Main

Oestrogen receptor (ER) and human epidermal growth factor receptor homologue 2 (HER2) are established biomarkers in invasive breast cancer and form the backbone of clinical decision-making related to targeted therapies in the adjuvant setting. Although data from external quality assurance schemes (such as UK NEQAS ICC), successful participation in which is mandatory for UK laboratories, indicates excellent performance for testing these receptors in local laboratories nationally, there is relatively little published evidence comparing local results to central re-testing of local ER and HER2 expression in large clinical trial data sets. In particular, information from central laboratory testing/validation of series of invasive breast carcinomas that have been designated as ER-positive and HER2-negative is limited; reports have largely described data from central re-testing of breast cancers, which have been recorded as HER2-positive in local laboratories such as in the Breast Intergroup Trial N9831 (Roche et al, 2002; Perez et al, 2006). Some of these early publications have indicated alarming proportions of discrepancy in defining HER2 positivity. There are fewer publications comparing central repeat testing of hormone receptors from clinical trial samples but Viale et al (2007) examined 6291 of 8010 tumours from women in BIG1–98 and found that central review confirmed 97% of tumours were hormone receptor-positive (defined as ER and/or PgR⩾10%). Using tissue microarrays of tumours in the Tamoxifen and Exemestane Adjuvant Multinational (TEAM) trial, of 4325 cases with sufficient material only 42 were ER-negative (0.99%), of these 28 were PgR positive and only 14 ER-negative/PgR-negative tumours were identified (0.3%; Bartlett et al, 2011b).

The accuracy of defining hormone receptor-positive and HER2-negative invasive breast cancer in local centres is clearly vital for patient management outside of the clinical trial setting, but also has significant resource and cost implications within randomised trials where ER and/or HER2 are critical components of eligibility. The question remains whether local biomarker results are sufficiently robust to allow trialists to avoid the costly re-analysis of biomarkers in central laboratories to confirm patient eligibility. To address this question we have examined data in the UK setting within OPTIMA prelim.

Materials and methods

The Optimal Personalised Treatment of early breast cancer usIng Multiparameter Analysis preliminary study (OPTIMA prelim) (ISRCTN42400492) was the feasibility phase of a randomised controlled trial designed to validate the use of multiparameter assay-directed chemotherapy decisions in the UK National Health Service (Bartlett et al, 2013, 4, 2016; Stein et al, 2016). Patients were aged ⩾40 years at entry with surgically treated ER-positive, HER2-negative primary invasive breast cancer, with 1–9 involved axillary nodes or, if node negative, a tumour of at least 30 mm in maximum dimension. Patients were randomised to standard care (chemotherapy followed by endocrine therapy) or an Oncotype DX test (Genomic Health Inc., Redwood City, CA, USA) was performed on the surgically resected tumour to assign patients either to standard care (if 'Recurrence Score' (RS) was >25), or to endocrine therapy alone (if RS was ⩽25). In this feasibility study, ER and HER2 were both re-assessed by a central laboratory (UCL Advanced Diagnostics) after registration into the trial to confirm eligibility prior to randomisation.

Oestrogen receptor was assessed centrally on whole tissue sections by immunohistochemistry (6F11; Leica Biosystems, Wetzlar, Germany) and an Allred score of 3 or more was regarded as positive, as per national guidance at that time (Harvey et al, 1999). If central ER results were discordant with the local report, and there was any doubt, the assay was repeated with a second antibody (EP1, Dako, Santa Clara, CA, USA). Human epidermal growth factor receptor homologue 2 was re-assessed centrally with dual-colour dual-hapten brightfield in situ hybridisation (DDISH) (Ventana Medical Systems, Tucson, AZ, USA) and, as per UK National Guidelines, a ratio of Her2 to chromosome 17 centromeric probe (CEP17) of 2.00–2.20 was considered to represent borderline/positive gene amplification, while a ratio of Her2 : CEP17 of >2.20 was regarded as Her2 gene amplification (Bartlett et al, 2011a). If DDISH proved unsuccessful, fluorescence in situ hybridization (FISH) was attempted using the HER2 PathVysion probe (HER2 PathVysion; Abbott Molecular, Des Plaines, IL, USA). Human epidermal growth factor receptor homologue 2 immunohistochemistry (4B5; Ventana Medical Systems) was applied in cases where no result was achievable by either HER2 ISH technique.

Results

Between October 2012 and August 2014, 442 patients were registered into OPTIMA prelim, but 11 patients were subsequently withdrawn prior to central testing. Thus a total of 431 patients, recruited from 35 sites, had their tumours tested centrally. Nineteen patients with 21 tumours showed discrepancies in receptor status between local and central laboratory results (4.4%; 95% confidence interval 2.5–6.3%). The remaining 412 patients (95.6%) with concordant results went on to be randomised into OPTIMA prelim.

Seven tumours in seven patients (1.6%) were found to be ER-negative on central re-testing (Table 1). Two of the seven were heterogeneous, with an uncommon admixture of ER-negative and ER-positive cells identified in the surgically excised tumour. Two appear to represent true errors in local laboratory tests; as local laboratory re-testing on the same sample found the tumours to indeed be ER-negative (personal communication). In one case, an interpretive difference remained between the local and central testing; the core and the excision specimen were both re-assessed locally as showing low-level ER expression (Allred score 3 in the core biopsy) by the local pathologist. Unfortunately, despite liaison with the laboratories it has not been possible to discover whether ER status has been re-assessed locally for other two discrepant tumours.

Table 1 Details of the seven patients (from 431 patients registered and tested centrally) with discrepant oestrogen receptor results
Full size table

In total 15 tumours in 13 patients (3.0%) from the 431 patients tested centrally were discrepant for HER2 results (Table 2). One patient had one tumour that was centrally categorised as ER-negative and also showed Her2 amplification (ratio of Her2 : CEP17=3.59). Seven others also showed Her2 amplification (ratio of Her2 : CEP17 ranged from 2.39 to 3.92). An additional patient had one tumour that was Her2 amplified and one tumour that was borderline amplified (ratio of Her2 : CEP17=2.78 and 2.11, respectively). The remaining four patients had tumours showing borderline Her2 gene amplification (ratio between 2.00 and 2.20); including one patient with two tumours both showing borderline amplification. Only 3 of the 15 tumours demonstrated what some consider ‘high-level’ gene amplification (ratio >3.00) (Starczynski et al, 2012) and none what others have described as ‘high-grade’ amplification (ratio ⩾4.00; Seol et al, 2012).

Table 2 Details of the 15 discrepant tumours (13 patients) for HER2 status
Full size table

It has not been possible to ascertain if there has been repeat HER2 testing (immunohistochemistry, or FISH or DDISH) on all of these 13 cases; for five women (with seven tumours) the local team have managed the patient as per the central, HER2-positive, results without apparent re-testing. In three further cases data have not been obtainable. In four cases local re-testing has been undertaken: in two cases (one by FISH, other method uncertain) the local laboratory results have remained HER2-negative (both tumours borderline amplified by Her2 : CEP17 ratio centrally), that is, results remaining discrepant; one case was agreed to be HER2-positive by re-testing locally by FISH; the final case on local repeat testing had a Her2 : CEP17 ratio of >2.00 but the local pathologist maintained that the tumour should be regarded as HER2-negative because of low average Her2 copy number. One case was negative immunohistochemically but showed Her2 gene amplification (3.92).

Discussion

Central re-testing of HER2-positive breast cancers has shown high levels of variability in some clinical trials; for example, HER2 positivity was only confirmed in 85.8% of 2535 patients in the North Central Cancer Treatment Group N9831 intergroup adjuvant trial (Perez et al, 2006). Some of these trials, however, pre-date stringent guidelines for HER2 assessment and reporting, and the reasons for discordance are often not clear. The value of central re-testing of breast cancers defined locally as HER2-negative as an eligibility criterion for other, more recent, clinical trials has not been well studied. Outwith clinical trials generally lower degrees of discrepancy have been reported (Vani et al, 2008; Kaufman et al, 2014), for example, Kaufman et al (2014) identified that only 4% of 552 patients with metastatic HER2-negative carcinoma (defined locally) in a large observational cohort were HER2-positive on central re-testing. These data are essentially similar to the results in our UK clinical trial where 3.5% of tumours defined locally as HER2-negative were HER2-positive on central re-analysis.

These data highlight that ∼3% of patients could be being excluded from HER2-directed therapies due to a potentially faulty local result in real-world testing in the United Kingdom. However, of note, we report here the proportion of cases that are discordant between local and central laboratory testing. Although for two cases, repeat re-testing of the same samples locally confirmed the tumour was ER-negative (rather than ER-positive as initially reported), for others it is only possible to record that the other results were ‘discordant’. It is not per se the case that the central laboratory is correct and the local laboratory inaccurate, as both adhere to the same quality assurance and reporting guidelines.

Central repeat testing of hormone receptor status from clinical trial samples have reported similar, albeit slightly higher, levels of difference between local and central laboratories than we have found. Viale et al (2007) examined 6291 of 8010 tumours from women in BIG1–98 and found that central review confirmed 97% of tumours were hormone receptor-positive, although this incorporated both ER and progesterone receptor, and with different cut-offs than applied as routine in the United Kingdom (i.e., defined as ER and/or PgR⩾10%). Indeed, the authors note that, of 105 carcinomas that were reported locally as ER-negative, 73 had >10%, and 8 had 1–9% positive cells. This highlights the difficulty of non-standard definitions globally for hormone receptor positivity and the need for pathologists, as well as all other members of the multidisciplinary team, to be aware of study protocols and definitions.

Notwithstanding that these results compare favourably to the (albeit limited) published data, there are a number of possible explanations for discrepant results between local and central laboratories. Additional challenges include variation in methodology (e.g., immunohistochemistry vs FISH vs chromogenic in situ hybridisation (DDISH) for assessment of HER2 status), as well as differences in the antibody clones used, variation in the material assessed (cores vs surgical excision specimens) and pathologist interpretation. It is well recognised that variation between core biopsy specimens and surgical excision is uncommon (<2% of cases showing heterogeneity; Arnedos et al, 2009; Lee et al, 2012), although this clearly does occur and may potentially explain variations in receptor status if different specimens are submitted for central testing than examined locally. Indeed, this variation appears to explain at least two of the seven cases with discrepant ER status in this study.

These potential discrepancies are all applicable even if the central review is undertaken in ‘real time’, that is, prior to patient randomisation, as in OPTIMA prelim. Nevertheless, particular care must be taken when analysing historical data on ER status, even in meta-analysis of clinical trials, or when comparing to present day results; data extracted from local reports may be based on entirely different methodologies; Collins et al (2008) examined (on TMA) 1851 cases where tissue and histology reports were available and highlighted that in 82% of the cases the original assays were biochemical. Even where immunohistochemistry was applied both locally and centrally as the technique of choice, agreement was only 92% for ER status (310 of 336 specimens; Collins et al, 2008). Again, the 1.6% difference seen in OPTIMA prelim compares favourably.

Despite all the potential technical and interpretive differences in biomarker analysis, the results from OPTIMA prelim indicate good concordance between local laboratories and a central re-testing centre in the United Kingdom in classification of invasive breast cancers as ER-positive and HER2-negative. Such re-testing in large randomised clinical trials recruiting thousands of patients is very expensive and, in the setting of this group of patients (as opposed to HER2-positive disease, for example, where discrepancies may be higher), the value is questionable.