The HER2 oncogene encodes a transmembrane tyrosine kinase (human EGF-like receptor 2) located on chromosome 17q21 12. HER2 protein belongs to the epidermal growth factor family 3 and is important for cell differentiation, adhesion and motility 4. HER2 gene amplification and protein overexpression exists in about 20% of breast cancers 56 and is linked to a poor prognosis 7. From a clinical point of view, HER2 receptor has become important as a target for antibody-based therapy with trastuzumab (Herceptin^®) 8. Trastuzumab was originally approved for the treatment of metastatic HER2-positive breast cancers because of a significant survival benefit when given in combination with paclitaxel 91011. More recently, adjuvant treatment of primary, HER2-positive breast cancers with trastuzumab has been shown to improve patient outcome markedly 12. Thus, determination of HER2 status in every breast cancer patient to select for adjuvant treatment with trastuzumab is becoming a standard worldwide. There has been much debate on how to test HER2 status. Other than fluorescence in situ hybridisation (FISH), immunohistochemistry is liable to technical and interpretational variability 131415. Accordingly, it has been shown that response to trastuzumab is strongly associated with HER2 gene status irrespective of protein expression determined by immunohistochemistry 1617. Importantly, no standardised immunocytochemical assay is available for cytological specimens. For these reasons, testing the HER2 gene status by FISH is now widely regarded as the gold standard 18.

HER2 status is commonly determined in the primary tumour, because biopsies from metastasised lesions are not always available. In three previous FISH studies on 12 to 68 patients, the prevalence of a discrepant result between the primary breast cancer and the distant metastasis ranged from 0% to 22.2% 1920212223.

Because of the controversial data and the therapeutic importance of HER2 testing, we compared the HER2 status in a large series of primary breast cancers and their matched distant metastases by FISH. In addition, we attempted to elucidate reasons for these discrepancies.

Our results demonstrate that discrepancies in HER2 gene status between primary breast cancers and matched metastases do occur and may be related to technical and interpretational difficulties.

The HER2 status of the primary breast cancers and the matched metastases was concordant in 92.4% of the 105 patients at the time of the initial evaluation (κ = 0.76; 95% confidence interval (CI) 0.61 to 0.92). This substantial concordance was increased to 97.1% after re-evaluation of all discordant cases (κ = 0.85; 95% CI = 0.73 to 0.98).

HER2 amplification was found in 22 (21%) of the 105 primary tumours and in 21 (20%) of the matched distant metastases. The HER2 status between primary tumours and distant metastases differed in eight (7.6%) of the patients at the time of the initial evaluation.

The primary tumours of all eight discrepant cases were reanalysed on routine tissue sections, and the hybridised cytological specimens were carefully scored again. On the routine tissue sections we calculated an exact ratio from at least 20 scored cells. The clinicopathological characteristics and FISH findings of these patients are summarised in Table 3.

Re-evaluation showed that in one patient with HER2 amplification of the primary tumour, rare tumour cells with HER2 amplification in the pleural effusion had been overlooked as a result of rarity of malignant cells and a highly predominating background of reactive mesothelial cells and macrophages (Figure 1a,b). In another patient, tumour cells in the pleural effusion but not on the TMA specimen of the primary tumour showed HER2 amplification. However, FISH on the routine tissue section of the primary tumour revealed a heterogeneous HER2 status with only a few amplified tumour cells (Figure 1c,d).

In another patient, metastasis showed low-level gain but no amplification of the HER2 gene (HER2/CEP17 ratio 1.25, with 3 to 10/4 to 6 signals), whereas the primary tumour had initially been scored as amplified on the TMA spot. Routine section FISH analysis of the primary tumour also revealed low-level gain with a HER2/CEP17 ratio of 1.6 (with 1 to 2/2 to 6 signals). A further two patients had a 'borderline' amplification. One patient had an average HER2/CEP17 ratio of 2.1 (with 3 to 9/2 to 4 signals) in the primary tumour but a ratio of only 1.38 in the metastasised cells of the peritoneal effusion (with 3 to 23/3 to 15 signals). The other patient had a HER2/CEP17 ratio of 2.1, with 2 to 8/2 signals after repeated counting of 20 tumour cells, whereas the metastatic cells had a ratio of 1.69 (5 to 10/3 to 6 signals).

Three patients (2.9%) remained with an unexplained discrepancy even after re-evaluation of the cytological specimens and repeated FISH analysis of the routine tissue sections of the primary tumour (Figure 1e,f). Two of the patients had a primary tumour with a normal copy number of HER2 and chromosome 17 but with HER2 amplification in the metastasis. The other patient had high-level amplification with a ratio of more than 3 in the primary tumour (10 to 20/2 to 4), in contrast with a ratio of 1.83 (3 to 6/2 to 3) in the metastasis.

HER2 amplification was significantly associated with nuclear grade but not with pT and pN category (Table 4). Time from resection of the primary tumour to the diagnosis of metastases was not significantly associated with the pT category (p = 0.248), the pN category (p = 0.1383) or the nuclear grade (p = 0.3777; detailed data not shown). Metastases tended to occur earlier in tumours with HER2 amplification than in those without (44.1 ± SEM 7.8 months versus 71.8.6 ± SEM 7.3 months; p = 0.0639).

The whole metastatic cascade from primary tumour to axillary lymph node metastasis and distant metastasis was analysed in 31 patients (29.5%). One patient with discordant HER2 status between the primary tumour (HER2/CEP17 ratio 2.1) and the distant metastasis (HER2/CEP17 ratio 1.38) had a 'borderline result' in the lymph node (HER2/CEP17 ratio 1.9). There was complete concordance of the FISH results in the remaining 30 cases, 4 of which showed amplification.

HER2 amplification identifies patients who are likely to respond to therapy with trastuzumab, a humanised antibody directed against the HER2 protein 2728. Approval of trastuzumab was originally restricted to patients with HER2-positive metastatic disease. On the basis of the positive results of recent international trials 12, approval is now being extended internationally to adjuvant treatment of HER2-positive breast cancers. In the mean time, HER2 analysis of all newly diagnosed breast cancers has already become a standard in many institutions. Because of the high costs of trastuzumab and the risk of cardiotoxicity in some patients 29, it is crucial to make a precise selection of patients for treatment to guarantee optimal clinical benefit while retaining cost effectiveness.

In patients with metastatic disease, selection for therapy with trastuzumab has traditionally been based on the HER2 status of the primary tumour. The reported prevalence of discordance of HER2 status between primary tumour and metastasis ranges from 0 to 22.2% when assessment with both immunohistochemistry and FISH is considered 20212330. Part of these discrepancies is likely to be due to the well-known technical and interpretational limitations of immunohistochemical HER2 assessment 131531. Because FISH is considered the gold standard for HER2 testing 3233, we investigated the discordance rate based solely on FISH in a large consecutive series of metastatic breast cancers. We found a discrepancy in 8 (7.6%) of the 105 patients. This is in the range of previous studies with FISH and confirms considerable stability of HER2 gene status even in tumours that developed distant metastases more than 21 years after initial surgery (Table 5) 1921. The discrepancies included three tumours with positive primaries but negative metastases, as determined by FISH, and five positive metastases but negative primaries. There would therefore be a risk of both undertreatment and overtreatment of these metastasised breast cancers if the treatment decision were based only on the HER2 status of these primary tumours.

The reasons for discrepancy in HER2 FISH status between primary breast cancer and metastases have not been investigated in previous studies. Here, detailed re-evaluation of the HER2 FISH status by scoring the specimens again or by hybridising routine tissue sections allowed us to discover reasons for HER2 discrepancies. Not all discrepant results represented true biological conversion. Instead, we uncovered interpretational difficulties as a reason in five (4.7%) of the patients. One had a slight gain of HER2 signals that was regarded as amplified in the first evaluation. In two patients, the discrepancy was explained by a HER2/reference ratio that was slightly lower or higher than the threshold of 2, which we refer to as 'borderline'. It is conceivable that such a borderline ratio is more prone to inter-observer variation than amplification with a high ratio or dense gene clusters. In a recent inter-laboratory survey, there was a considerable variability in interpretation of cases with low-level or borderline amplification 34. In contrast, those authors found excellent reproducibility in HER2 FISH analysis for tumours with no amplification or high amplification of the HER2 gene. This highlights the need for consensus on the use of an equivocal/borderline interpretative category. Accordingly, the package insert of the PathVysion includes the statement 'a ratio at or near the cutoff (1.8 to 2.2) should be interpreted with caution'. The clinical relevance of this type of interpretational limitation is not clear, because the impact of the ratio level on the likelihood of therapy response is still unknown. One could hypothesise that the response rate is low in breast cancers with a borderline FISH result.

Rarity of HER2 amplified cells was responsible for a discrepancy in a further two patients. In one of these, rare HER2-amplified cells were initially overlooked on a cytological smear from metastasis with a high background of reactive cells. In the other patient, rare amplified cells were not detected initially because of intratumoral heterogeneity of HER2 status within the primary tumour. In this patient, we must assume the presence of clonal selection of the rare HER2-amplified cells in the primary for distant metastatic spread. Our finding also emphasises the importance of careful and thorough evaluation of the hybridised specimens, because heterogeneity with small cancer foci prevails in rare cases 3536.

In three of the eight patients (patients 1, 2, and 8 in Table 3) we could not identify any interpretational reason for the discrepancy. Two of these were negative, as determined by FISH, in the primary but positive in the metastasis. In the third tumour, high-level amplification in the primary contrasted with low-level gain in the metastasis. It is possible that these three tumours represented true conversion of HER2 status by clonal selection or genetic drift during metastatic progression. However, we cannot exclude the possibility of undetected heterogeneity even in these cases, because only a small percentage of the entire tumour volume is represented on a histological section or on a cytological smear 3738.

The patients of this study were selected on the basis of the availability of cytologically diagnosed distant metastases. Fine-needle aspiration cytology of solid lesions or exfoliative cytology (for example malignant effusions) is a commonly used method to diagnose or confirm metastatic disease. Our results confirm that cytological specimens are well suited to HER2 FISH analysis 223940. It has previously been shown that results of HER2 FISH analyses from cytological specimens are highly concordant with matched histological sections 4142. It is therefore unlikely that our results were affected by the different types of tumour material of primary tumours and metastases.

The standard morphological parameters (pT category, pN category and grade) are strong prognostic factors in newly diagnosed breast cancer 43. Interestingly, none of these parameters was significantly associated with the time interval from initial treatment to cytological diagnosis of distant metastasis in our selected series of patients. Because the time of cytological sampling of metastatic cells is not always identical with the time of first clinical detection of metastasis, prognostic data must be interpreted with caution in our study. Nevertheless, our data indicate that within the group of breast cancers that are capable of distant metastasis, the dynamics of metastasis is driven by biological or environmental factors that are not fully reflected by morphological features. The observed tendency of HER2 amplification towards early metastasis is concordant with the known adverse prognostic role of HER2 amplification 44.

HER2 FISH status is highly preserved as breast cancers progress to metastatic disease. However, a discrepancy in HER2 status exists in a small fraction of patients. As well as true conversion of HER2 status, interpretational difficulties due to borderline FISH results, rarity of tumour cells and intratumoral heterogeneity were identified as important reasons for the discrepancy. Irrespective of the reason, however, every discordance in HER2 status is a diagnostic reality and can lead to inaccurate treatment decisions with medical and economical effects. To guarantee optimal care of individual patients, we advocate HER2 analysis of distant metastases in all patients irrespective of the result in the primary tumour. The feasibility of this approach will ultimately be determined by economic factors. Future patients with distant metastases after adjuvant treatment for HER2-positive breast cancer will pose additional diagnostic and therapeutic challenges. It is possible that treatment with trastuzumab leads to clonal selection of HER2-negative tumour cells, as observed in individual patients 35. Thus, HER2 testing of metastasis may become a necessity in every patient.

CEP17 = centromere probe of chromosome 17; CI = confidence interval; DAPI = 4',6-diamidino-2-phenylindole; FISH = fluorescence in situ hybridisation; HER2 = gene encoding human EGF-like receptor 2; SSC = saline sodium citrate; TMA = tissue microarray.

LB has received speaker's honorarium and refunding of travel costs, and research funding from Abbott Molecular Inc. CT has received refunding of travel costs by Abbott Molecular Inc. All other authors declare that they do not have competing interests.

CT contributed to the study design, was responsible for statistical evaluation and drafted the manuscript. SS participated in study design and FISH evaluation. UW, RS and GC selected and provided matched specimens. HN performed and scored FISH on the histological specimens. BG, MH and BDV performed and scored FISH on the cytological specimens. IZ performed statistical analyses. LT was responsible for and supervised FISH on the histological specimens. GF was involved in evaluation of the cytological specimens. LB conceived the study, supervised the experiments, revised the manuscript critically for important intellectual content and gave final approval of the version to be published. All authors read and approved the final manuscript.