College of American Pathologists
Printable Version

  IHC, FISH still sharing
  HER2 spotlight


August 2005
Cover Story

Like the perennial debate over which team is better—the New York Yankees or Boston Red Sox, the debate about FISH versus IHC for HER2 measurement appears to be everlasting. Of course, in the case of HER2 detection, the stakes are much higher.

As the anti-HER2 antibody trastuzumab becomes used in more clinical situations, clinicians increasingly want to identify women with breast cancer who are more likely to benefit from this expensive and often effective therapy. Like baseball fans who cite voluminous statistics, those who advocate measuring HER2 protein overexpression by immunohistochemistry, or IHC, and those who favor HER2 gene amplification by fluorescence in situ hybridization, or FISH, cite numerous arguments and extensive data to support the advantages of their preferred method. A prototypical exchange might go like this:

FISH advocate: FISH directly visualizes the underlying genetic abnormality.

IHC advocate: Whats important for carcinogenesis is HER2 protein overexpression. It appears that in occasional cases overexpression can occur without gene amplification.

FISH advocate: IHC gives false- positive results for some 3+ samples and has a gray zone of 2+ results.

IHC advocate: There is no compelling evidence that IHC 3+/FISH-negative cases have poorer prognosis. Nor is there compelling evidence that IHC 2+/FISH-positive cases are more likely to benefit from trastuzumab. And the few IHC 2+ cases can be reflexed to FISH.

FISH advocate: Interpreting IHC is too subjective. Surveys show that IHC results vary when done by inexperienced laboratories.

IHC advocate: Interobserver variability with IHC among experienced laboratories is very good. And FISH results, too, can vary when done by inexperienced laboratories.

FISH advocate: More laboratories are using FISH, both as the initial test and for confirmation of IHC results. FISH is especially helpful for needle core biopsies, where IHC has artifacts.

IHC advocate: IHC is faster, less labor-intensive, and less expensive. That’s why most laboratories use IHC as their initial screening test.

The situation can indeed be confusing. What’s most important, says Kenneth Bloom, MD, medical director at Clarient Inc. in Irvine, Calif., and clinical professor of pathology at the University of Southern California, is that IHC and FISH "are both great tests." Dr. Bloom does favor IHC. "I don’t want to sound negative about FISH," he says. "I use it every day and it’s a great test. But a test that is done under 100x power in a dark room with no landmarks on the slide to me is not the epitome of a screening test." In his experience, a large majority of laboratories use IHC as the initial screen and then reflex to FISH in some cases, while about 30 percent of laboratories use FISH as their first test.

To D. Craig Allred, MD, professor of pathology in the Breast Center at Baylor College of Medicine, Houston, it makes more sense financially to do IHC for all primary screening. IHC done accurately provides a definitive answer on 80 percent or more of cases in a rapid, cost-effective manner. Several expert guidelines recommend referring IHC 2+ cases to FISH, and reporting the one-third or so of samples that are amplified as positive. "That’s probably a good algorithm," Dr. Allred says. "Doing FISH up front would be no more accurate and would cost a lot more and take a lot more time."

FDA-approved kits have made IHC easier, Dr. Allred notes. "If you follow the manufacturer’s directions, you should get accurate results most of the time."

Patrick Roche, PhD, senior director for reagent development at Ventana Medical Systems Inc., favors protein testing as the first-line assay, based on his past experience at the Mayo Clinic. "Intermediate protein expression [IHC 2+] requires evaluation of gene status," he says. Thomas Grogan, MD, founder and chief medical officer at Ventana and professor of pathology at the University of Arizona Medical Center, agrees with Dr. Allred: "Follow the recommended controls religiously," he advises. He believes that instrumentation and automation can improve test standardization. "Instrumentation allows people to get the same results from laboratory to laboratory."

Ann Thor, MD, chair of the Department of Pathology at the University of Oklahoma Health Science Center, also sees most laboratories doing IHC for initial screening, with FISH for IHC 2+ cases. "There may be a few institutions doing FISH for all cases," she says, "but given the cost and reimbursement issues, probably not a lot. And there is no evidence that you should use FISH on all cases."

A recent shift has been more reporting of nonadjusted data for the Abbott/Vysis FISH kit. "Some institutions now give the centromeric score, the HER2 gene copy number, and the adjusted score [ratio]," Dr. Thor says. "There has been some concern that adjusted score may miss some women who may benefit from trastuzumab."

In the research setting, a developing trend in cooperative group clinical trials is to measure "molecular partners" of the HER2 receptor. "Many trials have moved beyond trastuzumab and HER2 testing to look at combining traditional agents with newer reagents that target HER2 and a binding partner," Dr. Thor says.

One pathologist who is using much more FISH is Atul Bhan, MD, director of the immunopathology unit at Massachusetts General Hospital and professor of pathology at Harvard Medical School. "We do FISH on all core biopsies," says Dr. Bhan. For excisional biopsies, use of FISH is based on whether the specimen is IHC 2+. "We now have a tendency to do FISH on all cases except those that are 3+," Dr. Bhan says. "As I see it, oncologists would be happy if FISH was done at some time on all cases."

Raymond R. Tubbs, DO, chair of the Department of Clinical Pathology at the Cleveland Clinic Foundation, thinks that the dominance of IHC for initial screening "is likely to change in the near future as bright-field assays become more mainstream and experience is gained with these assays." Dr. Tubbs and colleagues have been developing such an assay, which has the potential to show gene copy number and protein expression on the same slide. "Most pathologists do not want to take on the additional instrumentation and training of personnel required for FISH," he observes. "Having an assay that fits in with how we use a microscope and with our workload makes a lot of sense." In his view, bright-field measurement of HER2 genes will eventually become "the method of choice."

To compare the performance of IHC and FISH, Dr. Bloom has done an extensive analysis of the available data. He focuses on the FDA’s review when Genentech submitted an amended claim to have patients qualified for trastuzumab therapy by FISH. (IHC was the method used in clinical trials and the method included in the original trastuzumab approval, though both FDA-approved IHC assays differ from the predicate IHC assay.) The FDA’s review found that an IHC 3+ score and FISH positivity were virtually identical in relative risk reduction with chemotherapy plus trastuzumab, both for time to progression and for overall survival. "So independently either FISH or IHC helped you predict who would best respond to trastuzumab therapy," Dr. Bloom says.

The FDA also looked at two discrepant categories, IHC 2+/FISH-positive and IHC 3+/FISH-negative. FISH positivity did not confer any clinical advantage for IHC 2+ cases, Dr. Bloom says. "That would argue that reflexing IHC 2+ cases to FISH wouldn’t get you much of a benefit," he says. On the other hand, IHC 3+ cases did benefit from addition of trastuzumab in time to progression (though not in overall survival) even when they were FISH-negative. Dr. Bloom cautions that each category contained small numbers of patients, 32 and 43 respectively, out of 422 total patients, so the data were not strong enough for statistically firm statements. But they clearly don’t support the notion that FISH is more accurate.

So where did the idea come from that FISH is superior? Investigators comparing the two methods often combine data from IHC 2+ and 3+, Dr. Bloom finds. "Since IHC 2+ appears to have no significant benefit, that artificially makes FISH look better." One trial of single-agent, first-line trastuzumab therapy for metastatic breast cancer analyzed outcomes for IHC 2+ and 3+ separately. Response rates were 35 percent for IHC 3+ and 0 for IHC 2+, indicating that the IHC score clearly discriminated responders. For FISH, response rates were 34 percent for FISH-positive specimens and seven percent for those that were FISH-negative (Vogel CL, et al. J Clin Oncol. 2002;20:719-726).

A study by the Cancer and Leukemia Group B demonstrated the equivalence among IHC and FISH, as well as polymerase chain reaction, in determining HER2 status for clinical purposes. Among patients with node-positive breast cancer, all three methods demonstrated a similar prognostic relationship for disease-free survival and overall survival. All methods were also equivalent in predicting therapeutic response to high-dose doxorubicin adjuvant therapy in patients with HER2-positive tumors; none of the methods—or combinations of methods—was statistically superior (Dressler LG, et al. J Clin Oncol. 2005;23:4287-4297).

Based on his Mayo Clinic experience, Dr. Roche says that high-volume laboratories experienced at doing both assays can achieve greater than 92 percent correlation between IHC 3+ and FISH amplification. Some people believe that a few percent of IHC 3+/FISH-negative cases are true positives, overexpressing HER2 protein by some mechanism other than gene amplification.

Further information about the reliability of IHC comes from the National Surgical Adjuvant Breast Project B-31 trial, for which preliminary results were announced this year at the meeting of the American Society of Clinical Oncology. In this trial, trastuzumab is being evaluated in the adjuvant setting for women with primary breast cancer and no evidence of metastatic disease. Women who are HER2-positive are randomized to get standard chemotherapy plus or minus trastuzumab. Patients initially became eligible for enrollment after HER2 testing in their local hospitals. After the first 104 cases, Dr. Bloom says, the central laboratory retested the blocks. "What they found alarmed them big time," he says. About 20 percent of cases called IHC 3+ by laboratories other than reference laboratories turned out to be negative, with the miscall rate rising to 50 percent for laboratories using a method other than Herceptest. "Their initial response was that IHC is worthless," Dr. Bloom says.

But the NSABP set up a program to certify laboratories that do more than 100 cases per month who could show that their results correlated to FISH. Many laboratories met this criterion and their results were accepted as qualifying patients for enrollment. Review of the next 240 cases found that the error rate for IHC 3+ from these certified laboratories was two percent. "Basically this showed that once you know how to do the IHC assay, you can do it very reliably and reproducibly," Dr. Bloom says. This finding also shows the value of a certification mechanism.

When FISH-positive specimens from hospital laboratories were repeated in the central laboratory, the discrepancy rate was 4.5 percent. "So the argument that FISH is easier to do or more robust did not hold in a real-world setting, although the concordance was still excellent," Dr. Bloom says.

Dr. Thor draws a similar conclusion about FISH. "Some data from large cooperative groups such as NSABP suggest that small laboratories that don’t perform FISH a lot may not perform as well as laboratories that have been certified," she says, which suggests the value of certification for FISH, too. "Experience and setting up proper controls are equally important [to do FISH accurately] as with IHC.

"The most important factor in doing either test reproducibly is to follow the guidelines with the kits as closely as possible," Dr. Thor continues. She especially advises against shifting to more-sensitive secondary and tertiary reagents, which can lead to calling normal levels of expression as overexpression. "CAP guidelines are very useful in that regard," she says. The CAP is now doing HER2 testing Surveys. "When we get those data, they will be important information for pathologists. They will show us how we’re doing."

At the University of Oklahoma Health Science Center, only two pathologists read out IHC HER2 samples. "I do think it is important that whoever reads HER2 needs to be very familiar with the criteria of various scores and understand nuances of reporting," Dr.Thor says. "Reading HER2 is more demanding than other IHC analytes, such as cytokeratin." It is equally important to be familiar with scores and reporting when doing HER2 by FISH. For both assays, positivity should be scored only on the invasive part of the tumor, not on the in situ component. Two-thirds of insituductal carcinomas are positive for HER2, whereas only one-third of invasive tissues are.

Dr. Bloom predicts that FISH results will become more problematic as more small laboratories bring up the assay. "Smaller laboratories are getting pressured to set up FISH on their own," he says. Community oncologists perceive FISH as being easy, he finds, "like connecting the dots." In actuality, it is like any other staining test, except that DNA is more stable than protein. He sees autofluorescence of native proteins as a confounder. "Depending on your training level and experience, you can easily overcall that autofluorescence as a positive gene signal."

Dr. Allred acknowledges the "groundswell for FISH."

"Certainly there is some push from oncologists to do FISH because of the misconception that it is more accurate. If you get a chance to educate them, you can often change their minds," he says. Dr. Bhan has experienced "a drive for FISH from clinicians." To standardize results, he and a colleague, Elena Brachtel, MD, do all the HER2 IHC at Massachusetts General Hospital. By protocol, they do FISH on excisional biopsies only for IHC 2+ staining. "But it does get changed by demands of clinicians," he says. "Our protocol is getting a bit modified because of reports of a small number of cases that don’t show overexpression by IHC but that may show gene amplification. So the trend at our institution now is to do FISH on all cases at some time during treatment."

Dr. Thor says some people simply do not believe the literature that suggests that IHC is sensitive and specific. "I have done personal comparisons of hundreds of cases and I have never found an IHC zero or 1+ sample that had HER2 gene amplification." Use of FISH can be influenced by an oncologist who is eager to use trastuzumab. "But," she repeats, "I don’t know any data that says using FISH is justified for IHC 0 or 1+ cases."

A relative advantage of FISH is that it is not subject to edge, crush, or retraction artifacts which can be prominent on core needle biopsies, while IHC is subject to these artifacts. In edge artifact, the most common confounder, tissue close to the edges of the specimen stains stronger than in the center because antibody pools at the sides. "Some people see 3+ along the edge and 1+ in the middle and misinterpret the specimen as 3+ or 2+," Dr. Bloom says. "So they overcall HER2 results." The trend over the last couple of years to do more testing on core biopsies has exacerbated this problem. "We need to be very aware and seasoned in HER2 IHC interpretation," Dr. Bloom cautions. He suggests reflexing IHC 3+ scores on core biopsies to FISH in small-volume laboratories or repeating the IHC on a lumpectomy specimen if it becomes available.

An issue specific to FISH is how to report results. Ideally, any FISH assay would have a probe for chromosome 17 centromere and a probe for the HER2 gene, according to Dr. Tubbs. That would enable both gene counting and calculation of a ratio. "One option to having a CEP 17 probe is to establish cutoffs that account for polysomic states," he says. That’s what he is doing for the bright-field assay that is in development. "A lot of work that we have done with Dr. Tubbs suggests that you don’t really need a CEP 17 probe," says Ventana’s Dr. Roche. "Cases with low-level amplification most often result from duplication of chromosome 17, not from HER2 gene amplification. And they are usually associated with low-level protein expression." Dr. Bhan at MGH says he doesn’t know of a single outcome study that settles this question. Thus, he includes raw figures in his report. Above the line is the ratio result, with values greater than two being called amplified, while below the line are the two other figures—number of HER2 genes visualized and number of copies of chromosome 17. Clinicians can make their own interpretation if they wish.

Dr. Bloom believes that one way to improve IHC scoring on HER2 is to use image analysis. (He discloses that he is working for a company that has a vested interest in this technology.) "What we really try to do by IHC is visually relate the intensity of the HER2 stain to the number of HER2 receptors on the surface of the cell," he explains. Unfortunately, while the human eye is good at reading patterns of staining, it is not so good at detecting subtle changes in staining intensity. A comparison showed that pathologists using image analysis achieve greater interobserver agreement and better concordance with HER2 gene status (Bloom K, Harrington D. Am J Clin Pathol. 2004;121:620-630).

Dr. Grogan advocates automated performance of IHC or FISH as a way to increase accuracy. "The advantages of automation are standardization and transportability of results," he says. "The score should look the same at every site. Then you can focus in on interpretation." Ventana has an automated PMA-approved IHC test for HER2. Interlaboratory concordance was very high with the automated IHC assay, according to Dr. Grogan.

As a potential solution to the either/or debate over IHC and FISH, Dr. Tubbs predicts that the bright-field assay that he and several colleagues are developing, which uses enzyme metallography to visualize HER2 genes, "will very soon make it possible on one slide to use H&E to detect margins and other tissue features, to do all the things done with IHC, such as progesterone and estrogen receptors, ki-67 and HER2 protein expression, and to do bright-field analysis for HER2 gene copy number." And all this will be possible using a standard microscope. Introducing deposition of metallic silver to localize HER2 genes "takes bright-field detection to another level of sensitivity, making visualization of individual gene copies possible," Dr. Tubbs says.

He and his colleagues recently reported two abstracts on the performance of the assay, called EnzMet GenePro, one at the 2004 San Antonio Breast Cancer Symposium and the other at the 2005 U.S. and Canadian Academy of Pathology meeting. A paper reporting validation of the assay compared with FISH and its reproducibility among observers is in press in the American Journal of Surgical Pathology.

"We’re very enthusiastic about chromogenic gene detection," Dr. Grogan says. "We’re now working on the launch of the chromogenic assay for detection of the HER2 gene."

Assays that predict prognosis and response to therapy for breast cancer will become more complicated in the future. "We still don’t know how trastuzumab works," Dr. Bloom points out. "Our basic screening test is intended to define who is eligible for trastuzumab therapy, not who will respond." Predicting response is the next goal. Accumulating data point to key pathways and downstream molecules that are altered by HER2 overexpression. "Assessing downstream molecules will most likely be done by IHC," Dr. Bloom believes.

"Not all patients even with high levels of HER2 expression respond to trastuzumab," Dr. Thor says. At most, HER2-positive tumors show a 50 percent response rate to trastuzumab plus conventional chemo therapeutic agents. "Other biological tests are being developed to see who will respond to trastuzumab," she adds. Dr. Thor explains that HER2 (also called erbB2) belongs to the receptor tyrosine kinase, or RTK, type 1 gene superfamily. It can induce signaling independent of a ligand, particularly when in a highly expressed state. It can also bind to other members of its family—epidermal growth factor receptor (EGFR, erbB1), erbB3, and erbB4—whose activity and binding affinity are modulated by ligands. FDA-approved drugs exist against some of these other targets, including drugs that target more than one member, such as gefitinib. Trials are underway to determine whether patients get additional benefit using combined therapeutics.

Dr. Thor notes that EGFR is an especially important family member that indirectly interacts with estrogen via its upregulation of critical growth factor ligands that bind to EGFR. "Interactions between hormonal promotion of tumor growth and RTK-associated promotion of tumor growth is an active area of investigation," she says. Clinically, patients who have estrogen receptor-positive cancers that are resistant to tamoxifen often have abnormalities of EGFR and HER2/erbB2. "So this strong biological signaling link has potential practical significance," she says.

"Probably four to five years from now," she continues, "we will be using a whole different testing system." Like Dr. Bloom, she sees future tests looking at downstream intermediates in the signaling pathway, rather than the receptors themselves, to see whether the pathways are active. Or measurement of downstream intermediates may be combined with one agreed-upon test for HER2.

While agreeing that multi-analyte signatures will be the key to future breast cancer evaluation, Dr. Tubbs foresees them not only in the form of protein assays, but as genomic or gene expression signatures, including expression arrays, array-based comparative genomic hybridization, or both. Such molecular analyses will be in a "low-complexity" form, he predicts—a few genes at a time, rather than hundreds or thousands. And such low-complexity signatures may be readily adapted to conventional IHC and in situ hybridization formats, but with a twist—slide-based cluster analysis.

He identifies another trend with breast cancer and other tumors: finding a way to integrate the information we have now, along with low-complexity genetic signatures, into a weighted pattern that correlates more precisely with malignant potential, prognosis, and response to therapy. "We in the pathology community need to think about how to do hierarchical clustering of those markers," he notes. Dr. Tubbs points to the work of pathologist David Huntsman, MD, of the University of British Columbia, on hierarchical clustering of IHC biomarkers as an example of such work (Makretsov NA, et al. Clin Cancer Res. 2004;10:6143-6151). "I think that integrated hierarchical cluster analysis is the future of biomarkers, rather than single analyte interpretation in an isolated fashion," Dr. Tubbs says.

Summarizing the IHC/FISH debate, Dr. Allred says, "If done correctly, each assay gives comparable results." And each has an area where it gives low-level abnormal results, he adds. "Even in good laboratories, those areas are relatively difficult to interpret. At the end of the day," he concludes, "FISH is a surrogate for the true target, overexpression of the protein. So to try to make the case that FISH is a more accurate test doesn’t make sense."

Dr. Bloom agrees that IHC and FISH are very good tests and that both require training and experience. For oncologists, he uses this analogy: "If you asked a radiologist which is better, MRI or CT, the radiologist would say they are complementary." Somehow, he says, "we need to alleviate the fight between IHC and FISH in pathology.

"We’ve done a huge disservice in letting oncologists preach to us which test we should use," Dr. Bloom contends. "As laboratory professionals it is incumbent on us to step up and say we understand the tests we are using and that we will choose the best methodology based on the submitted sample." When he talks to oncologists, he finds most will accept that. He has seen laboratorians who initially could not convince oncologists that HER2 testing by IHC is reliable who then did both tests on the next 50 cases and showed a correlation so tight that the oncologists were convinced they didn’t need to perform primary FISH testing or reflex all IHC-negative cases. Unfortunately, Dr. Bloom adds, "Our passiveness has allowed oncologists and others to dictate testing methods. Pathologists reflexing all IHC-negative cases to FISH are shirking their responsibility as the laboratory expert. I don’t think that serves the patients best."

William Check is a medical writer in Wilmette, Ill.