Gene panels for breast cancer recurrence risk have arrived. In fact, they’ve been around since the mid-2000s. And now, like guests at a wedding reception, it’s a matter of figuring out where to seat them. Clinicians have already welcomed them to the head table and in many cases are raising toasts in their honor. Some pathologists, on the other hand, have yet to be fully convinced these guests aren’t wedding crashers.
Like it or not, tests such as Oncotype DX (Genomic Health Inc.), MammaPrint (Agendia), and Mammostrat (Clarient)—to name just a few—are making their presence felt.
Clinicians favor these tests for a simple reason: the results help them decide if patients with breast cancer need chemotherapy. More broadly, the tests reflect a shift in thinking among physicians, one that emphasizes molecular profiling of tumors. They’ve arrived on the scene when physicians are also starting to question the value of lymph node status to help determine treatment.
George W. Sledge, MD, finds these changes remarkable. Not all that long ago, he might have pink-slipped a test that would help parse treatment decisions. When the NIH held its consensus development conference on adjuvant therapy with breast cancer in 2000, he recalls, the agreement was, basically, that everyone with a tumor greater than one centimeter ought to be treated with chemotherapy. “There’s no question that resulted in us hugely overtreating patients,” he says. “So I think a test that reduces the quantity of human suffering by half in that group is a useful test,” says Dr. Sledge, professor of medicine and pathology, Indiana University, Indianapolis, and immediate past president of the American Society of Clinical Oncology.
In clinical practice, these tests are functioning like traffic managers. “We now see fewer patients getting chemotherapy who would have gotten it before,” says Thomas Julian, MD, professor of surgery, Drexel University College of Medicine, Philadelphia, and director of breast surgical oncology for the West Penn Allegheny Health Care System, Pittsburgh. “We’re also seeing a few who are getting chemo who might not have gotten it before. So it’s changed in both directions,” says Dr. Julian, who is also senior surgical director for medical affairs for the National Surgical Adjuvant Breast and Bowel Project.
These are still early days for the tests. While their makers are already looking to possible applications for other types of tumors (Genomic Health began delivering a 12-gene Oncotype DX test for stage II colon cancer in January 2010), practicing physicians continue to sort through what the tests can tell them about patients with breast cancer, particularly those with estrogen receptor-positive, node-negative tumors.
Oncotype DX is a real-time RT-PCR assay measuring RNA expression in 16 cancer-related genes and five reference genes, using paraffin-embedded tissue. Results are given as a recurrence score between zero and 100, which are translated as low risk (a score of 18 or lower), medium risk (19 to 30), or high risk (31 or above). The MammaPrint microarray assay measures expression of 70 genes in fresh tissue; it categorizes patients as either high risk, with a so-called poor signature, or low risk (a so-called good signature) for recurrence. There is no intermediate category. Mammostrat is an immunohistochemistry test measuring five markers: p53, HTF9C, CEACAM5, NDRG1, and SLC7A5. The results are combined into a quantitative risk index: low, moderate, and high. For now, only MammaPrint has FDA clearance.
Though Dr. Sledge and his colleagues at IU use the Oncotype DX, he holds a pantheistic view of the tests. “My sense, my bias, is that MammaPrint and Oncotype and indeed Chuck Perou’s [Charles M. Perou, PhD] intrinsic subtypes are just different ways of measuring the same underlying biology. So I suspect they all give us essentially the same answer. There’s a population of patients, whether we call them luminal A, or Oncotype DX low recurrence score, or low risk MammaPrint, or whatever, that’s probably characterized by estrogen receptor positivity, HER2 negativity, and low proliferative index.”
Kenneth J. Bloom, MD, chief medical officer of Clarient, sees the common ground among the three tests as well. “It’s just three different ways of approaching the exact same problem. Which tells us it’s not as if only one approach will get you the answer.”
Terry Mamounas, MD, professor of surgery, Northeastern Ohio Universities College of Medicine, and medical director, Aultman Cancer Center, Canton, Ohio, uses the Oncotype DX routinely for node-negative, estrogen receptor-positive patients to help decide whether chemotherapy should be added to hormonal therapy. Low-risk patients are unlikely to benefit from this addition, but high-risk patients benefit a great deal, though test results don’t make such decisions automatic. “Obviously for somebody who is medically not fit for chemotherapy, we don’t use the test,” Dr. Mamounas says. “Because we’re not going to give chemotherapy even if the test dictates that we should.”
Likewise, the test is not useful in patients whose tumors are HER2 positive. The test nearly always will show such patients to be at high risk; moreover, the paradigm for treating such patients is with chemotherapy and trastuzumab (Herceptin). “We don’t use the test to tell us otherwise,” Dr. Mamounas says.
Another consideration is tumor size. The test is most useful for tumors of around five millimeters or greater in size, he says. “For patients with very, very small tumors—one, two, three millimeters—even bad biology may be offset by the very early detection of a lesion. So there’s no need for the test.”
Dr. Sledge follows a similar pattern of use: for patients who are lymph node negative, ER positive, and HER2 negative, with “moderate-size tumors—say, tumors that are over a centimeter but less than four or five centimeters,” he says. “I don’t know that we have a hard and fast rule, but that’s kind of the general area we’ve used it in from the beginning.” (He’s used it since about 2005.)
Dr. Julian says the tests are now common parlance among his clinical colleagues. “When you go to meetings, and the medical oncologists are talking about treatments, this is something that’s always tossed up: Well, what was the Oncotype score, the recurrence score?” Obviously, it’s on clinicians’ minds, he says.
It’s on pathologists’ minds as well.
Elizabeth Hammond, MD, agrees these tests are useful, although she suspects they may best prove their mettle in second- or third-generation assays. It’s simple biology: phenotypic expression of a genetic alteration of ER or HER2 status is the result of multiple cell-signaling pathway changes. “Looking at multiple expressions of that problem with a gene panel, either by RT-PCR or some other method, will in the long run give us better information,” she says.
“But that assumes we’ve done the work to understand what the information is telling us,” continues Dr. Hammond, pathologist, Intermountain Healthcare, and professor of pathology, University of Utah School of Medicine. She suspects such information will prove to be more accurate, but will require the same rigorous analyses that are applied to other tests. “From an analytic perspective they’re wonderful tests,” she says. “What we have to do is prove they have high clinical relevance.”
Dr. Bloom wavers between enthusiasm, skepticism, and irritation—sometimes in the same sentence—when he considers these tests. For pathologists, these assays may be somewhat of an acquired taste, like organ music.
Clarient, as noted, has a test of its own, and Dr. Bloom clearly feels the test is useful. Likewise, he acknowledges that competitors such as MammaPrint and Oncotype DX are good tests.
He’s also sympathetic to the plight of oncologists. “They’ve got tough decisions,” he says. “They have to make those decisions under uncertainty, and a lot of times without mature data. They see things they get very interested in and want to jump on it.”
That’s when the role of skeptic takes over, as it should for any pathologist, he says. “I think part of our job is a little bit of being a naysayer,” Dr. Bloom says, “and saying to clinicians, ‘I know you’re excited about this stuff—but let’s not get ahead of ourselves.’ ”
He’d also like pathologists to show their clinical colleagues the value of the pathology data already available. Not every pathologist passes muster in this regard, he says. “Frankly, I think there are more than a few pathologists who go to tumor board and reduce everything to the same three sentences no matter what,” he says: tumor size, ER/PR/HER2 status, and lymph node involvement. “That’s it. No further discussion.” That closes the door to more nuanced conversations. “In my experience”—he’s the former director of laboratory operations at Rush-Presbyterian-St. Luke’s Medical Center, Chicago—“most oncologists are happy to engage and bounce decisions off you. That’s part of your job as pathologist.”
But since Oncotype DX and MammaPrint are send-out tests, with turnaround times of 10 to 14 days (results from the IHC-based Mammostrat, also a send-out test, are available to local pathologists within 48 to 72 hours, according to the company), he’s worried that pathologists will be excised from the conversation when the results come back to the oncologists.
Dr. Julian concedes the point. “It’s interesting—we don’t really get a whole lot of discussion about this with our pathologists.” But, he adds, the tests may enhance the conversations between clinicians and patients. (Some pathologists, of course, might feel they need to be part of patient conversations as well.)
Dr. Julian, who uses Oncotype DX, says the discussion begins with a patient’s postoperative visit, when she finds out she has invasive, ER-positive, node-negative breast cancer. The question naturally arises, “What next?” The patient is then told that the Oncotype DX test will break her risk factors into one of three categories—low, intermediate, or high—to help determine the need for chemotherapy. “So it sets up the information ahead of time, before they go to the medical oncologist,” who’ll discuss the results in more detail with the patient. “Patients actually do seem to understand that,” he says, noting that a recent in-house study showed that about half the patients said they felt these discussions helped them comprehend the reasons behind their treatment recommendations.
But that doesn’t address Dr. Bloom’s concern—that results won’t be shared at tumor boards, which leaves pathologists out of the interpretive decision. These meetings should also be a place for pathologists to explain the subtleties of multi-index PCR-based tests, he says. “My general impression is that people hear ‘molecular’ and think it’s better. Frequently that is not the case. But if you don’t understand it you don’t know what questions to ask,” says Dr. Bloom. “And if it was your pathologist who was talking about it at tumor board repetitively, you would get those nuances over time, because they would be being described to you with case example after case example. But I worry about when you get a result by yourself in your office without that collaborative spirit, if you ever truly understand the nuances around the testing.”
In teaching the CAP’s breast cancer predictive markers course throughout the country, Dr. Hammond encounters another concern: pathologists say they’re frustrated by oncologist colleagues who are using a gene panel test when they don’t believe local pathology results for ER, PR, or HER2.
But she reserves her biggest fears for inappropriate use of the test. “I have great concerns about Oncotype DX being done in women who are estrogen receptor negative or women who are lymph node positive,” she says, since the test has not been validated in these other groups.
Clinicians share those concerns, though that hasn’t kept them from eyeing other uses. Some are tentatively exploring its use in node-positive women, based on a retrospective analysis of archival tissues from a randomized trial (Albain KS, et al. The Lancet Oncology. 2010;11:55–65). This analysis, looking at Oncotype DX, was “very eye-opening,” says Dr. Julian, showing a segment of ER-positive, node-positive postmenopausal women who did not receive benefit from chemotherapy beyond what antihormonal therapy could deliver.
Dr. Julian struggles with the possible implications. “That’s a gut-wrenching situation.” Telling a relatively young, node-positive woman that she doesn’t need chemotherapy is an historic change, he says. “We’ve always said, ‘Gotta have chemo, gotta have chemo, gotta have chemo.’ Then… boom! Now we are saying, ‘You know what? You may not need chemo.’ ”
Dr. Mamounas and his colleagues at Aultman have begun using Oncotype DX in node-positive patients. “But very carefully,” he says.
The data for node-positive patients obviously are less robust than data for node-negative patients (though the company reports that Medicare reimburses for the Oncotype DX for patients with one to three nodes positive for metastases or micrometastases). The aforementioned trial, SWOG 8814, randomized patients between chemotherapy and hormonal therapy versus hormonal therapy alone. The overall trial showed a small benefit, Dr. Mamounas says, but when researchers looked at it according to the recurrence score, they found no appreciable benefit for patients with low recurrence scores—only for high recurrence scores. Potential benefit for patients with intermediate recurrence scores is uncertain.
These results in and of themselves were not enough to persuade Dr. Mamounas to change his approach, but taken in conjunction with published data from the NSABP B-20 trial—showing that tumors with low Oncotype scores are sensitive to endocrine therapy but not chemotherapy—he felt it merited a look. He rarely uses the Oncotype DX in node-positive premenopausal women, however, because there are no data for this group of patients, he says.
Future studies looking at the use of Oncotype DX in node-positive patients, including a recently opened SWOG study, S1007, called the RxPonder trial—a prospective, randomized trial—could be galvanizing. In this study, researchers will look at whether patients with a recurrence score of 25 or less benefit from chemotherapy. They’ll also look at whether it’s possible to refine the cutoff point to identify where chemotherapy benefit might begin, says Steven Shak, MD, chief medical officer of Genomic Health.
“It’s going to be a very, very provocative and interesting study to get through,” says Dr. Julian. “Historically, everybody’s had it drummed into them that node-positive disease requires chemotherapy. That’s how all the textbooks are, that’s how the literature is, that’s even how patient-related information is set up: Nodes are positive, you need chemo—no ifs, ands, or buts. Now it’s like: hmmmm.”
It makes sense that gene panels might be useful in node-positive patients, says Soonmyung Paik, MD. “Response to chemotherapy has nothing to do with whether tumor cells are in the lymph node,” says Dr. Paik, director of the Division of Pathology, NSABP Foundation, Pittsburgh, and a director of the Samsung Cancer Research Institute, Seoul, South Korea. “It depends on the biology of the tumor.” Though he helped develop the Oncotype test, he says he has no financial ties to the company.
Dr. Shak reports that the company is also looking at possible uses of the Oncotype DX in ductal in situ carcinoma of the breast. Results from a study that was done when the assay was being developed “suggested that there might be a different, DCIS algorithm that might reveal the recurrence rates regardless of whether tamoxifen was given,” Dr. Shak says. (The company announced on May 24 that it planned to make the Oncotype DX DCIS score available by the end of the year, based on positive preliminary results from the ECOG 5194 validation study. It plans to submit data from the study for presentation at the 2011 San Antonio Breast Cancer Symposium, which will be held in December.)
Clinicians are also struggling with how to treat patients classified as being intermediate risk by Oncotype DX. The TAILORx trial, now underway, might provide some answers. Accrual was reached last fall, with more than 11,000 patients screened, reports Dr. Shak. More than 7,000 patients with recurrence scores in the mid-range group are randomized to treatment with hormonal therapy with or without chemotherapy. For the trial, the scores are somewhat altered: The low-risk category tops out at 11 (dropped from 18), while the lower boundary of the high-risk score is 25 (lowered from 31).
Dr. Sledge is hopeful but cautious. “If it gives me a definitive result, I’ll be delighted with applying that result to my patients,” he says. “I think the greater likelihood is that we’re going to find there’s a spectrum of results, that maybe at the low end you’ll get less benefit from chemotherapy, and at the high end you’ll get more benefit from chemotherapy. And I think that’s going to be the area that will be troublesome statistically.”
His comments touch on one other area of concern for both pathologists and oncologists—the notion of what, exactly, low, intermediate, and high risk means. Dr. Paik recalls that when the Oncotype DX was being developed, some collaborators saw little need for providing a numerical score, insisting the low-/intermediate-/high-risk labels would suffice. While that’s easier for patients and physicians to understand, it can be misleading, says Dr. Paik. A low-risk score of 17 might have far different implications than a low-risk score of 5. “I actually don’t like that the report is given as low, intermediate, and high risk,” he says. “I want to use only a score. But I don’t think oncologists or patients are trained to look at the data in a continuous manner.”
It’s not surprising that Oncotype DX provides a user-friendly, quantifiable score, since it had its genesis in an era when laboratories were still struggling to provide consistent IHC and FISH results, notes Dr. Sledge. Indeed, says Dr. Paik, when work on the Oncotype DX began in 2000, a motivation was the discovery, from the Herceptin trials, that HER2 testing in the community was unreliable. “We wanted a prognostic test that was highly reproducible,” he says.
The Oncotype DX now reports single-gene scores for ER, PR, and HER2 alongside its recurrence score. It began doing so in 2008, says Dr. Shak, at the request of clinicians who wanted more quantitative information and were concerned about uncertainties in ER, PR, and HER2 testing. But because it’s an RT-PCR test, some question their usefulness. Dr. Hammond doesn’t object to the reporting of ER/PR/HER2 results by RT-PCR per se, and even sees some benefit, since it gives physicians another means of evaluating test results. But she sounds the typical pathologist caution about the need to validate those results. No one, everyone agrees, should act on RT-PCR results alone.
Clinicians who want RT-PCR HER2 results are “doing it because they believe in it themselves,” Dr. Hammond says. But, she adds, no organization has compiled reported evidence to provide a useful guideline for using RT-PCR for HER2 testing. The CAP’s HER2 panel will reconvene at some point to consider this very question, she adds.
David Dabbs, MD, chief of pathology, Magee-Women’s Hospital, Pittsburgh, worries about discrepant ER/PR/HER2 results between IHC/FISH and RT-PCR and says it’s a problem at his institution and two others.
In an abstract (No. 603) from June’s ASCO meeting, Dr. Dabbs reported on a quality assurance study looking at HER2 testing at his hospital as well as at Cleveland Clinic and Riverside Methodist Hospital, Columbus, Ohio, versus HER2 results from Oncotype DX. Reports the abstract: “Of the total 507 MWH cases, 236 CC cases, and 100 RMH cases; 99% (464/468) of MWH, 99.5% (221/222) of CC, and 100% (94/94) of RMH negative cases were also negative by GHI RT-PCR assay. However, all MWH (n=13), CC (n=8) and RMH (n=2) equivocal cases were reported as negative by GHI. Of the 26 MWH positive cases, 8 were positive and 18 were either negative or equivocal by GHI (concordance of 31%). Concordance for positive cases was 17% (1 of 6) at CC and 25% (1 of 4) at RMH.” (Dr. Dabbs’ manuscript based on the study was accepted at CAP TODAY press time for publication in the Journal of Clinical Oncology.)
Dr. Dabbs finds these results alarming. In an interview with CAP TODAY prior to the ASCO meeting, he spoke generally about his study, calling the findings “serious discrepancies. It’s a patient safety issue,” says Dr. Dabbs, who has been in contact with the FDA about the results and subsequently has begun submitting every new discrepant result to the agency via MedWatch.
“Any pathologist who does either immunohistochemistry or FISH would have no problem concurring with our results,” he contends. “The results that are positive are flagrantly positive.” The FISH HER2s are performed by a molecular pathologist in his lab who personally reviews every case, he adds. “There could be a number of reasons for the poor percent positive agreement. The most obvious seems to be suboptimal microdissection based on their findings,” Dr. Dabbs says.
In addition, Dr. Dabbs says that if the HER2 component is unreliable, “Then there should be substantial concern about the other gene sets of the multivariate test.” Says Dr. Dabbs: “The Oncotype recurrence score is predicated on RT-PCR results for 16 genes. If an individual gene assay (e.g. HER2) is not robust, then the recurrence score becomes questionable. And there is no means by which to measure this because of the proprietary nature of the test.”
Genomic Health has done its own studies showing high concordance with central laboratories—greater than 95 percent, says Dr. Shak—between FISH and RT-PCR results, but Dr. Dabbs sees value in providing an additional, independent look. He says he canvassed 19 other laboratories when he discovered the discrepancies at his lab. Cleveland Clinic and Riverside Methodist both had data and agreed to participate in a study.
Dr. Dabbs echoes Dr. Hammond with his worries. “Some clinicians are using the Oncotype test off-road. It’s inappropriate. There’s no data to support what they’re doing.” He says in only one case has a clinician contacted his lab about a discrepant result, asking for guidance. “But that’s only one doctor who called us. I have no idea what the others are doing. It’s disturbing.”
Not everyone shares Dr. Dabbs’ alarm. In followup e-mail interviews with CAP TODAY, several sources commented on the abstract prior to the ASCO meeting.
Dr. Paik reiterated that Oncotype is not recommended for HER2-positive patients. “And since Oncotype DX is indicated for only ER-positive node-negative tumors,” he writes, “the incidence of HER2 positivity among tested cases would be small and enriched for cases with borderline or low amplified HER2 positivity or enriched for some unusual cases such as lobular or tubular cancer that were somehow HER2 positive. It is not unusual to find low HER2 mRNA levels in such cases,” he says, based on his experience measuring HER2 mRNA using three different methods (microarray, QuantiGene Plex, and NanoString) in 1,600 cases of HER2-positive breast cancer from NSABP trials.
He adds that the Oncotype DX has a built-in control for HER2 mRNA measurement—the level of GRB7 mRNA, which behaves almost like HER2 mRNA. In his view, reporting HER2 results via the Oncotype DX helps when the clinical assay is negative or equivocal but mRNA is high. That should prompt clinicians to request re-testing of HER2 using a clinical assay, he says, “since it is most likely a false-negative result.”
Dr. Mamounas says it is hard to judge the significance of the results of Dr. Dabbs’ abstract, since, he writes, “other studies have shown high concordance between HER2 by RT-PCR and HER2 by IHC or FISH. Ultimately, the best predictor of HER2/neu positivity can only be determined by its prediction of response to trastuzumab.”
Dr. Sledge, who was not asked to respond to Dr. Dabbs’ abstract, says discrepant results are uncommon in his experience. “We pretty much always send for an Oncotype DX after we know that a patient is estrogen receptor positive. So presumably the only discrepant result you might find is the patient is estrogen receptor negative by Oncotype. I can’t remember the last time that happened. So it’s certainly not been an issue in my practice.”
It’s not outside the realm of possibility, however; if he did encounter discrepant results, he says, he’d repeat the test by IHC. If he had a FISH-negative HER2 case that was subsequently called positive by RT-PCR, “That would be disturbing,” he says. “I’m not even sure how I would face it, because it hasn’t happened in my practice.”
In the best-case scenario, Dr. Bloom says, a discrepant result should launch a discussion between the clinician and pathologist and a followup test. “I don’t think you just say, ‘They did it by PCR—they’re wrong.’ The result is legitimate, but you have to explain where the legitimate result came from—and you have to demonstrate positive cells in the patient.”
Dr. Mamounas says he only sporadically sees discordance with RT-PCR results. Such discrepancies may result from how the specimens are obtained—for example, the in-house ER/PR/HER2 testing might be done on core biopsy samples, while the Oncotype is done on the surgical specimen. Most discrepancies are easily resolved, he says, after followup conversation with the local pathologists.
He himself has not found the Oncotype ER/PR/HER2 RT-PCR results useful. “I think the whole score, and the way it’s weighted, is more important,” he says. “Otherwise, you’re picking apart a test that’s developed and validated as a single score.”
Nonetheless, he says the RT-PCR information might at times help him make a decision. He gives as an example an 80-year-old with an equivocal HER2 by traditional methods; if the RT-PCR result is negative, it might help him decide against using Herceptin and chemotherapy. “It can almost act like an occasional tie-breaker,” he says.
It’s hard to say exactly where the learning curve might trail off into outright inappropriate use. Dr. Mamounas says he’s seen his clinical colleagues develop a more nuanced approach to using the Oncotype test as they become more familiar with its parameters. Initially, he says, clinicians were using the test in ER-positive patients with small tumors and for patients who were typically not candidates for chemotherapy based on their clinical presentation. “Although it does not make a lot of sense to send the test for a patient who refuses chemotherapy upfront,” he says, he’s seen some patients change their minds after learning their Oncotype score places them in a high-risk category. And Dr. Dabbs notes that oncologists have also learned to refrain from ordering the test for patients with tubular cancers of the breast, since these tumors respond well to antiestrogen therapy alone and would not be candidates for chemotherapy.
As these tests send patients down different treatment roads, they’re also highlighting the different mindsets pathologists and clinicians can bring to a problem.
Surgeons and oncologists hear “if, if, if”—as in the case of the prospective SWOG trial—and get excited. “If SWOG comes out and shows that the Oncotype DX is incredibly predictive of what happens to the [node-positive] patient, afterwards, it really will start reducing even more the necessity then to go after positive lymph nodes,” says Dr. Julian.
Pathologists, on the other hand, hear “if, if, if,” and cringe. Together, they’re like contrapuntal voices in a choral mass, with one group singing “Rejoice!” and the other warning, if not “Repent!,” then at least “Be careful!”
As the tests continue to evolve and possibly be used in new settings, the chorus will likely keep singing—and even, as recent events are showing, perhaps grow slightly louder.
Karen Titus is CAP TODAY contributing editor and co-managing editor.