Immunohistochemical (IHC) testing for estrogen receptors (ER), progesterone receptors (PR) and human epidermal growth factor receptor −2 (HER-2) is now carried out on the tissue samples of all newly diagnosed patients with invasive breast cancer in the vast majority of hospital laboratories [1, 2]. The results indicate the likely benefit to patients of hormonal therapies (tamoxifen, aromatase inhibitors) targeting estrogen driven tumors or antibody based therapies that target the HER2 oncogene (trastuzumab and trastuzumab deruxtecan) [3–5]. Testing with these markers, in addition to the cell proliferation marker Ki-67, also provides valuable prognostic information by predicting how aggressive a tumor is and consequently if a patient is likely to benefit from chemotherapy [6, 7].

The results of these IHC assays all require some degree of quantitative assessment in order to determine whether or not the tumors express sufficient amounts of receptor to indicate a beneficial response to therapies. However, issues surrounding the reproducibility of IHC testing for ER, PR, HER2 and Ki-67 and their assessment are well documented, to include the negative impact of laboratories producing erroneous results [8–14]. Consequently, to ensure the reliability of results it is recommended that laboratories conform to international evidence based quality assurance (QA) guidelines on how tumor samples should be prepared, tested and reported for these biomarkers [1, 3, 4, 15]. One of the criterion stated in these guidelines is that laboratories carrying out these tests should participate in an external quality assessment (EQA) program [16], also known as performance testing (PT), such as the programs offered by the UK National External Quality Assessment scheme (UK NEQAS) [17], College of American Pathologists (CAP) [18] and the Royal College of Pathologists of Australasia (RCPA) [19]. In Vietnam, one of the main EQA programs offered to laboratories is run by the Center for Standardization and Quality Control in Medical Laboratory of Ho Chi Minh City (CSQL) [20]. However, prior to the current study, CSQL had not offered EQA for ER, PR, HER2 and Ki-67.

In order to assess the reliability of IHC testing for these important predictive and prognostic breast cancer biomarkers and in collaboration with the CSQL we organised a pilot EQA ring study involving hospitals from the south, north, and middle of Vietnam, estimated to be approximately half of all clinical hospitals in the country testing for these markers. Moreover, as cases are often referred for testing between hospitals where there are preparatory (pre-analysis) variables such as fixation and paraffin processing, we considered it imperative that the EQA study included the testing of samples prepared in all of the participating laboratories. The logistics of instigating this type of ring study, where each laboratory simultaneously tests not only its own cases but that of every other participating laboratory, is complex and not currently offered by established EQA programs. The hypotheses are that the assays are highly reproducible; by the same laboratory on the same cases when tested on separate occasions (intra-laboratory reproducibility) and by different laboratories testing the same cases (inter-laboratory reproducibility).

The distribution of scores on each of the forty-breast cancer cases for each of the four markers, ER, PR, HER2 and Ki-67 are shown in Table 3 and images of the results obtained on four of the forty tumors (case numbers 1-4, in Table 3) are shown in Figures 1–4. Based on the IHC profile (Table 3) the cases comprised; 19 luminal A, 3 luminal B, 9 HER2 enriched and 9 triple negative invasive breast carcinomas. Across, most profiles, there were 13 tumors with HER2 low expression (1+, or 2+ with no HER2 gene amplification).

In the current study the CSQL organized an EQA ring study to determine the reliability of testing for ER, PR, HER2 and Ki-67 in ten hospitals from the south, middle and north of Vietnam, estimated to represent approximately half of all Vietnamese hospital laboratories currently testing clinical cases for ER, PR, HER2 and Ki-67. As cases are often referred between hospital laboratories for testing we tested the reliability of testing on tissues prepared in all ten of the participating laboratories comprising four different cases from each laboratory (forty tumors in total). Consequently, a more rigorous assessment than the testing of single slides, comprising smaller numbers of tissues and often prepared in just one laboratory, as offered by many EQA programs.

McHugh (2012) [26], recommends using both proportional (percentage) agreement and Cohen’s kappa coefficient for determining the level of inter-rater agreement, as both have their strengths and limitations, and we followed this recommendation in the current study (Tables 4–8).

Whilst the reproducibility of testing of the same forty different cases on two separate occasions (intra-laboratory reproducibility), was significant for all four antibodies, it was by no means identical, as shown by the relatively low kappa scores, with greatest variation seen with the testing of Ki67.

The inter-laboratory reproducibility of testing, using clinically relevant cut-points for defining likely benefit to treatment strategies, was relatively high. Almost perfect agreement (kappa score of 0.81–1.00) was achieved for the testing and scoring of ER (kappa 0.822), with 92% (CV 4.8%), of results in agreement. Only one laboratory (laboratory A in Table 5) achieved the expected result on all forty cases of breast cancer, and this for the testing of ER. Similar to the intra-laboratory results, greatest variation was seen with Ki-67 testing (Table 8).

For over twenty-years HER2 has been assessed in clinical laboratories most commonly using a combined IHC and in situ hybridization (ISH) approach that classifies invasive breast carcinomas as either HER2 positive (3+ and 2+ cases with HER2 gene amplification) or HER2 negative (0, 1+ and 2+ cases without HER2 gene amplification) [1, 3]^. More recently, the results of the DESTINY-Breast 04 clinical trial have shown significant improvement in progression free survival and overall survival for women with advanced and metastatic breast cancer with tumors expressing low levels of HER2 expression (1+ and 2+ without HER2 gene amplification) and treated with the antibody drug conjugate, trastuzumab deruxtecan (T-DXd), compared to a control group receiving the physicians choice of therapy [5]. Subsequently, T-DXd has been approved by both the US Food and Drug Administration (FDA), the European Medicines Agency and the UK National Institute for Health and Care Excellence (NICE) for patients with unresectable or metastatic breast cancer [27, 28]. Some of the latest pathology guidelines for reporting HER2 from both the US and the UK emphasize the importance of appropriately scoring HER2 low breast cancers [15, 29]. Consequently, for HER2, each of the four scores; 0, 1+, 2+, 3+ now have clinical implications and therefore each is of importance. Therefore, for the assessment of HER2 in the current EQA the expected result was the central value only, the median, which in our study was also equivalent to the mode i.e., the single discreet score achieved by the majority of participants.

Overall, 77% of pathologists scores for HER2 were in agreement (Table 6), with a kappa concordance value of 0.794 (95% CI 0.753–0.836); a level of agreement considered as substantial by many studies utilizing Cohen’s kappa coefficient to compare agreement amongst pathologists. However, the level of agreement in the scoring of just the HER2 low cases was significantly lower, with only 50% of scores in agreement and with a kappa value of 0.323 (95% CI 0.223–0.424). In comparison, the scoring of HER2 3+ and HER2 0 was almost perfect (kappa 0.879, 95% CI 0.835–0.923), with 89% of scores in agreement (Table 7). Consensus agreement amongst a group of expert UK pathologists scoring HER2 have reported similar findings [23]. These results question the reliability of the IHC assay to accurately predict benefit to T-DXd for breast cancer patients. A conclusion similar to that reached by other authors [30, 31].

The International Ki67 in Breast Cancer Working Group (IKWG) suggest that the use of Ki-67 IHC, as opposed to more expensive multi-gene assays, is adequate for determining prognosis for breast cancer patients with ER positive disease with very low (5%) or very high (≥30%) Ki-67 scores. However, they recommend that inter-laboratory variation of the IHC assays is too great to rely on intermediate Ki-67 scores falling between these two values [6]. Indeed, in the current study testing for Ki-67 resulted in greater variation, than in the testing for HER2, ER and PR (Tables 5, 6, 8). Technical differences in the assays employed by the ten participating laboratories may have accounted for some of this variation, most noticeably the use of at least four different Ki-67 antibody clones, compared to the use of just two different clones for the other markers (Table 1). Evidence suggests that some of the Ki-67 antibody clones commercially available are more reliable than others [32]. Difference in antigen retrieval stringency may also have played a role, as an appropriate antigen retrieval step is essential for this marker [6, 33]. The importance of choosing the most appropriate antigen retrieval protocol for ER, PR and HER2 is also well documented [11, 34] and those laboratories that failed to achieve the expected result have been recommended to check to ensure that their assays are fully optimized, particularly with respect to choice of antibody and antigen retrieval protocol.

In Vietnam the testing and scoring of Ki-67 is not as established as for the other three markers. No doubt this also added to the increased variation in results. In this respect, whilst the authors of this paper are familiar with the latest IKWG guidelines, and similar initiatives to standardise Ki-67 scoring, they have found the 5% interval score that makes a global estimation, without counting specific cells, to be the most practical and convenient method by which to record Ki-67 expression [6, 35]. Indeed, there is some evidence to show that histopathologists can accurately discriminate between proportions that are only 5% different, albeit in equal sized images [36].

Recently Probert et al (2023) [7] investigated Ki-67 expression and breast cancer mortality in over 8,000 United Kingdom National Health Service (NHS) patients with ER positive and HER2 negative disease. This study found that there was very little difference in adjusted cumulative breast cancer mortality risk for women with 0%–5%, 6%–10%, 11%–19% and 20%–29% Ki67 positive tumors. In comparison, patients having tumors with higher Ki-67 scores of 30%–39% and 40%–100% had significantly greater mortality risk. Interestingly, standardization to guard against inter-laboratory variability had little impact on the results [7].

In our EQA ring study, whilst Ki-67 testing resulted in a greater level of variation in results than for the other three markers, employing the clinically relevant cut point of ≥30% for prognosis as used by Probert et al [7] resulted in an overall agreement of 84% for testing and scoring of the tumors. This is an important finding as whilst there is still room for considerable improvement in consistency of Ki-67 testing by Vietnamese laboratories, it does support the findings made by Probert et al (2023) [7] i.e., that interlaboratory variation may be less of a problem than previously thought if a simple cut point for prognosis is used and reinforces the potential use of Ki-67 on a more regular basis for clinical decision making. When using the cut point of ≥20% of Ki67 stained cells, as used to predict favorable response to abemaciclib combined with endocrine therapy for the adjuvant treatment of hormone positive, HER2 negative, node-positive, high-risk, early breast cancer [25], the level of overall agreement in our EQA ring study was lower, though still relatively high with an overall agreement of 81% for Ki-67 testing.

These are important findings for developing countries such as Vietnam, as IHC markers are considerably less expensive than multi-gene assays, such as Oncotype Dx [35]. In a population of over 90 million people, with over 24,000 new cases of breast cancer each year and rising [37], and with the financial burden of cancer care frequently falling on the shoulders of the patient [38], estimating Ki-67 by IHC would be a far more affordable way to determine prognosis and predict response to therapy than the use of expensive multi-gene assays. Going forward, it also emphasises the importance of EQA initiatives to help ensure the reliability of Ki-67 testing.

In summary, this is the first EQA ring study held within Vietnam for ER, PR, HER2 and Ki67 and sets the base line as to the current level of reproducibility for these assays within the country. The study reports on the level of both intra- and inter-laboratory reproducibility and utilised tissue samples prepared in all of the participating laboratories. There was a significant level of reproducibility for all four biomarkers, with ER testing giving the least variation and Ki67 the greatest variation in results. However, the reproducibility of testing cases for HER2-low expression was relatively poor. As there is evidence that participation in EQA improves the quality of results [11], we are confident that continued participation in the program will increase the level of reproducibility for these important biomarkers and help ensure their reliability for current and future clinical use. This work represents an advance in biomedical science because it provides evidence for the reproducibility of breast cancer biomarker testing in Vietnam.