors and a mixture of activated oncogenes for which therapies were in development for.
What are the next steps for this research? Could NCI-MATCH serve as a prototype for future trials?
Some very logical next steps will come out of this work. Using a genomic platform to triage patients to multiple therapies could be continued in the same vein, [such as] reconsidering or refining which molecular alterations/therapy pairs are most fruitful to explore in this way.
I would suggest a couple of alternatives. One would be to go further down the path of common cancer types and delve further into matching. Imagine a scenario where in a certain cancer type, we might not be interested in broadly pursuing combinations of a novel targeted therapy with a standard, classic chemotherapy backbone. However, that combination could be highly pertinent in breast, lung, or colon cancer. We can imagine tailoring [treatment] for particularly common cancers. This is the type of study where we [should] begin to introduce “backbone therapies” or combination regimens that are tailored to specific tumor types.
If we flip that, we could [enroll patients] by identifying the most common genetic alterations. Perhaps the most efficient thing to do would be to delve further into PIK3CA-mutant cancers, which as I said are quite common. That alone could be an efficient place to investigate multiple new regimens, such as next-generation drugs or combinations, within that common genotype.
A third variation could be to [evaluate] just rare tumors. Never before in the history of therapeutic development, NIH-supported or otherwise, have we had a large investigational opportunity for patients with individual rare cancer types to undergo this type of testing and be subsequently triaged to clinical trials. That is a major unmet need because 20% to 30% of patients with colon cancer have a rare [subtype of cancer].
All of those are permutations that we could think about. Escalating to combination therapy is a very obvious mandate in the field; cancers are complicated. The idea that we are going to be able to produce deep and durable responses with single drugs is something that we didn’t think about at the dawn of my career 20 years ago. Now, in 2020, that is still the case.
The NCI is mounting 2 trials called I-MATCH, which includes immunotherapy combinations and immunotherapy/targeted therapy combinations, and COMBO-MATCH, which is [evaluating] purely targeted therapy combinations. As I suggested, I believe there are other permutations that could be valuable to the field.
Flaherty KT, Gray RJ, Chen AP, et al. Molecular landscape and actionable alterations in genomically guided cancer clinical trial: National Cancer Institute molecular analysis for therapy choice (NCI-MATCH). J Clin Oncol. Published online October 13, 2020. Accessed November 3, 2020. doi:10.1200/JCO.19.03010