The FDA, in its approach to companion diagnostics, seems to favor single-assay platforms while the field moves toward multiplexed platforms, says Dr. Marc Ladanyi, here at Memorial Sloan-Kettering Cancer Center with colleagues involved in BRAF mutation testing, Maria Arcila, MD (center), and Laetitia Borsu, PhD.

December 2011
Feature Story

Anne Paxton

Take a fatal and nearly untreatable cancer. Invest hundreds of millions of dollars to develop and test a drug that shrinks tumors or improves survival in half of patients with a common mutation. Link diagnosis to a PCR test that you also manufacture. And win an accelerated FDA approval for the test-drug combination.

Altogether, no mean feat. And it’s essentially what Roche accomplished to bring its metastatic melanoma drug Zelboraf (vemurafenib) on the market this year.

By all accounts, Zelboraf should bring new hope to thousands of melanoma patients and perhaps a billion dollars a year in revenue to Roche. “It’s the first ever joint FDA approval of a drug and a DNA-based companion diagnostic, and the beginning of the molecular companion diagnostics era,” says John W. Longshore, PhD, director of molecular pathology for Carolinas Pathology Group, Carolinas Medical Center, Charlotte, NC.

Zelboraf, a kinase inhibitor, has been proven effective only for melanoma patients with the BRAF V600E mutation. But the linkage of Zelboraf to the Roche Cobas 4800 BRAF V600 Mutation Test is creating something of a furor. In its Aug. 17 approval of Zelboraf, the FDA’s Center for Drug Evaluation and Research specified that the drug is indicated “for the treatment of patients with unresectable or metastatic melanoma with BRAF V600E mutation as detected by an FDA-approved test” (emphasis added). So when the FDA Center for Devices and Radiological Health granted pre-market approval for the Cobas test the same day, the Roche assay became the de facto companion diagnostic for Zelboraf.

That pairing leaves an estimated 50 to 100 laboratories, which have been using laboratory-developed tests (LDTs) to detect BRAF mutations for years without FDA approval, in a state of uncertainty. The upside to the approval is that the FDA clearly recognized companion diagnostic testing as a key part of this new paradigm of targeted therapy, says Marc Ladanyi, MD, attending pathologist in the Molecular Diagnostics Service and a senior investigator in the Human Oncology and Pathogenesis Program and the Department of Pathology, Memorial Sloan-Kettering Cancer Center in New York City. “The negative aspect is that because of the approach the FDA is mandated to follow—to approve actual physical products—it has approved one specific test kit for BRAF V600E mutation, instead of setting minimum analytical requirements for the mutation test.”

In his view, “This is placing laboratories in a difficult position—and maybe down the line, an impossible position.”

Zelboraf is a highly effective drug for a disease that has had few treatment options until now, Dr. Longshore says. The Cobas in vitro diagnostic and Zelboraf drug applications were both submitted to the FDA in April, so their approval only four months later was exceptionally quick—among the fastest IVD or drug approvals seen to date, Dr. Longshore says. “It demonstrates the priority that the FDA is placing on drugs that do not just fulfill an unmet medical need, but are accompanied by a companion test to ensure that the right patients are selected for therapy.”

According to Roche, the Cobas 4800 BRAF V600 test is already “live” at five sites in the United States, among them Dr. Longshore’s Carolinas Medical Center, Clarient Diagnostic Services in Aliso Viejo, Calif.; Central Pennsylvania Alliance Laboratory in York; and Laboratory Corporation of America. Beyond U.S. borders, Roche has submitted new drug applications for Zelboraf in the European Union, Switzerland, Australia, New Zealand, Brazil, India, Mexico, and Canada.

As the new test-drug combination sweeps the country, however, simmering in the background is the fact that labs have been conducting BRAF testing for several years with LDTs, though not primarily for melanoma, says Roger D. Klein, MD, JD, chair of the Association for Molecular Pathology’s Professional Relations Committee, medical director of molecular oncology, BloodCenter of Wisconsin, and clinical assistant professor of pathology, Medical College of Wisconsin.

“Historically, the largest indication has been in colon cancer. In the investigation of HNPCC [hereditary non-polyposis colorectal cancer] in patients who have exhibited high-level microsatellite instability, BRAF mutations are associated with sporadic disease,” Dr. Klein says. “Further, BRAF is mentioned in the NCCN [National Comprehensive Cancer Network] colon cancer guidelines for consideration as a prognostic and/or predictive marker for evaluating responsiveness to EGFR inhibitors.” BRAF testing is discussed, too, in the NCCN thyroid cancer guidelines for diagnosis and prognosis.

While the majority of BRAF testing in Dr. Longshore’s laboratory has been for patients with metastatic melanoma, “We are also seeing an increasing number of referrals for colon adenocarcinoma where some reports in the medical literature have shown that the BRAF V600E mutation is associated with both a poor prognosis and a poor response to anti-EGFR therapy.” In fact, Dr. Ladanyi says, a number of centers have set up BRAF testing for colorectal cancer because of data showing that patients with BRAF mutations do not respond well to the EGFR inhibitors Erbitux (cetuximab) and Vectibix (panitumumab). “Some places don’t routinely test metastatic colon cancer for BRAF and others do. We do here, and there are many more colon cancer patients than metastatic melanoma patients,” Dr. Ladanyi says. So although the Zelboraf approval will lead to a significant increase in overall demand for BRAF testing, “for labs like ours, it may not represent a huge increase.”

With LDTs such a fixture in the laboratory world, the Cobas test could hit some speed bumps in winning acceptance. Since the BRAF mutation of interest for colorectal cancer, V600E, is the same as for metastatic melanoma, many people will already have developed PCR tests for BRAF V600E, but the FDA has approved none of those assays for colorectal cancer. “Then the issue becomes, why do we need a different assay for the same mutation?” Dr. Ladanyi says. Conversely, Roche’s Cobas BRAF test could presumably be used on colorectal and thyroid cancers, at least for the V600E mutation. But if that were to happen, the Cobas test would become more like an LDT itself. Says Dr. Klein: “You would have to validate the test as if it were an LDT. You can’t simply verify it because it’s only indicated for metastatic melanoma.”

Another important factor facing Roche’s new test-drug combination, Dr. Ladanyi believes, is that molecular diagnostic labs that are trying to extract as much genetic information from cancer samples as possible are realizing they need to be able to screen for multiple mutations at once. “It’s not going to work if we have to have not just a different assay for every gene but potentially a new assay and maybe separate technical platforms for non-V600E BRAF mutations. Obviously that would be an incredibly inefficient situation—both in terms of health care dollars and in terms of utilizing very small patient biopsy samples to get the maximum amount of genetic information to help the patients.”

The FDA, in its approach to companion diagnostics, seems to be favoring single-assay platforms, Dr. Ladanyi says, whereas the field is moving more and more toward multiplexed platforms. For example, Memorial Sloan-Kettering uses a Sequenom Mass-Array platform to do highly multiplexed mutation analysis, looking at mutations in eight genes at the same time, and Massachusetts General Hospital does a similar multiplexed assay on a different platform, he notes. “But multiplex platforms, which are now the trend at the larger cancer centers, are going to be more complicated to get FDA approval for. So we’re not sure how that’s going to work.”

Moreover, Dr. Ladanyi notes, it shouldn’t be forgotten that in many cases, people set up LDTs not because they want to avoid paying a higher price for an FDA-approved product, but because there is no FDA-approved product. “There may be no ASR options out there. In the meantime, we have oncologists who are clamoring for testing to begin, so of course labs develop their own tests to meet that need. LDT activity has been critical to progress in many different cancers, so if you create a situation where that whole process is blocked, it’s just going to be very problematic. It’s going to be a disaster,” Dr. Ladanyi says. For example, “Where would we be in terms of CML treatments if we were still waiting for an FDA-approved test for BCR-ABL?”

Dramatic findings in a study sponsored by Roche have intensified the   companion diagnostics debate. In the study, which has been submitted for publication in the Archives of Pathology & Laboratory Medicine, researchers compared results of testing by the Cobas BRAF Mutation Test and by Sanger sequencing on samples from 477 patients drawn from the BRIM2 and BRIM3 clinical trials that led to the approval of Zelboraf. In testing these samples, the researchers found the Cobas BRAF V600 Mutation Test had no failures, compared with Sanger sequencing which had a 9.2 percent failure rate, and the Cobas test showed higher sensitivity in detecting V600E mutations than Sanger, as confirmed by deep sequencing. The Cobas test also detected a majority of V600K mutations in the cohort, which is significant because patients with that mutation have shown some response to Zelboraf.

Of the data from the BRIM2 trial, Dr. Longshore says: “A significant number of samples had false-negative calls [reported as wild-type when they had a V600 mutation] with Sanger sequencing, and a similar number of samples had invalid results with traditional sequencing. Most of these were due to the inability to generate a PCR amplicon to sequence. The net effect for both of these problems with Sanger sequencing-based BRAF analysis is that about 13 percent of patients would have been denied access, or at least would have had delayed access, to Zelboraf therapy.” In contrast, only about one percent would have been erroneously deemed ineligible for Zelboraf due to a false-negative with Roche Cobas.

The surprising thing about the study (of samples from the BRIM2 and BRIM3 trials) was that Sanger sequencing gave a false-negative in 20 percent of the samples, says Kenneth J. Bloom, MD, medical director, Clarient Diagnostic Services, who co-wrote abstracts of the study presented at the American Society of Clinical Oncology meeting in Chicago in June and the European Multidisciplinary Cancer Congress in Sweden in September. “The Sanger sequencing detected no mutation in eight percent of cases, and in another 11 percent of cases it didn’t get any results at all. That was a shock, because when this study started, Sanger sequencing was going to be the gold standard.”

Those results didn’t surprise Dr. Ladanyi. “We know for many different genes and mutations that Sanger sequencing is not an especially sensitive method and is probably the least sensitive detection method around. In most cases you need at least 25 percent tumor cells in your sample, and for some cancers that’s a problem. Most people are trying to use tests that can detect cancer cell mutations even if the cancer cells make up only one to 10 percent of the overall cells in the tissue, and I think that’s about where the sensitivity of the Cobas test is.”

At Sloan-Kettering, two methods are used. Sequenom, which is mass spectrometry-based genotyping, is the main method and somewhat more sensitive than Sanger sequencing. “Our secondary, even more sensitive method is Locked Nucleic Acid PCR sequencing [Arcila ME, et al. J Mol Diagn. 2011;13:64–73], which is a way to suppress amplification of normal allele by using a specially modified nucleic acid called a locked nucleic acid during PCR, followed by Sanger sequencing of the mutation-enriched product,” Dr. Ladanyi says. “The oncologists know that our in-house BRAF V600E mutation testing is more sensitive than conventional Sanger sequencing and they’ve not requested the Cobas test.” But where unmodified Sanger sequencing is used, if oncologists see the data showing Cobas is more sensitive, “they will ask for the Roche test, and that’s understandable.”

Jan Nowak, MD, PhD, director of the Molecular Diagnostics Laboratory at NorthShore University Health System, Evanston, Ill., views the comparison study findings as not significant, and he is skeptical as well of the sensitivity and turnaround time claims Roche has made. “They set it up to make Sanger sequencing look bad, by using specimens that had fewer than 20 percent tumor cells, then saying ‘Our test is more sensitive.’” But that’s a bit of a straw man fallacy, he believes. “If you choose specimens that have appropriate cellularity, that have 20 percent tumor cells, Sanger sequencing is probably at least as sensitive and probably more specific. It’s true that because the Roche assay is a PCR assay, it can detect the mutation in specimens with fewer tumor cells, but that doesn’t mean that it’s really more sensitive in a functional way.”

Dr. Nowak’s laboratory uses an LDT very similar in design to Roche’s test, with a 24-hour turnaround, and he believes it’s a better test than the Cobas because it can distinguish non-V600E mutations. He thinks claims of a faster turnaround time and low cost may need to be taken with a grain of salt. “While the Cobas test, once you push a button, may only take two hours, in reality you aren’t going to be doing the test every day on a single specimen.” To achieve a per-test cost of $120 to $150, which Roche says is possible with the Cobas test, “you’re going to have to batch the tests, and you need a piece of equipment that costs $60,000.” The volume of melanoma testing in his lab is maybe one a week. “So if they came in and put their instrumentation in the lab for a volume of one or two specimens a week, what would be the cost of doing runs? I suspect it’s going to be expensive.”

One downside of the kind of targeted PCR testing that the Roche study found superior, Dr. Klein notes, is that “there are other mutations even for melanoma. I don’t know that we fully understand their impact, but in general, Sanger sequencing does a better job of detecting other potentially important mutations, because it’s not looking at a specific target; it’s looking generally to see if there is a mutation. There are other mutations you’ll pick up with Sanger sequencing that you’re not necessarily going to pick up with the Cobas.”

There are also a number of questions about the clinical specificity of the Cobas test, Dr. Klein suggests. “The main issue is we generally don’t know, when we’re giving a drug that is postulated to inhibit a mutation-containing gene, what the effect is of allele proportion of the mutation on responsiveness. It could be that there is a differential effect of the drug based on mutation level, or a threshold allele proportion above which patients respond. It could also be that a significant percentage of patients in whom a mutation is not detected will also respond. If enough patients without detectable mutations respond, it may even be a mistake to rely on the test.”

“Moreover, the percentage of neoplastic cells in the sample from which the DNA is extracted is going to relate to the analytical sensitivity, because if you have a lot of stromal or other types of cells in there, your allele proportion is going to decrease,” Dr. Klein points out. In practice, this depends on proper performance of enrichment procedures like macrodissection or microdissection. Although there is no question allele-specific PCR will have a lower limit of detection than Sanger sequencing, it’s possible Roche’s comparatively superior analytical results for the V600E mutation with the Cobas test are explainable by some other factor, he says.

“Low-level mutations can be caused by two things: a low percentage of cancer cells that have the mutation because of cellular heterogeneity, or a low percentage of cancer cells in the specimen. So a possible explanation for the apparent increased V600E detection rate of the Cobas test relative to Sanger sequencing was actually the specimen quality with respect to neoplastic cellular content.” As a result, he is reserving judgment on whether Sanger sequencing is less effective than Cobas. “There are a lot of unknowns.”

Dr. Klein says he does not agree with Roche’s efforts to limit BRAF V600E testing to the Cobas test. There are advantages to using the Cobas test, he concedes. “Because the Cobas is standardized and tends to have safeguards making the test more straightforward to use, I think it potentially broadens the population of pathologists who can get involved with the testing, which is a good thing. The less specific molecular technical knowledge that a pathologist needs in order to perform the test, the more pathologists who can do it. But it’s not a good thing to force pathologists and laboratories to use their product.”

Since Clarient was one of the three U.S. sites included in the comparison study, “we got to actually run the Cobas and found it very easy to use, not labor-intensive, and straightforward,” Dr. Bloom says. “Any certified molecular lab should be able to run this test. The real issue for most pathologists has been the need for a Cobas instrument. If a pathology lab owned a Cobas instrument, it was generally in microbiology, not in the AP section. I’ve heard many pathologists say, ‘I already own an ABI instrument or an Illumina instrument, so do I really have to purchase a Cobas just for this one assay?’ What happens if the next companion diagnostic comes out on yet another platform?”

A large reference lab like Clarient can afford to acquire many instruments, Dr. Bloom says. “But it is not practical for most institutions to have four or five platforms running in their lab. It is a challenge at times to keep up one platform, let alone five. So I don’t see this as being a positive trend for pathology and I can see why a hospital-based pathologist would be greatly concerned. If I were a medium-sized hospital lab without access to a Cobas, this would be devastating.”

Nevertheless, Dr. Bloom doesn’t believe it’s realistic to stop linking a companion diagnostic to a specific vendor. In fact, he says, “What this study showed is you have to link to a specific vendor’s product.” Before this study, “Sanger sequencing would have been something that we and the FDA would never have given a second thought about. Sanger was supposed to be the gold standard to prove the Cobas was doing its job. Many pathologists and oncologists believe that Sanger sequencing is the gold standard for assessing mutations, and have a problem understanding how classic sequencing, where you find every mutation, could perform worse than PCR where you’re identifying a specific mutation. But in point of fact, it was in these studies.”

He has many LDTs in his lab, including BRAF testing for thyroid, lung, and colon cancer, but he believes the Roche study reminded everyone—by accident—that better analytical validity does not establish better clinical validity. “Everybody wants to say, ‘I can find that target in my own way with an LDT,’ but that doesn’t address the fundamental issue: When you call it a positive, does the patient respond to therapy?” In considering the Cobas BRAF-Zelboraf combination, he says, “I imagine the FDA had to grapple with the fact that an apparently sure thing, Sanger sequencing, took 20 percent of patients who might have benefited from the drug out of the picture. I think that by including the requirement for the detection of the V600E mutation by an FDA-cleared assay in the labeling for vemurafenib, the FDA was emphasizing that when it comes to clinical performance, the proof is in the pudding.”

This message from the FDA may bode ill for some LDTs, Dr. Bloom believes. “I think the FDA has drawn a line in the sand and said, ‘You’re going to have to show me clinical performance.’ Maybe not for all LDTs, but likely for those used to select patients for targeted therapies. And that puts the biopharmaceutical industry and pathologists between a rock and a hard place. Clinical trials can take several years to complete; therefore, the diagnostic test used to enroll patients in pivotal trials may be several years old by the time the trial results are presented. During the time of the trial, labs may have developed tests or methods that have superior analytic performance, but the clinical performance of the new test will remain unknown unless the clinical trial is re-done enrolling patients based on the new test result.” Dr. Bloom predicts there will be more and more pressure on labs to have real clinical data if they want to make clinical claims about the performance of their LDTs.

The logistics of adopting a new test will pose other challenges. The Cobas-Zelboraf combination is unlike Abbott Diagnostics’ Vysis ALK FISH test for Xalkori (crizotinib), Dr. Ladanyi points out, because the Cobas BRAF test is brand new. “The Abbott test was already the preferred test being done at many centers. So Abbott just got approval for a test that was already being used by just about everybody.” The same was true of HER2 in breast cancer. “The reagents for FISH or IHC were already being very widely used when the FDA approval came, so it was less of a controversy. For Zelboraf, I think it will be interesting to see what happens, because the test that got approved is not a widely used test, and again there’s the cost of acquiring the platform. That’s significant.”

Hovering over the scientific and policy debates about LDTs is the role of reimbursement, because insurers could theoretically start denying coverage unless a Cobas test is used. “That seems like it would be a little odd, because in all likelihood the Cobas test will be a little more expensive and maybe a lot more expensive than LDTs. So in a way, it would be counterproductive,” Dr. Ladanyi points out.

Clinicians can always order non-FDA-approved tests, but there can be substantial risks to that, Dr. Bloom says. “The eligibility criterion for the drug states that you need an FDA-cleared assay, so if an oncologist put a patient on the drug without such testing, they are technically off-label. The payers have up to 18 months to say, ‘Even though we’ve paid for this drug in this patient, that patient was not eligible and we want our money back.’ Whoever paid for the drug may be on the hook, and that can run into millions of dollars depending on the size of a practice.” He believes this has happened with other drugs before and has bankrupted some oncology practices in the past.

How many laboratories that have their own BRAF LDTs are likely to switch to the Cobas test? Says Dr. Klein: “My guess is that some labs may voluntarily switch, particularly those that have capital budgets to do it. It depends to some extent on the laboratories’ relationship with their treating physicians and the pressure they get in that context from the hospital.” In his experience, “Most treating physicians have trusted pathologists to use our medical and technical judgment. I don’t think they typically care whether something is FDA-approved—nor would they even necessarily know.” Another factor is the price of the Cobas instrument, estimated at about $60,000. “Doing the Cobas test requires purchase of an instrument and prepackaged reagents, and generally, depending on your volumes, LDTs are going to be significantly less expensive.”

The costs of an LDT for BRAF are less than the Cobas test, but there are significant advantages to choosing an FDA-cleared test from a performance standpoint, Dr. Longshore believes. “The sample preparation, nucleic acid isolation, PCR setup, hardware for running the assay, and the interpretation guidelines have already been evaluated during the clinical trials required for a 510(k) or PMA submission. Additionally, the analytical sensitivity, cross-reactivity, interfering substances, and specificity have been determined and validated during this process.”

While any change in method can be costly, Dr. Longshore adds, “this is a task that occurs often in molecular labs due to testing improvements and changes in versions of cleared tests.” He believes that, over time, more labs will choose to adopt FDA-cleared testing methods as they become available for assays that previously have been performed as LDTs.

For the time being, the College has expressed concern about a draft guidance  the FDA released this year, “In Vitro Companion Diagnostic Devices,” which lays out the agency’s current thinking on how LDTs should be regulated. Restricting the use of LDTs, as the draft guidance proposes, would disrupt patient care, because LDTs play an important role in advancing modern medicine, the College pointed out in its comments to the FDA. Under CLIA, each clinical laboratory is required to verify the analytic validity of every diagnostic test by establishing analytic sensitivity, analytic specificity, accuracy, precision, reportable range, and reference intervals. “Further, for CAP-accredited laboratories, all molecular testing systems must have clinical validity established as well.”

“Virtually all new laboratory tests are developed as LDTs before being commercially developed,” the College noted. The CAP recommends that therapeutic labels refer to the analyte being tested or performance characteristics required rather than to any specific tests that are FDA-approved. “Pathologists, medical directors of laboratories, should be able to continue to choose to use the test that will provide the best diagnostic information for their patients. Often these tests are LDTs.”

The Association for Molecular Pathology has expressed similar concerns, Dr. Klein says. “The draft guidance doesn’t appear to be directed toward laboratories, although it’s not clear what it may signal for labs in the future. Moreover, it certainly could put pressure on at least some places to use the FDA-approved test.”

The FDA stance on LDTs remains unclear, but the Cobas-Zelboraf approval puts a qualitatively different pressure on the agency, Dr. Ladanyi says, to resolve these LDT issues. “There is concern that some kind of reporting to the FDA will be required, and the question is how manageable that’s going to be for the majority of labs in terms of time and money.” Potentially labs could be obliged to switch tests for no real scientific reason, Dr. Ladanyi says, “just a purely regulatory reason and a potential reimbursement reason.”

Dr. Nowak, too, rejects the idea that tests have to be approved by the FDA. “If payers insist that labs do a particular test because it’s referred to in the package insert, that’s going to fail, because BRAF testing is important for thyroid and for colon cancer, and the Cobas test is not approved for those mutations. People have been doing the same tests as LDTs for years with good quality, and if you turn it into a black box kind of thing, where you have to use one particular assay, it becomes immediately apparent when someone wants to test for that same analyte, a BRAF mutation, in a thyroid or brain or colon cancer, that it will have to be done as an LDT.” Dr. Nowak believes that the wording in the FDA package insert should be “as detected by an FDA-approved test or other assay that has similar performance characteristics.”

Dr. Klein is not predicting a dramatic move by the FDA to regulate most LDTs in the near future. “There are a lot of political dynamics. Moreover, what can the agency realistically accomplish given its current resources and personnel? Does FDA even have the legal right to regulate pathology practice? There’s a lot we don’t know. CAP accredits 7,000 labs and most of them perform LDTs. I don’t know how the FDA would regulate them all.” LDTs have historically been an important part of pathology practice, Dr. Klein adds. “I think there is a possibility we could see the FDA begin to encroach upon pathology practice. This is something the pathology community is going to have to watch very closely, and when it’s appropriate for the optimal care of patients, respond to.”

In the meantime, pathologists will need to adjust to the era of companion diagnostics, Dr. Klein says. “At the center of personalized medicine is diagnostics, and at the center of diagnostics is the pathologist. So I think even general pathologists will be serving as consultants on the most appropriate way to test and the meaning of tests. The pathologist is going to have to understand the interaction between the specimen and perhaps in some cases the morphology and the drug responsiveness and likelihood of finding a mutation.” The pathologist will also, he says, need to be the person who understands all facets of both diagnostic and therapeutic responses, conveys that information in a coherent manner, and educates the treating physician about it.