Important points were identified and potential solutions to the companion diagnostic and drug development pathway were offered by individual workshop participants and speakers throughout the day (see Box 6-1). These issues included the type of regulatory pathway that should be used for companion diagnostics, how clinical utility evidence could be generated, which payment models were likely to succeed, and how NGS could further transform thinking about companion diagnostics. The speakers identified several goals that, if met, could help facilitate the use of companion diagnostics in the future (see Box 6-2).
COORDINATING REGULATORY PATHWAYS
The discussion of IVDs and LDTs again raised the question of the economic incentive to develop companion diagnostics. As McCormack said, “The whole model will collapse unless the playing field is made level. Why would diagnostic companies want to invest years … and tens of millions of dollars to lose it shortly after you cross the finish line? It just doesn’t make sense.” Thompson agreed that this is an area for concern because in a system for generic diagnostics, FDA initially approves the companion diagnostic, and then a lab could decide to make a very similar test, validate it, and then sell it as a companion diagnostic for an associated drug. The pioneer of the test achieves FDA approval, but others can produce the same generic test that is subject to CLIA instead.
Several speakers reiterated the need to demonstrate the equivalence of IVDs and LDTs, in part through a follow-up of patients. CLIA is a law,
BOX 6-1
Possible Solutions for Co-Development Challenges Suggested by Individual Speakers
Workshop co-chair Debra Leonard highlighted suggested solutions and potential action items presented at the workshop by individual speakers.
• Regulatory oversight of LDTs is needed. This could be accomplished through a more robust CLIA accreditation process, a risk-based approach to regulation by FDA, or by choosing one path to assure the safety and effectiveness of all companion diagnostics, regardless of their source.
• Providing a mechanism to regulate follow-on tests may help provide more comparable economic incentives for IVD and LDT developers to produce diagnostics.
• In predicting drug response, rather than a particular test being specified, the companion diagnostic analyte and the performance characteristics of a test should be specified.
• The same molecular tests to diagnose and classify diseases could also be used for patient selection for drug clinical trials.
• Well-defined standards of evidence for the clinical utility of tests and the process for generating that evidence are needed for determining intended uses.
• It will be necessary either to demonstrate the equivalence of LDTs to FDA-approved companion diagnostics or else to rigorously monitor LDT performance, by making LDT performance data publicly available.
and thus LDTs are not going to go away or be subsumed under a single regulatory system without congressional legislation to do so, said Leonard. She recognized the need for greater regulatory oversight of LDTs, though she admitted that this would still not “level the playing field” between IVDs that go through the FDA approval process at significant expense and LDT developers. One possibility would be to formalize regulatory oversight of LDTs through a CLIA laboratory accreditation process, she said.
If a more robust CLIA is indeed a solution for regulating companion diagnostics, then part of that solution should be to eliminate FDA’s role in regulating the tests because it would be duplicative, Thompson said. If an improved CLIA is “enough to assure safety and effectiveness of these tests, why wouldn’t it be enough for all tests?”
One broadly discussed approach to demonstrating the equivalence of IVDs and LDTs, as well as to reducing variability in diagnostics, was the establishment of local or national testing and outcomes data repositories that would be used to gather and generate evidence to improve patient care. A workshop participant said that FDA has been discussing the possibility
BOX 6-2
Goals for Companion Diagnostics Identified by Individual Speakers
Workshop co-chair Debra Leonard summarized the goals for the future as presented by individual speakers during the workshop discussions. These goals were
• The development of a single test that could be used to simultaneously make a diagnosis, indicate treatment, and assess the adverse reaction risk for drugs upon clinical validation for each use.
• The creation of a global, value-based payment system for companion diagnostics (including next-generation-based testing) that would be based on evidence that considers overall patient care and achieving specific outcomes.
• A learning health care system, that, with research and payer support, uses clinical data to improve patient care moving forward.
• The establishment of a national testing and outcomes database to generate evidence for improving clinical care.
• The implementation of regulatory guidance for next-generation sequencing that would allow for the development of tests and would both direct patient care and be used for drug trials.
• The development of an FDA process to alter the drug label to account for cleared or approved new tests for existing or new drugs that would provide an alternative to requests that are now driven only by pharmaceutical companies.
• The institution of a new reimbursement method that would account for next-generation-based testing to provide more patient data at the same or at lower cost than multiple, individual diagnostic tests.
of such a database as a useful regulatory tool. Well-curated resources that contain genotype–phenotype information and other sorts of evidence would be extremely useful. It would also be practical, because similar resources are being developed for other purposes. A database could be used both for clinical interpretation of test results and for discovery, Leonard said. However, many laboratories may be reluctant to put the evidence that they generate into such a repository. Buller suggested starting small before deciding whether to increase the scale of the project and also creating incentives for laboratories to participate.
McCormack asked whether FDA should first triage all tests to determine which ones could be overseen by CLIA and which ones would need FDA approval. This approach would alleviate the burden of asking FDA to review the devices with less risk. A risk-based strategy would require
coordination between FDA and CLIA pathways to determine which tests go down which pathway, said Burke.
Gutman questioned whether FDA would be able to perform such a triage function, given its limited resources. In order to perform in such a way, FDA might need to use classification panels for assigning risks, and this process would likely need several years to be phased in, he estimated. A master file to record all tests could be established by an organization, with an independent body then suggesting which tests need the full approval process. However, no clear suggestion emerged of which organizations should be responsible for these roles—whether FDA, CLIA, or another body. Such lists are already maintained through CMS so that laboratories can be reimbursed for these tests, noted Victoria Pratt, chief director of molecular genetics at Quest Diagnostics at the time of the workshop, but the lists are not public.
Reducing Test Variability
Given how variable the results of tests can be, both tests performed in the same laboratory and tests done in different laboratories, the accuracy of all testing needs to be improved, said Buller. One possibility, he suggested, would be to create more centralized models for testing, perhaps involving consortia of hospitals. This might be especially useful in light of evidence suggesting that test accuracy improves with the number of tests a laboratory does. Another approach would be to define performance metrics for testing and then reward good performance and data transparency. Mansfield mentioned the possibility of establishing laboratories of excellence around the country where particular diagnostic tests would be sent. This would reduce variability because all laboratories would perform the same tests and could have the same performance levels.
GENERATING EVIDENCE FOR TEST VALUE
Leonard and Burke agreed that generating good evidence about how to get the right drug to the right patients at the right time remains a major issue and that randomized controlled trials will not be possible in all circumstances, which will require new study designs. Neither CLIA nor FDA formally assesses clinical utility, although payers are very interested in this issue, Leonard said. Payers try to make evidence-based decisions based on cost effectiveness and the quality of patient care and outcomes. However, payers generally do not have a coding system that allows them to distinguish between the use of a test in a setting where good evidence exists for the test’s utility versus settings where such evidence does not exist. As a starting point, there is a way to generate robust evidence in the absence of a clinical trial by studying negative predictive markers to demonstrate negative clinical utility,
Robson said. The national testing database that was mentioned by Mansfield would also be a way to generate such evidence for determining test value.
Armstrong emphasized the need for high levels of evidence and standards surrounding tests, including LDTs, especially because even well-established tests can generate variable results. However, Mansfield raised the issue of follow-on tests where, because of a potential lack of clinical trial samples, it may be difficult to tell whether a test is better or worse than the original test. Buller agreed that retaining samples is usually unrealistic, but he also pointed out that discordant results are essentially doing the comparison. Is a false positive or false negative really false? Pfizer is making an effort to get clinical data whenever it supports a platform comparison in a particular country so that it can investigate these issues.
Incentives are needed for the development and use of tests that improve patient outcomes and that move toward value-based payments, Burke said. Sharon Terry of the Genetic Alliance observed that value-based pricing for diagnostics is going to be an “extremely steep climb.” Such pricing is unlikely to emerge from either the public or private sectors, she said; rather, it will emerge through business-to-business transactions among, for example, a diagnostic company, a pharmaceutical company, and an ACO. “Those are the only economic engines that are going to drive differential reimbursement,” she said.
One of the barriers to value-based pricing is that the recipient of the value is not the person paying for the value, a workshop participant said. Given that situation, shifting health care costs toward individuals, as has been occurring in the recent past, could have its benefits. It could bring more rationality into the system if the recipients of health care have a louder voice in deciding what they are willing to cover out of their own pockets.
The development of national guidelines for clinical utility would be beneficial for the health care provider community in a value-based payment model, the workshop participant said. In a typical third-party-payer model, the payer makes the decision about reimbursement; however, in a bundled payment model, the payer and the provider have input on the decision making over reimbursement, another participant said. In the case with the bundled payment, if the payer does not have sufficient expertise in genetics, then the determination of value could rest on the provider. Armstrong said that an individual provider should not be relied on to take on all of the responsibility, but that it could be possible to make use of the collective expertise within the health care system for decision making about payments.
Next-generation sequencing will be a disruptive technology, and intended uses and proper interpretation will be critical, Gutman said. Armstrong suggested that next-generation sequencing will initially be used in clinical areas where it already makes sense, such as in the evaluation of infants in the newborn intensive care unit. She also pointed out that Aetna has a very small number of people who understand the detailed potential and problems of next-generation sequencing. “It would be good to get help,” she said. “But we still have a fiduciary responsibility to administer a plan of benefits that a plan sponsor wants us to administer on their behalf. So we have to stick with technology assessments and evidence.” Payers are not research organizations. They are claims payment organizations. They may study a few issues, but the full range of what needs to be evaluated is immense. “Health plans are not the solution to fill all these evidence gaps that exist,” she said.
Finally, Swatkowski offered the perspective that companion diagnostics may be an interim step to understanding disease and mutations that are unique to particular patients. An all-encompassing diagnostic test would define “diagnosis, prognosis, and adverse reactions,” and that is the NGS platform, she said.
Standards will be needed as next-generation sequencing gathers momentum, said Koch. Roche has begun doing next-generation sequencing, and it is finding a great deal of variation across platforms and analytical tools. “To ensure that we do the right things for patients and have accurate results, standards will be required, whether [tests] are LDTs or [FDA-approved] IVDs,” Koch said, and he cautioned that next-generation sequencing is incredibly complex. “We have a challenge that is beyond technology here,” he said. “It’s really about how to understand the biology and appropriately translate it into something meaningful for patients.”