Reproducibility of Equivocal HER2 In Situ Hybridization Groups 3 and 4, and Comparison With HER2 mRNA Expression: The Thin Line Between Amplified and Nonamplified Breast Cancers

Approximately 15% of invasive breast cancers (BCs) involve amplification of the human epidermal growth factor receptor 2 gene (HER2) and overexpression of the corresponding protein.^1,2 A positive HER2 status is associated with poor prognosis. The development of HER2-targeted therapies has proved effective in achieving excellent outcomes for this subset of patients with BC. Recently, novel therapies called antibody-drug conjugates have also proved effective in BC with low and ultralow HER2 protein expression in the absence of HER2 amplification. Therefore, determining HER2 status is crucial for selecting the most effective personalized treatment.^3,4

The 2018 HER2 testing guideline from the American Society of Clinical Oncology and the College of American Pathologists (ASCO/CAP) recommends using immunohistochemistry (IHC) as the primary method for assessing HER2 status.⁵ When an equivocal IHC 2⁺ score is obtained, which occurs in about 15% to 20% of all invasive BCs, in situ hybridization (ISH) is required to further clarify the HER2 status.^1,5,6 Using dual-probe ISH testing, tumors with a mean HER2 copy number of 4.0 or greater and a HER2:CEP17 ratio of 2.0 or greater are classified as HER2-positive (ISH group 1), while those with a mean HER2 copy number less than 4.0 and a HER2:CEP17 ratio less than 2.0 are considered HER2-negative (ISH group 5).⁵ Three ASCO/CAP ISH groups are regarded as equivocal and require additional workup.⁵ ISH group 2 tumors have a mean HER2 copy number less than 4.0 and a HER2:CEP17 ratio of 2.0 or greater. ISH groups 3 and 4 each have a HER2:CEP17 ratio less than 2.0, with ISH group 3 tumors having a mean HER2 copy number of 6.0 or greater and ISH group 4 tumors showing a mean HER2 copy number of 4.0 or greater and less than 6.0.⁵ These equivocal ISH results require correlation with the HER2 IHC. In case of a double equivocal result, that is, an equivocal ISH result in combination with an IHC 2⁺ score, a second observer blinded to the initial ISH result must evaluate at least 20 nuclei. If the equivocal ISH result is confirmed by the second observer, ISH groups 2 and 4 are considered as HER2-negative with a comment without further testing on the same sample.⁵ It is uncertain whether patients in this ISH group benefit from HER2-targeted therapy in the absence of protein overexpression.^6–12 In contrast, ISH group 3 tumors with an IHC 2⁺ score are regarded as HER2-positive in case of confirmation of the ISH result. If the additional observer’s review shifts the count into a different ISH category, the final category should be determined through adjudication following internal procedures.⁵ However, the 2018 ASCO/CAP guideline does not specify the workflow of such an internal procedure.⁵

High reproducibility in ISH quantification has been reported, with a study by Polonia et al¹³ showing 97% concordance (k = 0.879). However, this study included few cases from ISH groups 3 and 4, and interobserver agreement for these groups remains unclear. To address HER2 ambiguity in ISH group 4, alternative assays like genomic profiling and real-time quantitative polymerase chain reaction (RT-qPCR) have been proposed. Studies comparing ISH and next-generation sequencing in ISH group 4 tumors have shown no detectable HER2 amplification by next-generation sequencing, regardless of the ISH result.^8,14 Additionally, discordant findings have been observed in mRNA quantification for ISH group 4, with studies showing varying relationships between mRNA levels, IHC scores, and ISH status.^6,9,15

OBJECTIVE

The objective of this study was to assess the interobserver reproducibility of ISH groups 3 and 4, using the originally reported score as well as scores determined by 2 additional observers who were blinded to the initial results. Additionally, we aimed to assess the HER2 mRNA expression level by using the MammaTyper RT-qPCR kit. The integrated HER2 IHC/ISH result, as determined by each observer, was compared to the HER2 mRNA expression level determined by MammaTyper.

MATERIALS AND METHODS

RESULTS

Clinicopathologic Characteristics

One male and 49 female patients with BC were included, with a median age of 61 years (range, 22–91 years). Tissue samples were collected from various sources: 30 (60%) were obtained from surgical resections, while 20 (40%) came from core needle biopsies. Of these 50 samples, 45 (90%) were taken from primary tumors, 4 (8%) from lymph node metastases, and 4 (8%) from distant metastases (liver, meninges, ovary, and skin). The most common histologic type was no special type (NST) BC, accounting for 42 of 50 (84%), followed by invasive lobular carcinoma at 4 (8%), of which 2 were pleomorphic and 2 were classic-type invasive lobular carcinoma. The other tumor types comprised 2 micropapillary carcinomas (4%), 1 mixed NST/micropapillary carcinoma (2%), and 1 invasive papillary carcinoma (2%) (Table 1). Most tumors were hormone receptor–positive: 46 (92%) were ER-positive, and 37 (74%) were PR-positive. In terms of HER2 IHC, 46 samples (92%) were scored as 2⁺, 3 (6%) were scored as 3⁺, and 1 (2%) was scored as 1⁺. For the ISH classification, 28 (56%) were categorized as group 4; 13 (26%) as group 3; 5 (10%) as group 1; and 4 (8%) as group 5. No ISH group 2 cases were available in our institution. The amplified and nonamplified BC samples were randomly added to obtain a heterogeneous study cohort, aiming to prevent bias of the observers involved in the interobserver variability study. Most tumors had an intermediate to high Ki-67 index (n = 38, 76%) and low level of TILs (n = 44, 88%) (Table 1).

Table 1.Clinical and Pathologic Characteristics of 50 Patients With Breast Cancer

View Fullscreen

ISH Interobserver Reproducibility

To assess the reproducibility of ISH, with an emphasis on the equivocal ISH groups 3 and 4, the ISH groups of the original report were compared to the additional observers included for this study. Figure 1, A illustrates the comparison between the ISH groups from the original report and those as assessed by observer 1. The overall concordance was 22 of 50 (44%). Groups 1 and 5 remained the same. In the discordant cases, 8 of 13 cases originally classified as ISH group 3 were assigned to a different group by observer 1. Specifically, 1 case changed to group 1 (amplified), 6 cases changed to group 4, and 1 case changed to group 5 (nonamplified). Furthermore, 20 of 28 cases originally classified as ISH group 4 were assigned to a different group by observer 1. From this, 1 case changed to group 1 (amplified), 1 case changed to group 3, and 18 cases changed to group 5 (nonamplified).

View Figure

• Download as Powerpoint
• Download Figure

Figure 1, B illustrates the comparison between the ISH groups from the original report and those assessed by observer 2. The overall concordance was 16 of 50 (32%). Group 5 remained the same among observers. In the discordant cases, 1 of 5 cases originally classified as ISH group 1 (amplified) was assigned to ISH group 5 (nonamplified) by observer 2. Moreover, 13 of 13 cases originally classified as ISH group 3 were changed to other groups by observer 2, including 7 cases to group 4 and 6 to group 5 (nonamplified). Furthermore, 20 of 28 cases originally classified as ISH group 4 were assigned to a different group by observer 2. From these, 4 cases changed to group 1 (amplified) and 16 cases changed to group 5 (nonamplified).

After integration of ISH and IHC data, the observed discordances between the original report and observer 1 resulted in 7 of 50 cases (14%) changing from a positive to a negative HER2 status, and 2 of 50 cases (4%) changing from a negative to a positive HER2 status. From the original report to observer 2, the data indicate that 14 of 50 cases (28%) changed from a positive to a negative HER2 status, and 4 of 50 cases (8%) changed from a negative to a positive HER2 status. In other words, of the 50 patients, 9 (18%) and 18 (36%) could have been treated differently, based on the counts of observer 1 and observer 2, respectively. Nine of 28 ISH group 3 and 4 cases (32%) fell into 3 different ISH groups, based on the initial report and the ISH recounts by the 2 additional observers. Five of these were initially reported as ISH group 3, among which 4 patients with BC might not have received anti-HER2 treatment, based on the majority opinion. The remaining 4 BCs were ISH group 4 cases for which the treatment would not have been altered by the majority opinion (Table 2).

Table 2.In Situ Hybridization (ISH) Groups for 9 Breast Cancers With Discordant Results Based on the Original Report and the Recounts by Observers 1 and 2, With HER2 Status Based on the Majority Opinion

View Fullscreen

Based on the overall majority opinion in the hypothetical scenario of 3 observers per double equivocal BC, 7 of 41 patients with BC (17%) would have been reconsidered as HER2-negative, and 1 of 41 patients with BC (2%) would have been reconsidered as HER2-positive, resulting in a potential treatment change for 8 of 41 patients (20%) (Supplemental Table; see the supplemental digital content containing 1 table and 3 figures at https://meridian.allenpress.com/aplm in the December 2025 table of contents). The level of discordance was not influenced by the specimen type (ie, biopsy versus resection specimen). Figure 2 shows examples of invasive breast tumors originally reported as HER2 ISH group 3 (Figure 2, A through C), ISH group 4 (Figure 2, D through F), and reclassified as group 5 by the majority opinion (Figure 2, G through I).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Invasive BCs with an equivocal HER2 IHC 2⁺ score are evaluated by additional ISH testing to investigate the presence of HER2 gene amplification. Depending on the mean HER2 gene copy number and the HER2:CEP17 ratio, tumors are classified into 5 different groups, with ISH group 1 tumors being amplified, ISH group 5 tumors being nonamplified, and ISH group 2, 3, and 4 tumors being designated as equivocal.⁵ The 2018 ASCO/CAP guideline on HER2 testing recommends to perform additional workup for such “double equivocal” tumors.⁵ As the selection of the precise tumor area wherein 20 nuclei are counted can differ among different observers, especially in resection specimens with large tumors, we questioned the degree of reproducibility of such “borderline” cases. Moreover, counting of HER2 signals can differ among observers, as this might be influenced by the distance between the individual signals. We therefore retrieved a series of double equivocal BCs, diagnosed since the introduction of the recent version of the ASCO/CAP guideline,⁵ to re-count the archived SISH slides. Four and 5 BCs, previously reported as nonamplified and amplified, respectively, were added to the series to avoid bias of the involved observers and served as a control group for the reproducibility study. For 32 of 41 ISH group 3 and 4 cases (78%), there was sufficient residual tumor tissue to allow for mRNA expression analysis by RT-qPCR. Our findings show that interobserver agreement for amplified and nonamplified cases was substantial. There was 100% concordance with the originally reported result for the 4 nonamplified BCs for each observer. For the 5 amplified BCs, concordance ranged from 80% to 100%, with only 1 amplified case reclassified as nonamplified by a single observer. This misclassification was most likely due to incorrect selection of the tumor area when counting heterogeneously distributed HER2 signals at ×630 magnification.

In contrast, the reproducibility of ISH groups 3 and 4 was limited. The combined concordance for ISH groups 3 and 4 was 13 of 41 tumors (32%) for observer 1, with 8 of 13 ISH group 3 tumors (62%) being reclassified and 20 of 28 ISH group 4 tumors (71%) being reclassified. The combined concordance for ISH groups 3 and 4 was 8 of 41 tumors (20%) for observer 2, with all ISH group 3 tumors being reclassified as ISH group 4 or 5 (ie, 0% concordance) and 16 of 28 ISH group 4 tumors (57%) being reclassified as nonamplified ISH group 5. In the case of an equivocal HER2 IHC 2⁺ score, a switch from group 4 to group 5 will not affect treatment, that is, these patients are considered to have HER2-negative BC, as there is insufficient evidence to support HER2-targeted treatment in ISH group 4 BCs.^6,22 However, a switch from ISH group 3 to ISH groups 4 or 5 will have an impact on the patient’s treatment, as IHC 2⁺ ISH group 3 tumors are considered as HER2-positive. Based on the counts of observer 1 and observer 2, in our series 9 patients and 18 patients, respectively, could have been treated differently. Our findings show that a large proportion of double equivocal HER2 BCs end up with a discordant ISH group result after evaluation by a second observer. None of the 13 initially reported ISH group 3 BCs was confirmed by both observers, and only 4 of 28 initially reported ISH group 4 cases were confirmed by both observers. Finally, HER2 status, and thus subsequent systemic treatment, could have been potentially different for 8 of 41 patients (20%) who were initially reported as having ISH group 3 or 4 tumors, if 3 observers would have been involved from the start. It seems therefore more prudent to involve a third observer, blinded to the previous results, to re-count at least 20 nuclei. In those BC cases flirting with the threshold between amplification and absence of amplification, the majority opinion could be used to establish the final decision regarding HER2 status. Of note, the observers who were involved in the ISH reassessment for this study had less experience with HER2 ISH evaluation than the observers involved in the original evaluation, which might have influenced our results.

Automated algorithms or so-called artificial intelligence (AI) could help in the automated scoring of HER2 IHC and ISH through imaging software. In the case of HER2 IHC, according to a meta-analysis by Wu et al,²³ the ability of AI to identify HER2 2⁺ was slightly inferior compared to the recognition of HER2 0/1⁺ and HER2 3⁺. This could be explained by the changes in guidelines affecting the definition of HER2 2⁺, and the greater heterogeneity in HER2 2⁺ than in HER2 3⁺. Concerning ISH, in a study by Xue et al,²⁴ the overall classification accuracy between the Aitrox model and manual counting was 85.33% (157 of 184). In group 5, the most common group, the consistency was as high as 95.90%. However, consistency was low in other groups owing to limited numbers of cases. Other models have shown similar reproducibility rates.^25,26 For now, standardized AI algorithms that can be extrapolated to different hospital settings are still required to demonstrate robustness and applicability. Of note, even across different computational pathology models for HER2 IHC, substantial variability was observed for all IHC categories (ie, 3⁺, 2⁺, 1⁺, and 0). This discordance was most likely due to these models being trained by pathologists, who do not entirely agree among themselves.²⁷

It needs to be emphasized that large-scale prospective data are not available about treatment response to HER2-targeting therapies in the adjuvant setting in this particular patient population. Follow-up data on treatment response to HER2-targeting therapies in clinical trial setting are required to guide future treatment decisions for these patients.⁶ A retrospective analysis showed that ISH group 2 cases might benefit from HER2-targeted therapies in the neoadjuvant setting, with a complete response rate of 27%.²⁸ Moreover, multiple studies have shown that the mean HER2 copy number strongly correlates with the chance of achieving a pathologic complete response after neoadjuvant chemotherapy in patients with HER2-positive BC, with lower HER2 copy numbers and lower HER2:CEP17 ratios being associated with a higher chance of residual disease.^17,29–31

Several studies have assessed mRNA quantification in ISH group 4 BCs. In a study by Tong et al,⁹ similar mRNA expression levels were reported for ISH dual-equivocal IHC 2⁺ (ISH group 4 nonamplified, according to ASCO/CAP guideline 2018⁵) and ISH-negative IHC 2⁺ tumors. Gupta et al⁶ observed that mRNA levels were higher in ISH group 4 tumors with IHC 2⁺ than with IHC 0/1⁺, although protein quantification (determined by quantitative fluorescence) was similar to that observed in HER2-negative cases. However, Marchio et al¹⁵ found significantly different HER2 mRNA levels between HER2-positive, HER2-negative, and IHC 2⁺ ISH group 4 carcinomas, with greater overlap of HER2 mRNA levels between IHC 2⁺ ISH group 4 and HER2-negative carcinomas. Given the interobserver discordance for equivocal ISH tumors demonstrated in our study, one may wonder whether ISH should be regarded as the gold standard in this ambiguous gray zone. RT-qPCR provides an objective single value for HER2 mRNA expression levels. On the one hand, RT-qPCR could cause a false sense of security by relying too much on this single value and its associated positive or negative HER2 status, determined by an arbitrary cut point. On the other hand, however, RT-qPCR might provide a more reliable view on the actual HER2 status of the tumor, as it analyzes all tumor cells, and not just a subjective selection of 20 tumor nuclei. Interestingly, we observed a difference between the HER2 status provided by the different observers and the mRNA-based HER2 status, with observer 2 showing the highest percent concordance of 72%. This observation insinuates that mRNA expression could be used in selected IHC/ISH double equivocal cases to make a final decision.

Clinicians generally want to obtain a definitive HER2 status as soon as possible. Future discussion about a potential clarification of the current version of the ASCO/CAP guideline should also consider how a timely report about the definitive HER2 status can be achieved for double equivocal cases, and how those results should be communicated. It is likely that turnaround times will increase if additional evaluation steps are added, such as implementing more observers for ISH and/or RNA-based testing.

In the past years, a novel antibody-drug conjugate targeting HER2—trastuzumab deruxtecan—has shown extended progression-free and overall survival in patients with HER2-positive, but also HER2-low, BC.^32–34 According to the European Society for Medical Oncology consensus and ASCO/CAP guideline of 2023, HER2-low BC is defined as BC with HER2 IHC 1⁺ or 2⁺ scores without HER2 amplification.^35,36 No substantial molecular or pathologic differences have been demonstrated between HER2-low and HER2-0 BCs, after correcting for the expression of hormone receptors.^37,38 For this reason, HER2-low BCs are not considered as a distinct molecular entity; rather they represent a term used for clinical-trial purposes.³⁶ In light of these findings, the final HER2 status of BC with mean HER2 copy numbers of 4.0 or greater and less than 6.0 becomes increasingly important, as patients with HER2-low ISH-negative BC could potentially benefit from treatment with trastuzumab deruxtecan, whereas patients with HER2 IHC 2⁺ ISH-positive BC are still eligible for treatment with the “classic” anti-HER2–targeting therapies such as trastuzumab and pertuzumab.

Our study has a few limitations. First, the cohort size was limited, as ISH group 3 and 4 cases are infrequently encountered and our institution did not have any archived ISH group 2 cases with available residual tumor tissue. Second, owing to the short follow-up time, we were able to retrieve follow-up data of only 25 patients with primary BC. Therefore, to determine which assay (MammaTyper versus combined IHC/ISH) correlates strongest with progression-free survival, prospective validation in a large multicenter patient cohort is required, particularly for those patients with IHC 2⁺ BC and an HER2 ISH finding bordering on the ISH threshold. Third, the level of experience of the observers, both in this study setting and in real life, might influence the obtained results. Fourth, our study used brightfield ISH instead of fluorescence ISH (FISH), and we therefore cannot extrapolate our results to FISH without additional validation. Future studies should investigate whether laboratories applying dual-probe FISH for HER2 observe a similar number of double equivocal breast cancers. We are currently exploring this via an internationally organized survey. Fifth, we were unable to investigate the impact of retesting on a subsequent specimen, as this is not routinely performed in our institution. One study on ISH group 2 cases demonstrated that 23% of ISH group 2 tumors convert from a HER2-negative to a HER2-positive status upon retesting, suggesting that tumor heterogeneity may substantially influence the HER2 ISH results.³⁹

In conclusion, the reproducibility of the equivocal ISH groups 3 and 4 is poor. Although the 2018 ASCO/CAP guideline for HER2 testing recommends adjudicating BC with double equivocal IHC and ISH results as per internal procedure,⁵ we show here that the number of observers involved in HER2 testing likely influences the final HER2 status, and thus, patient treatment. We therefore recommend to routinely involve a third observer in the case of “borderline” ISH BCs with discordant ISH results between 2 observers, and to rely on the majority opinion to establish a final diagnosis. Evaluation of HER2 mRNA is a more objective method to determine HER2 status and could be considered as an alternative solution for tumors with equivocal ISH.