Karen Titus
May 2025—Since the hunt began to identify low levels of HER2 in metastatic breast cancers, the action has revealed itself like the plot of a Henry James novel: Nothing much happens. Also, a lot happens. And each narrative thread is conveyed in hard-to-parse language.
The goal has been to qualify patients for the antibody-drug conjugate trastuzumab deruxtecan (T-DXd), which was shown in the Destiny-Breast04 trial to significantly improve survival in so-called HER2-low cases. Immunohistochemistry assays were designed to identify strongly positive cases, however, and thus not useful for those at the 0 and 1+ end of the spectrum.
Now, several years after the presentation of the D-B04 results at the ASCO 2022 annual meeting, which launched the low-end ship, pathologists and oncologists are adjusting to the implications of the most recent Destiny trial, D-B06 (Bardia A, et al. N Engl J Med. 2024;391[22]:2110–2122). In late January, the FDA approved the drug Enhertu for HER2-low or HER2-ultralow cases, as well as the Ventana Pathway HER2 (4B5) companion diagnostic for assessing these lower levels of HER2 in patients with metastatic breast cancer.
Closer to home, pathologists are turning to updated CAP cancer protocols to help navigate the changes—the 2025 Q1 release includes updates to the breast biomarker reporting template. And the CAP Immunohistochemistry Committee has begun offering a new product, the HERI Survey, for HER2 and ER, which should help laboratories navigate the latest interpretive challenges of these predictive markers.
The story, in short, remains compelling but dense, told in various dialects that use IHC 0, IHC >0 and <1+, IHC 0+, HER2 null, HER2-low, and HER2-ultralow. Says the chair of the Immunohistochemistry Committee, Dylan Miller, MD, “I think the language and semantics are definitely confusing for pathologists.”
As with reading Henry James, a deeper look at the HER2 saga has its rewards.
What questions did the latest Destiny trial answer? And what new questions did it raise?
“It’s a good drug,” he adds. The bystander effect of antibody-drug conjugates such as T-DXd is “a great concept.”
The primary goal of the D-B06, says Kimberly Allison, MD, professor of pathology and director of breast pathology at Stanford Medicine, was to look at bringing T-DXd as an earlier line of treatment option for metastatic ER-positive breast cancer. The trial also expanded enrollment to cases that had lower levels of staining than those of D-B04. Instead of limiting enrollment to 1+ and 2+/ISH negative, which are considered low, they enrolled those that were part of the IHC 0 category—specifically, IHC 0 with membrane staining.
The researchers refer to those cases as ultralow, though that term isn’t used in guidelines, protocols, or reporting templates. If D-B04 created a rugby scrum out of 0 and 1+ cases, D-B06 has done the same for 0s.
As Dr. Allison explains, current ASCO-CAP scoring and FDA-approved scoring have two categories within the IHC 0 group: 1) no staining observed, and 2) membrane staining that is faint/barely perceptible in less than or equal to 10 percent of cells.
“They haven’t changed the scoring,” she says. “In some ways that’s good news—we don’t have to use a different scoring system to identify patients who are eligible for T-DXd for this indication.”
It did, however, create a problem with reporting, she says. Currently IHC 0 is reported as a negative IHC 0 score—it doesn’t demarcate the two different categories.
One possible response to the D-B06 results and FDA approvals would have been to revise the ASCO-CAP guidelines. “Do we get the guidelines committee together again? Do another rapid update?” says Dr. Allison. But revising guidelines and getting the word out is like taking on War and Peace. Says Dr. Allison: “It takes a year or two.”
Since this was primarily a reporting issue, the CAP instead released, on March 19, an updated breast cancer biomarker reporting protocol.
“Essentially, for HER2 IHC reporting, for each score we have subcategories,” says Dr. Allison, who worked on the update along with Yale pathologist Uma Krishnamurti, MD, MBBS, PhD. Each is defined specifically, enabling pathologists to skirt the words “low” and “ultralow” that have entered mainstream oncology discussions.
The low and ultralow terms are mostly used by pharmaceutical companies and the oncology trials. The ASCO-CAP guidelines and CAP cancer protocols have continued to use the numeric score, letting oncologists interpret that in the context of the definitions used in the FDA indications and the clinical trials. Looking at the definitions of what each staining pattern describes does provide clarity, says Dr. Miller, professor of pathology, University of Utah, and IHC lab director for Intermountain Central Laboratory. “That doesn’t necessarily mean we’re all going to agree, and that there won’t be interobserver variability. But the definition is pretty clear.”
In the updated protocol, says Dr. Allison, a negative 0 score requires the pathologist to select either no membrane staining detected or membrane staining detected, with the actual specific wording used as defined in the ASCO-CAP scoring and as defined in D-B06. “The whole definition is there,” she says.
“We did the same thing for IHC score 1+,” she continues, “where you check the box and it’s negative, but underneath that you have to say the specific pattern that was present.” The category of equivocal score of 2 contains multiple options, or different definitions, “of how you could get to that 2+ score,” Dr. Allison says. The definition is also spelled out for 3+.
Providing the definitions should help guide pathologists to the correct score, she says. “It’s gotten complicated,” especially for pathologists who are toggling between breast and gastric scoring criteria. “I think it’s really helpful to have the specific definition, and the box that you check, for how you got that score.”
The definitions are not just in the notes anymore, she adds. “They’re brought right into the box you need to check. And then there are of course instructions embedded in the reporting worksheet tool that tell you relevance of, Why do we care if it’s no membrane staining versus a little bit? And it describes what low and ultralow are.”
Dr. Allison says she suspects the makers of T-DXd would prefer pathologists use the word “ultralow” in their scoring; that same desire came up with the word “low,” too, when that became of interest after the D-B04 trial was published. “But what we opted to do was have a standardized comment.” Ideally, this will help pathologists respond to clinical colleagues “who ask for ‘ultralow testing,’” she says with a laugh.
That, at least, is “how we’re handling it now,” she continues. The comments are optional, and labs can customize their own comments as well. But given the evolving indications and various shifts in language, “we thought it would be good to have standardized verbiage.”
“It will be more wordy,” she acknowledges. “I’m hoping those comments won’t be garbled over the LIS in some way.” And while standardized language can create clarity, it can also add complications, since HER2 testing is not specific to T-DXd. Using branded terminology across the board for just one indication would be a misstep, she says.
It’s too early to assess how well the new approach works—Dr. Allison hasn’t had a chance to use it in her own practice. It takes several months for such a change to reach vendors and then to go live in each LIS, she says. “So pathologists and oncologists need to ask that it be pushed through their systems locally,” given that it may not be a priority for the IT team.
That means confusion about how the 0+ category is identified will likely persist for a bit. “I think folks have handled it ad hoc, in different ways,” she says. Some might add a comment to an IHC 0 category that says, for example, “‘There is some faint staining, and so it’s considered ultralow.’”
Others are trying to include them in the 1+ group, Dr. Allison says. “Which isn’t per guidelines, but if you explain what you’re doing for this temporary phase, that’s probably OK.”
Dr. Miller thinks the protocols are a helpful way to stratify the 0 cases into two categories based on no expression (0 null) and weak staining, though he says the 0+ term used for the latter category is “not the best term. But they had to pick something to designate that ultralow range.” Moreover, he says, it does help with the problematic gray area that emerged after D-B04. “Now, separating 0 from 1 isn’t as big of a deal. If it is 0 with a little bit of expression, or enough expression to call it 1+, either of those are eligible. So clinically that’s a less important distinction.”
Dr. Miller and others on the Immunohistochemistry Committee are taking on these challenges as well.
In the old paradigm, Dr. Miller says, “All we really cared about were the strongly positive cases.” But with interest now focused on the lower end of the biological continuum, “we’re trying to bin what is a biologic continuum of expression into discrete categories.” (Or in Jamesian terms: Where does Isabel Archer fit in?)
He adds: “That’s always been a challenge with HER2. But now it’s especially relevant because there is always gray area at those thresholds, and that hampers reproducibility. So that’s what our committee is wringing our hands over,” he says. “How do we help labs with this?”
Looking at the current accreditation program anatomic pathology checklist, he notes that requirement ANP.22970, Annual Result Comparison–Breast Carcinoma, has long called for laboratories that read HER2 to compare their results to those of other laboratories through a comparison with published benchmarks. But as Dr. Miller notes, “We don’t have reliable published benchmarks for the ultralow or the low HER2 range.” ANP.22975, Immunohistochemical Predictive Marker Interpretation, is a new requirement as of 2024 and calls for a quality assessment to be done annually for each pathologist interpreting specific predictive markers to ensure consistent interpretations among pathologists.
But the various quality frameworks—analyte-specific quality assessment, proficiency testing, alternative performance assessment—apply to different aspects of testing, and finding the right fit with lower-end HER2 testing has been challenging, he says. “Big picture, what we can try to do is level set and calibrate our thermostats, so to speak, so that we’re all comparing apples to apples when we’re looking at these scores.”
One approach is to do exercises in which pathologists gather to look at the same slides and discuss interpretation and differences, and come to a consensus on how to interpret gray areas, artifacts, and other challenges. “That can narrow the reproducibility window,” he says.
Stepping up that approach, the committee is developing Survey products that offer whole slide digital images. By removing the variability inherent in assays and stainers, Dr. Miller says, pathologists should be able to focus on interpreting the same image, their analysis, and their scoring. The first product, the HERI Survey, is for HER2 and ER. Products for other predictive markers are in the works, he says. “We’re calling them interpretation-only Surveys for predictive markers.”
Dr. Allison, for one, is a fan. It’s unreasonable to tie ultralow detection to certification, she says. But practicing proficiency in scoring online images, while essentially skipping the technical component, “makes a lot of sense.”
This approach addresses another of the committee’s concerns, Dr. Miller says: that low-end accuracy will suffer if laboratories try to identify cases in isolation. “We try to set a ground truth or a gold standard for these by consensus among the experts on our committee.” So while pathologists should compare results among others in their institution, “there should also be a ground truth to be sure they’re all on target and not collectively an outlier. That’s how we’re approaching this and helping people get more reliable around those gray area thresholds.”
Reproducibility will never be fully tamed, he continues. “Everyone’s instruments will vary a little day to day.” That so-called drift phenomenon is not reproducible. There’s also wide variation from lab to lab in terms of which clones are used, as well as settings on autostainers. Nor is every case fixed the same way. Even environmental factors, including altitude and humidity, “can all have an effect,” Dr. Miller says. “And frankly, I’m not sure we’ll ever be able to get a handle on those things.” With those technical aspects out of bounds, “We’re focusing on the pathologist interpretation factor as much as we can, understanding that if we are all looking at the same slide, we can at least get close to agreeing on what that individual slide shows. Even if quantitative methods are the future, we’re being asked to do testing with the environment and circumstances we have now.”
Adds Dr. Miller: “These markers with companion diagnostic tests are increasing exponentially. We are doing our best to keep up, but it takes time to gather the data about how reproducible the assays are and to gather materials that can be used in a Survey.”
After D-B04, Dr. Rimm says, the ambiguity around IHC 0 versus 1 opened the door to what he calls “persuasion medicine.”
He says he’s aware of cases where directors of breast pathology received calls from oncologists asking them to review cases that were read as an IHC 0.
Dr. Rimm doesn’t want pathologists to be persuaded. “That’s not how it should work. We shouldn’t be pressured to give a diagnosis so a drug can be given, but rather, the drug should be given based on the biology of the tumor.” The Yale pathology lab, he reports, didn’t change its 0 rate after the publication of D-B04. At the same time, he’s quick to recognize, “We didn’t have very many other options,” given the limitations of IHC assays on the low end of the spectrum.
Many assumed D-B06 would offer more clarity, but Dr. Rimm says he thinks AstraZeneca made a mistake—“although they didn’t know this when they designed the trial”—in not giving the drug to all HER2-nonamplified cases. The low category was already problematic, he says. The data he’s seen on laboratories’ performance in distinguishing between 0 and 1 suggest labs can’t do better than 80 percent. “That’s still one in five that would be inappropriately called,” he says, “and then the therapy would depend on the pathologist, not the biology of the tumor.” The current 0 conundrum makes a difficult situation even worse, he says. “They’re asking us to tell an even more subtle difference apart.”
Not only is it more subtle for pathologists to read, but these samples are subject to preanalytic variation and stainer variance, he adds, echoing Dr. Miller. That variance concerns Dr. Rimm even more than the difficulty of individual pathologists’ readings. A true 0 in one lab might be an ultralow in another lab, given that labs do not standardize their stainers.
Why weren’t IHC 0 cases included in D-B06? Dr. Rimm isn’t the only one who was disappointed, as correspondence from Dr. Allison and colleagues in the New England Journal of Medicine (Wolff A, et al. N Engl J Med. 2025;392[8]:829) made clear. (And, as the authors note, “The shorthand ‘IHC >0 and <1+,’ which was used to describe IHC 0 with some staining, is an oxymoron.”)
D-B06 followed the concepts used in the initial HER2-positive trials for T-DXd, which required a positive result for enrollment, Dr. Allison explains. D-B06 likewise assumed that only patients with the target should be included. “Logically, that makes some sense if you’re using it as a biomarker,” Dr. Allison says. “But the big concern—and pathologists recognize this—is that IHC no staining is not a true result.” It’s not what the test was designed to do—the limit of detection is not sufficiently sensitive in the lower range. In fact, she says, in D-B06, cases that were IHC 0 locally were low or ultralow on central retesting more than 60 percent of the time.
The Destiny-Breast06 results are new enough that there are no peer-reviewed studies looking at concordance and accuracy of ultralow as a category. “But I know in general pathologists are not pleased about it,” Dr. Rimm says with a laugh.
Rather than sifting through the IHC 0 and IHC 1+ cases like miners panning for gold, pathologists might be better served by tests that get to the heart of HER2 biology, Dr. Rimm says.
At Yale, he and colleagues originally set out to provide measurements in attomoles per square millimeter of HER2. But it made sense to add the TROP2 target to their quantitative fluorescence assay, to make it a selective assay rather than just a measurement assay. This can be used for two approved TROP2 targeting therapies: sacituzumab govitecan (SG) and Dato-DXd. The New York State Department of Health recently granted approval for the assay, called Troplex, which measures TROP2 and HER2 in attomoles per square millimeter on a single slide.
With the list of approved drugs likely to grow, Dr. Rimm says, pathologists will need to help oncologists figure out which drug should be used first in each patient. He’s already posed that question to his clinical colleagues.
Their response? “I get different answers,” he says. “There is no way to decide.” Some talk to their patients about which toxicities might be easiest to tolerate. Others say they compare trials, “even though they’re not supposed to.”
He says the drug pipeline is moving quickly. “I think we’re just seeing the tip of the iceberg.” More than 250 known antigens are being tested as ADC markers, he says, referring to a review by Heather Maecker, et al. (Maecker H, et al. MAbs. 2023;15[1]:2229101). “There are a lot of potential ADCs out there,” he says. And the approvals are already moving beyond breast and earlier than the metastatic setting.
Dr. Rimm cites the work of Laura Huppert, MD, of UCSF, to help make sense of issues related to drug sequencing; her work showed that whichever ADC is given first is more effective (Huppert LA, et al. NPJ Breast Cancer. 2025;11:34). He’s also hopeful that the Trade-DXd Trial, by the Translational Breast Cancer Research Consortium, will provide useful answers. The trial is looking at sequencing of T-DXd and Dato-DXd.
“But what happens when a third or fourth drug comes out in that same setting?” Dr. Rimm asks. There are currently seven different trastuzumab ADCs in various stages of clinical trials, he says. “I don’t have an answer for how you’re going to tell which of those to use first. But when they have two different targets, at least you can look at the biology of the tumor and see which target is more prevalent.”
Like any outside-the-box thinker, Dr. Rimm likes to look both forward and back. He remains troubled by the lack of precision among pathologists. An assay like a blood glucose typically has a coefficient of variation of less than 1, “let alone less than 10,” he says. Bioanalytic assays such as a CEA or CA 19-9 have higher CVs, though they’re steadily improving.
But in preliminary research in his lab looking at the CV among pathologists, he and his Yale colleagues have found it to be “in the high double digits. It can exceed 400 in some cases, which is just scary.”
Hence his dogged interest in pursuing methods that would enable quantitative measurement of cells. “Quantitative assays are on their way. It’s just a matter of time. Then we can measure the CV of the assay, and we can give patients and oncologists, with confidence, an outcome measurement instead of a reading. So I predict that measuring will be more and more important in anatomic pathology,” echoing the changes that happened in clinical chemistry in earlier decades. “Now one wouldn’t even think of doing a nonquantitative glucose.”
Though his work involves quantitative fluorescence, “There are other people who are working on this as well, with other tools,” he says. Not only will ASCO-CAP guidelines reflect the growth of such testing, he predicts, but “I think oncologists will demand it. If you give a reading, and your competitive lab gives a measurement, they’re probably going to go to that lab as opposed to bringing the specimen to you.”
He’s not the only one anticipating the need for new tests. But given the extremely complex predictive biomarker landscape, Dr. Allison says she’s thankful for the relative simplicity of HER2, where one test can be used for all indications. In the future, “I hope it remains simple for the things we’re good at,” she says.
“And for the things we’re not good at,” she continues, if there does turn out to be a threshold that matters, “we are going to need more sensitive assays.”
She also envisions use of panel-based, scenario-based biomarkers for cases where there are multiple ADC options with different protein targets. But the need for granularity is not a given. “There are very powerful, sensitive ADCs that don’t need a lot of protein present to do what they need to do,” she says, noting that the SG ADC indication for TROP2 in breast was approved without a biomarker requirement, for example.
Dr. Miller has his own thoughts about the need for new assays. One potential approach would be to mimic the approach used for PSA or troponin, he says, where the normal assay is used for screening but a highly sensitive assay is used for another specific purpose.
In the meantime, pathologists will need to respond to the demands of the here and now.
Clinicians tend to have three main questions, Dr. Allison says. The first is: How are we reporting this?
The second question is: Should we keep testing? If a case is reported as IHC 0—Dr. Allison refers to it as the historic IHC 0—should the pathologist examine it again and report it more specifically, or perhaps do an additional round of staining on the same sample?
The other common question is: How many samples do you need to get a non-0/no staining result? Dr. Allison is sympathetic. “They’re looking for treatment options in metastatic patients. And I think pathologists sometimes forget it’s not really relevant in the nonmetastatic setting.” For such patients, asking whether additional testing can push them toward a useful treatment is fair game.
She’s seen data at this year’s USCAP meeting and at earlier San Antonio Breast Cancer Symposium meetings that show that testing multiple samples can invariably achieve a non-null result. This suggests to her that most IHC 0s are false-negatives, due to ischemic time, for example, or insufficiently looking for the focal area of staining.
That even a tiny amount of staining is considered relevant gives her pause. Perhaps this isn’t a targeted therapy in the way that the medical community typically has thought about targeted therapies, she says. For HER2 overexpression, for example, “you’ve got something that’s an oncogenic driver. It’s highly expressed; you’re targeting it with antibodies and antibody-drug conjugates.” But an ADC is more of a mechanism of drug delivery, she says, and because of the huge bystander effect, very little protein needs to be present. “Eventually we may not need to test at all. But we’re not quite there yet.”
Where does Dr. Rimm think this will end? “I think we’re more likely to go back to the original ASCO-CAP guidelines and just do 0, 1, 2, and 3. And then [in D-B15] they will either get approvals for the 0s or not.”
In short, a lot could happen. And it could also look like nothing has happened.
Henry James would approve.
Karen Titus is CAP TODAY contributing editor and co-managing editor.