To the editor: We read with great interest and some concern the paper by Gown et al1 in this issue of Modern Pathology regarding their commercial laboratory quality experience with HER2 testing. Gown et al tested 6604 breast cancer patients in the PhenoPath Commercial Lab for HER2/neu status. The cases were derived from 100 different hospitals and cancer centers in 29 different states in the form of paraffin block or cut sections of core biopsies, surgical specimens, or metastatic sites. They used the Dako polyclonal antibody A0485 (component of the HercepTestâ„¢) and performed FISH assay using the Vysis probe on equivocal cases, as per standard guidelines. The authors describe the quality of the immunohistochemical data obtained using AO485 antibody as an ASR (i.e., without other components of the Herceptâ„¢ kit) and the Vysis FISH method, and demonstrate unacceptable lack of correlation between these two methods at a level of 30.6%.

The authors then describe the common occurrence of HER2 staining of normal breast elements (although the number of cases showing HER2 staining of the normal breast epithelium was not specifically quantified in their paper). The authors go on to devise a method to ‘correct’ for this anomalous HER2 staining by visually ‘subtracting’ the degree of staining of the normal epithelium from that seen in neoplastic cells. The authors do not mention if they applied the normalized scoring system to biopsies in which no normal epithelium was present. They replace the FDA/Hercept test scoring system (0, 1+, 2+, and 3+) with what they term a normalized scoring system, which they derived by visually subtracting the score, that is intensity of staining of the normal epithelium from the score of the tumor cells. According to the authors, using the normalized scoring system reduced the discordance rate of immunohistochemistry and FISH for the 3+ group from 30.6 to 5.3%. There was no significant change in discordance of the negative tumors having scores of 0 or 1+.

The group concludes that using the normalized scoring system gave them a 94.7% concordance between immunohistochemistry and FISH for negative and positive results. They suggest the introduction of this normalized scoring system and propose that changes should be made to the FDA scoring criteria to meet the new ASCO/CAP mandate for 95% concordance for test validation.

There are several concerns with this proposal. First, Gown and colleagues have vast experience and may well be able to apply consistently the normalized scoring system subtraction method, whereas for many others this approach would introduce a huge variable that will be a step backward in the attempts of our discipline to enhance the quality and consistency of HER2/neu testing in breast cancer. Our belief is that the normalized scoring system method described by Gown and colleagues is built on a wrong premise, and although of some academic interest, it should not be adopted for the following reasons:

  1. 1)

    This approach (normalized scoring system) is in conflict with the recommendations of organizations (public and private), including and not limited to the College of American Pathologists and American Society of Clinical Oncology (CAP/ASCO), which have made concerted efforts to improve the standardization in HER2/neu testing using immunohistochemistry. The ASCO/CAP guidelines categorically state that staining of the normal epithelium is one of the exclusion criteria to reject a test result.2, 3

  2. 2)

    HER2 gene is unique in that the most common cause of protein overexpression is due to gene amplification (score ≥2); this is rarely, if ever (<3% of cases), due to the enhanced transcription of a normal HER2 gene count.4 Another 1–2% of cases with polyploidy of the tumor may produce enough protein to result in a positive immunohistochemical test that is associated with an ISH score of <2. These cases should not be counted as false positive because they have shown to respond to Herceptin therapy.5

  3. 3)

    Immunohistochemistry is a qualitative test, which in the case of HER2/neu is used to produce a semiquantitative result based on technical and clinical validation. The use of the normalized scoring system will introduce another variable for an immunohistochemistry semiquantitative score, which adds to the subjectivity of the test that is basically qualitative by nature.

  4. 4)

    The ASCO/CAP guidelines2, 3 are the result of an extensive review of the literature that was carefully performed by an expert panel (oncologists and pathologists) bringing their experience to the table. Strict guidelines were proposed, dealing with the pre-analytical, analytical, and post-analytical aspects of HER2/neu testing, to enhance the quality of the test and to ensure consistent and accurate results in at least 95% of cases with positive or negative HER2/neu status. This consensus has proven to be of general value and should not be changed without careful consideration.

  5. 5)

    Immunohistochemistry standardized methodology, if calibrated carefully using cases or cell lines with positive, negative, or borderline status, produces a test with no staining of the normal breast epithelium.6, 7 This is a pragmatic method that makes it much easier and more straightforward to score the percentage of positive cells and the intensity of staining in the tumor. The 3+ category requires strong complete (chicken-wire) membrane staining in at least 30% of cells. In most positive cases, 90 or 100% of cells are actually positive. The acceptance of 30% cutoff in the ASCO/CAP guidelines is to allow for the loss of the protein that may be associated with fixation.8

  6. 6)

    The authors clearly highlight the problem that labs in North America face with up to 20% discordance between labs.9, 10, 11 The causes of this discordance are manifold. The most important cause may not be the scoring criteria but rather the inadequate appreciation (or disregard) of standardized methodology, coupled with limited participation in external QA programs. As Gown et al themselves have indicated in a previous study, strict adherence to scoring criteria and calibrations of the test, plus the use of internal controls and participation in external QA programs can result in excellent immunohistochemical/ISH concordance results of >90%; many labs have achieved that level according to published QA programs.12, 13

  7. 7)

    The ASCO/CAP guidelines specifically state that staining of the normal epithelium is a reason to reject a test result. The HER2 staining of normal breast elements reported by Gown et al differs from general experience where laboratories set up the assay in such a way that normal breast epithelial staining is absent or minimal. Under these ‘usual’ circumstances, the normalized scoring system subtraction is not applicable, as normal staining is not present to an appreciable degree. Thus, cases showing the staining of the normal epithelium should be rejected and the necessary corrective taken.

  8. 8)

    The use of ASR products in non-expert hands, where validation is difficult, is fraught with a myriad of potential variations stemming from pre-analytic factors that have given rise to the poor immunohistochemical–FISH concordance rates that the authors cite. In a commercial reference, the laboratory environment, such as with Gown et al, does face a ‘unique’ problem in performing immunohistochemical staining for HER2 on patients subjected to remarkably variable, and entirely unknown fixation protocols. This is an indicated need as intervention to assure that patient specimens have optimized pre-analytic tissue fixation protocols.

  9. 9)

    The ad hoc committee on standardization of immunohistochemistry has recommended the standardization of all aspects of the immunohistochemical method, preferring the use of test platforms (IVD) that have both technical and clinical validation and that are common across multiple laboratories.

Conclusion

For the reasons cited above, we hope that the results reported by Gown and colleagues will not be adopted by pathologists, who may obtain a better outcome by attending closely to all aspects of their testing procedure, including the pre-analytic, analytic, and post-analytic phases, rather than attempting to adopt the normalized scoring system approach, which adds a new level of subjectivity to what already is a difficult assay.

Although the goal of this study is laudable, we believe that the current guidelines for HER2 scoring are appropriate and that proper intervention to assure quality tissue specimens for HER2 testing should be the goal of every laboratory.