WASHINGTON – At this week's annual meeting of the American Association for Cancer Research, a number of similarities emerged among labs offering cancer gene panels on next-gen sequencing platforms. For example, several groups that offer such services reported that lung cancer is the most common tumor type they see, followed by breast cancer and colorectal cancer. Likewise, these labs all see the same top three mutated genes: TP53, PI3K, and KRAS.
But these groups diverge a bit when it comes to deciding what clinical action to take based on the information gleaned from such tests. Academic labs increasingly view NGS-based cancer panels as a cost-effective — and tissue-saving — alternative to running a series of multiple single-gene companion tests, but they are bumping up against challenges in clinical decision-making because the results invariably lead to regulatory questions about the off-label use of targeted drugs. While there have been a number of published reports showing that sequencing results led to a dramatic response to an off-label use of a targeted therapy, the success rate of this approach is still unclear.
Most speakers at the conference who discussed their NGS-based cancer sequencing tests said that they provide oncologists with information on "actionable" mutations to help them determine the appropriate course of treatment using targeted therapies. However, it turns out that different groups have varying definitions for "actionable."
This question was raised in the meeting's opening plenary session by Charles Perou, a professor of genetics, pathology, and lab medicine at the University of North Carolina's Lineberger Comprehensive Cancer Center, who noted that while he believes next-gen sequencing will eventually become part of cancer care, he is concerned that genomics knowledge is currently "ahead of our ability to therapeutically target" tumors since many mutations identified by sequencing are in targets for drugs that aren't yet on the market.
He said that there's still an open question around what should be considered clinically actionable, noting that this category can not only include drugs that have been approved by the US Food and Drug Administration for the tumor that is being studied, but also FDA-approved drugs for unapproved tumor types, as well as drugs that are still in clinical trials.
Perou noted that while sequencing can identify druggable targets, it's up to the clinician to prioritize potential treatments. Referring to a case in which a patient passed away before getting her sequencing results, he said that the data indicated that she could have received three possible personalized interventions. "I very much doubt we would have given this patient the three drugs at once, and then the question is, 'Which of the three should we have given her?' I don't know."
In an AACR session on next-gen sequencing, Shashikant Kulkarni, medical director of Washington University's Genomics and Pathology Services group, said that his team considers "actionable" to include FDA-approved drugs for the patient's cancer type, as well as off-label use of FDA-approved drugs. Using these criteria, he said that about 45 percent of the cancer patients his group has sequenced to date have actionable mutations.
Meanwhile, Vincent Miller, senior vice president of clinical development at Foundation Medicine, explained that the company has a broader definition of actionable mutations, which includes those for which there is an FDA-approved targeted therapy in the same or in another tumor type, or a clinical trial for a therapy targeting the alteration. Based on this definition, he said that the company's FoundationOne assay has identified at least one actionable mutation in 76 percent of the more than 2,220 samples it has run to date.
In another session on NGS-based cancer testing, Arul Chinnaiyan, director of the University of Michigan's Center for Translational Pathology, said that people ask him all the time what portion of patients receive a clinical recommendation based on the results of his group's MI-OncoSeq assay. "It depends how you define actionable," he said. Out of the 150 patient samples his group has sequenced so far, he estimated that about 50 percent had "actionable" mutations, though a smaller number were actually directed to a specific therapeutic regimen based on the results of the test.
In a panel discussion on regulatory considerations for cancer biomarkers, Elizabeth Mansfield, director of personalized medicine at the FDA's Office of In Vitro Diagnostics within the Center for Devices and Radiological Health, noted that "people use 'actionable' in many different ways," and added that she thinks of the term as "something that really has evidence."
"It's my impression that we have a lot of information but not a lot of evidence," she said. "We can get, on an individual level, a clue that a particular cancer has a mutation or several mutations that might look like drivers, and there's a drug approved that might target that, but it doesn't always work out that simply."
Mansfield said she likes Foundation's approach of sending patients into clinical trials based on their mutation status, but she's wary of treating patients with off-label drugs based solely on this information. As far as gathering more evidence to help inform clinical decision-making, "it's a question of how you're actually going to do it," she said. "Are we still going to have clinical trials that look like clinical trials or are we going to try to aggregate information from lots of N-of-1 experiments?"
Foundation's Miller agreed that the community needs "mechanisms to annotate what happens when doctors choose an off-label approach and the patient doesn't respond." One issue, he said, is that there's an "inherent bias" to publish positive results. "So we might see the N-of-1 who has a TSC mutation who gets an mTOR inhibitor and has a [complete response] and the paper ends up in a great journal, but where's the denominator?"
John Palma, director of medical affairs at Roche Molecular Systems, said that this evidence will be particularly necessary for low-frequency cancer mutations for which there is little incentive for pharmaceutical companies to design clinical trials because the populations are so low.
"I think the avenue is to collect all of the information for decisions that are being made … so we can start to build the body of evidence around particular markers that are low frequency," he said in the panel discussion. "You can't do a trial [for] any of these particular types of mutations, but collectively we can use that information to help bolster the value proposition that any one of them provides."
Waiting for a Sponsor
The regulatory environment for NGS-based cancer panels is still murky. Several participants at the panel discussion and other sessions noted that CLIA labs can easily include in sequencing panels genes for which the FDA has approved single-gene companion diagnostics. This allows them to survey a patient's cancer-related mutations in one assay instead of a series of tests, which saves time as well as precious tissue. In the example of lung cancer, for instance, patients are currently tested for EGFR, KRAS, and ALK, but new evidence suggests that mutations in other genes such as BRAF, ROS, and MET can help guide therapy. Labs that comply with current FDA regulations and test for these mutations gene-by-gene will ratchet up their costs and quickly run out of tissue.
Mansfield acknowledged that the community is outpacing FDA's ability to provide guidance with regard to gene panels.
"The way we've set the companion diagnostic model, it's pretty much, 'Come forward with one test for one drug.' So if you did want to do multiple tests on one tissue, it does present the problem of running out of tissue, and it would be much more efficient to run all possible tests on one platform at one time," she said. However, she added that the FDA needs to address a number of questions with regard to such tests: "How would you actually label that test? What would the evidence be behind that test? Would it be a companion diagnostic?"
Mansfield added that these questions are actually "ancillary issues that just have to be decided." The important thing, she noted, "is actually having the test and ensuring that it really works, so that if I've got six different drugs for lung cancer with six different biomarkers behind them, I can be sure that the test is telling me that I have one of those biomarkers or not."
While the FDA is "very interested" in next-gen sequencing as a companion diagnostic, "our problem is that we can't approve something that doesn't get submitted to us. We don't have the power to reach out and say, 'We want a next-gen sequencing test to come in here right now.'"
The agency is currently "waiting for someone to come forward with this technology," she said. "We agree that it offers great efficiency over individual tests, but it requires a sponsor to come forward and propose it and to give us data to review. And as soon as that happens, we'll be moving forward."
Mansfield said she is confident that the agency will find a way to address these questions. "Frankly, we at FDA can't stick our heads in the sand because Foundation and others are already doing this. So we've got to find a way to give patients the best treatments that they can get."