This article has been updated from a previous version to clarify information about ongoing efforts to resolve the discordant results of the two studies.
By Ben Butkus
A recently published study out of the University of Pittsburgh Medical Center has sparked a debate about the accuracy and usefulness of Genomic Health's multigene, quantitative PCR-based Oncotype DX test for separate reporting of HER2 status in breast cancer patients.
In addition, the research results suggest that qPCR-based testing in general may be suspect when it comes to assessing individual gene expression in carcinomas for pharmacogenomic purposes.
According to one of its authors, the study — which found a high false negative rate for Oncotype DX HER2 reporting compared to the most commonly used methods of immunohistochemistry and fluorescence in situ hybridization — should serve as a warning to pathologists tempted to use Oncotype DX as a surrogate for IHC/FISH in determining HER2 status and deciding whether to prescribe targeted therapy with trastuzumab.
According to Genomic Health, however, such a scenario is unlikely since the company markets Oncotype DX's quantitative HER2 score as an independent read-out only to help clarify equivocal IHC/FISH results, and has found no evidence of clinical labs relying solely on the test to prescribe trastuzumab, marketed as Herceptin by Genentech. HER2 is one of 21 genes measured in the Oncotype DX breast cancer recurrence test.
Furthermore, the results counter those of Genomic Health's internal studies, which have demonstrated a high degree of concordance between Oncotype DX and IHC/FISH HER2 reporting for HER2 status.
In the latest study, published online in October in the Journal of Clinical Oncology, David Dabbs and colleagues at the UPMC Magee-Womens Hospital examined 843 patients with available HER2 RT-PCR results from Genomic Health from three laboratories: Magee-Womens Hospital, Cleveland Clinic, and Riverside Methodist Hospital.
They further evaluated all IHC-positive and equivocal cases using FISH. Of 843 total patient cases, 784 (93 percent) were classified as negative; 36 (4 percent) as positive; and 23 (3 percent) as equivocal using IHC/FISH.
Of the 784 negative patient cases, 779 (99 percent) were also classified as negative by Oncotype DX HER2 scoring. However, all 23 equivocal patient cases were also reported as negative by Genomic Health's test. In addition, of the 36 positive cases, just 10 (28 percent) were reported as positive; 12 (33 percent) as equivocal; and 14 (39 percent) as negative using Oncotype DX HER2 scoring.
According to the study's authors, the results constituted an "unacceptable" false-negative rate of more than 50 percent for HER2 status using Genomic Health's assay, which could "create confusion in the decision-making process for targeted treatment and potentially lead to mismanagement of patients with breast cancer if only [Genomic Health] HER2 information is used."
In addition, the authors suggested that in light of their data, "it is imperative that laboratory directors examine their HER2 results" and essential for them to conduct "a rigorous evaluation of the other components of the Oncotype DX test measured by RT-PCR … to determine whether they are similarly compromised."
Further, they warned that oncologists "be careful not to take into account [Genomic Health's] HER2 assay in selecting patients for trastuzumab therapy and continue to rely on validated IHC and/or FISH assay."
In an interview with PCR Insider, Rohit Bhargava, corresponding author on the UPMC study, said that his lab undertook the study for quality-control purposes.
"In pathology, generally whenever we send something out … we always try to have a quality assurance measure in place that [represents] what we said versus what outside people say," Bhargava said. "It's an ongoing quality-assurance measure. We cannot directly compare the Oncotype DX recurrence score because we don't provide a recurrence score. But we do provide [estrogen receptor, progesterone receptor,] and HER2 on every single primary and invasive breast cancer. And they are reported anywhere in the country. Since these are also being reported by Oncotype … we have to keep a quality control check on it."
Bhargava noted that in the majority of cases the Oncotype DX HER2 score was fine. "The problem really came when we had called something as positive using accepted methods — the combination of IHC and FISH — and many of those were coming back as negative" with Oncotype DX.
Genomic Health launched Oncotype DX in 2004 to predict the likelihood of breast cancer recurrence in women with newly diagnosed, stage I or II, node-negative, estrogen receptor-positive breast cancer treated with tamoxifen. The company has since demonstrated that the test can also be used to quantify the likelihood of chemotherapy benefit.
The results of the 21-gene, RT-PCR-based test are presented as a recurrence score of one to 100. However, soon after Genomic Health introduced the test, physicians, knowing that the test included ER, PR, and HER2 expression measurements, began requesting that the company break out quantitative results for those three genes — which Genomic Health did and still does.
"Within the first six months even, people called and said, 'There are cases in which we have discrepancies, where our own confidence in ER, PR, and HER2 status is uncertain, so could you provide on your report additional information with regard to our own assay on ER, PR, and HER2," Genomic Health CMO Steve Shak told PCR Insider. "This was a very reasonable request given that … oncologists and surgeons often look at conflicting reports."
Shak added that when Genomic Health began offering HER2 scoring, "we were also very careful to say … that HER2 testing by IHC and FISH should still remain the primary assay at the site. And in fact, that's what UPMC does, and that's what all the other hospitals do. We've said … that our own assay should not be used as the sole test to determine who should be treated with Herceptin."
Perhaps fueling physician requests for separate quantitative HER2 reporting was a study that Genomic Health conducted in collaboration with Kaiser Permanente, PhenoPath Laboratories, and the University of California, San Francisco. That study, published in JCO last year, showed a HER2 concordance between FISH and RT-PCR of 97 percent — greater than the at least 95 percent concordance stipulated for HER2 testing in guidelines from the American Society of Clinical Oncology and the College of America Pathologists, and seemingly discordant with the results of the UPMC study.
Both Bhargava and Shak offered up multiple potential reasons for this disagreement.
"First, [Genomic Health's] study might have processed the specimens somewhat differently compared to the routine cases," Bhargava said. "There may be a difference in technique, based on microdissection. The microdissection may not have been done in a consistent way [in our study], as they may have done for their study."
Second, Bhargava said that breast cancers often consist of intermixed normal breast tissue and in situ and invasive carcinomas. "But when we report HER2, we are only reporting on the invasive component," he said. "That's the only thing that's been shown to have importance, and that's what the treatment is based on." The presence of normal tissue can result in a false negative if a non-morphologic assay like qRT-pCR is used. False positive results with non-morphologic assays are less common but can occur, he explained.
A problem may arise, however, because the in situ component is more often HER2 positive than invasive carcinoma. "If you have large amounts of in situ carcinoma that's positive, with invasive being negative, then IHC and FISH [morphologic assays] … will report that as negative. But if you are making a cytosol off of that, you don't know what different components you have, and that may give a false positive result with qRT-PCR."
According to Shak, possible explanations for the discordant results might include the fact that "there might be rare instances where RNA, DNA, and protein might not give the same result. Most of the time … the studies are pretty concordant … but that would be a topic for further scientific research, if that was the case."
Another possibility, Shak said, is that the UPMC study contained some bias in terms of the types of samples examined. In fact, despite Bhargava's claims, Shak said that the UPMC study cases were not necessarily "routine" in two different ways.
"One is that only a small percentage of cases that are HER2 positive are sent into our lab for recurrence score testing," he said. "So the cases that are sent in are likely those cases where there may already be some uncertainty or some question about the HER2 signal."
Secondly, Shak noted that the UPMC authors "called up 19 labs to see whether they had a problem and wanted to participate, and 17 of the labs clearly didn't need to participate. So it really didn't look at a representative experience, but quite a selective one."
Lastly, Shak said that the increasingly controversial issue of "focal amplification" of HER2 expression due to tissue heterogeneity might have come in to play in the UPMC research. Focal amplification generally describes a situation where researchers call a tissue positive for HER2 expression with between 5 and 50 percent of cells expressing the gene, as opposed to greater than 50 percent as stated in CAP guidelines. "There may now be a growing practice, without evidence, of calling focal amplification of HER2 positive," Shak said. "That's a controversial practice that deserves further study."
However, Bhargava told PCR Insider that his team "made absolutely sure that we repeated IHC and FISH on the same tissue block that we sent to Genomic Health … to ensure that it wasn't the heterogeneity of the tissue that caused the discordance."
Although Genomic Health stands by the data of its Kaiser Permanente study, "when we hear about any discordances, we always take it very seriously, even if it's only one," Shak said. "We investigate them on a case-by-case basis. If the site had a test that was first negative, and our report comes back looking like it is HER2 positive … we tell sites [to] do another assay and then look at the HER2 status. And in many of those instances, the HER2 on the repeat by IHC or FISH confirms the RT-PCR result."
Nevertheless, in light of the discordant results from UPMC and any other random discordant results, Genomic Health is "committed to conducting additional well-controlled and well-designed studies in trastuzumab-treated patients in order to explore the value of RT-PCR in comparison to IHC and FISH," Shak said.
Genomic Health also told PCR Insider that it has requested that the UPMC team send the 14 discordant results to an independent high-quality central reference lab for blinded testing.
Shak also stressed that use of the phrase "false negative" in the UPMC paper would be deemed inappropriate by many people "because it implies that one laboratory can call itself the HER2 gold standard and be 100 percent accurate." In fact, he added, even though the ASCO and CAP guidelines recommend the use of IHC and FISH for HER2 testing, the organizations have acknowledged that up to 20 percent of HER2 measurements might be discordant between labs using those methods.
Bhargava acknowledged this frequent discrepancy, but noted that RT-PCR-based techniques might be even less trustworthy in cases where physicians are examining an individual gene or gene product, particularly in carcinomas.
"RT-PCR is no doubt an excellent and very sensitive technique, and when you are designing the probes and primers accurately, it's highly specific and quantitative, as well," he said.
"But when you're looking at expression level, especially in carcinomas, whether it's breast or prostate, or any other carcinoma, there is always an admixture of non-invasive tumor tissue" separated by only a few microns, Bhargava added.
"So if you are doing gross dissection or even microdissection, you are still not taking the pure population of the tumor," he said. "You are extracting that mRNA from everything else. No matter how good you are, in some cases it's just impossible."
When HER2 expression is part of a multigene panel, as it is with Oncotype DX, some minor variations in scoring would likely be acceptable and not change the meaning of the recurrence score. However, because of the importance of HER2 in helping to determine treatment with trastuzumab, "even a slight problem in identifying whether it's positive or negative can make a big difference in patient care," Bhargava said. "Determination is absolutely critical. You have to measure this gene accurately."
The upshot is that "just a few years ago oncologists were not using [Oncotype DX]," Bhargava added. "Our main concern is that yes, you can use this, but be careful and don't over-rely on one assay. You have to look at other things. If they are not making sense, you have to be quite skeptical about the result."
Have topics you'd like to see covered in PCR Insider? Contact the editor at bbutkus [at] genomeweb [.] com.