Accurate assessment of HER2 status is essential for identifying patients who are candidates for therapy with the anti-HER2 monoclonal antibody trastuzumab. A rapid and accurate test for HER2 status is needed because an increased number of treatment decisions are based on the HER2 status of a patient's disease. In particular, increasing evidence suggests that response to certain forms of hormone treatment and chemotherapy are dependent on a patient's HER2 status [9
]. The results of our multicentre ring study show that CISH is a reliable test for the assessment of HER2
gene copy number.
The study demonstrates excellent correlation between FISH and CISH results, with 92% of cases scored as having high-level amplification by FISH also scoring as having high-level amplification by CISH. In addition, 94% of cases which were shown to have a normal HER2 gene copy number by FISH appeared to lack HER2 gene amplification by CISH. Inter-laboratory concordance with CISH was similar to that seen between FISH and outside CISH; 94% for ≤5 HER2 copies per cell and 92% for ≥6 HER2 copies per cell.
Intra-laboratory correlation of IHC with CISH was also good for cases scored as 3+ and 0 or 1+ by IHC (91% and 92% of cases respectively). Twenty (27%) of the 75 cases scored as HER2 2+ by IHC showed HER2
gene amplification as assayed by CISH. This is in agreement with the findings of other studies comparing HER2 overexpression with the degree of HER2
gene amplification. In these studies, the percentage of tumours with a HER2 IHC 2+ score and gene amplification was approximately 25% [12
Accurate scoring of cases with low-level HER2
gene amplification is technically challenging. Therefore, a high number of cases with a HER2/CEP17 ratio of 2.0–4.0 were specifically included for evaluation in this study. According to the FISH scoring guidelines, these cases are positive for HER2
gene amplification. However, 43% of cases scored as low level of amplification by FISH, but scored as non-amplified by CISH. Further detailed analyses should be conducted in this area to clarify this result. However, it should be noted that breast carcinomas with a low degree of HER2
gene amplification are rare (estimated to be 1–3% of all carcinomas), and inter-observer variability is greatest in this category of samples. Indeed, three out of seven cases with low-level amplification assessed by FISH included in this study gave different results when re-assessed by FISH in another laboratory (data not shown). A recent inter-observer study for FISH showed that, although agreement was excellent for tumours with normal HER
2 gene copy numbers or HER
2 gene amplification, there was marked inter-observer variability in these 'borderline cases' [27
]. Recently, a consensus panel has proposed adapted scoring guidelines for HER2 testing. An important recommendation from this panel was to consider reporting breast cancer cases with a HER2/centromer chromosome 17 ratio between 1.8 and 2.2 as borderline [28
]. Using this adapted scoring guideline, a tumour is assessed as HER2
amplified when the ratio is more than 2.2; or when the absolute number of HER2
gene copies is more than six.
The key to understanding FISH/CISH discrepancy in cases with low-level amplification lies in the nature of the two tests. A significant difference between FISH and CISH is that in the most commonly used FISH assay, 2-colour FISH, the copy number of the HER2 gene and of the centromere of chromosome 17 (CEP17) are assessed simultaneously. The signal from the centromere probe functions as an internal control; the final FISH score is based on a ratio of the signals from the two probes. Conversely, in the CISH assay, the copy number of HER2 is assessed directly. When classifying tumours with one to four copies, or with >10 copies of HER2 by CISH, the HER2 status of the sample is clear. However, in cases with HER2 copy numbers in the range of 4–10, the HER2/CEP17 ratio may be of importance. Some discrepancies between FISH and CISH may also be explained for polysomic cases where the FISH assay might indicate a negative testing result, as the HER2/CEP17 ratio is calculated, while CISH indicates a HER2-positive testing result as the numbers of HER2 signals are assessed. In general, we have also noted that the inter-observer variability in scoring cases with mid-range HER2 copy numbers is relatively large, and also that inter-observer variability in scoring HER2/CEP17 ratio with FISH is considerable (unpublished results). According to the Invitrogen scoring system for CISH, supplied by the manufacturer at the start of this study, a score of ≤5 HER2 copies per cell indicates no amplification, whereas a score of ≥6 HER2 copies indicates amplification. This meant that classification of cases with an average CISH score of 5–6 was unclear. However, since this study, Invitrogen has updated its scoring system to specify that a score of >5 signals per cell should now be considered indicative of HER2 gene amplification. As we wished to be able to categorize the subgroups in this borderline category more precisely, scoring 5 copies and 6 copies as separate categories was part of the study design. We believe that obtaining a reliable CISH result for these borderline cases is no more problematic than obtaining a reproduced FISH result for such cases. A practical approach to this problem is to count at least 60 nuclei when the number of HER2 copies ranges from 4–10, and to consider adding CISH using a CEP17 probe on a consecutive slide to confirm the result. Alternatively, and where possible, dual-colour FISH with probes for HER2 and CEP17 can be considered. It is to be hoped that it will be possible to analyze the studies showing benefit of trastuzumab therapy in the adjuvant setting for the benefit of the subgroup of HER2 low-level/borderline amplified tumours. This may help to better define the analytical approach to this category of tumours.
When own CISH was compared with FISH in determining the HER2 status of IHC 2+ cases, there was 100% concordance for HER2-positive samples and 94% for HER2-negative samples. The good intra-laboratory correlation of FISH with CISH suggests that CISH could potentially be used in place of FISH to determine the HER2 status of equivocal cases by IHC. As both tests produce similar results in the retesting of equivocal IHC cases, CISH could potentially fulfil the same role as FISH in the HER2 testing algorithm.
CISH displayed high inter-laboratory concordance, showing that results are reproduced between laboratories. Correlation between own CISH and outside CISH was similar to that observed between FISH and outside CISH with 95% concordance for cases with <5 HER2 copies per cell and 92% for cases with >6 HER2 copies per cell. The most notable discrepancies with CISH in a few cases are probably the result of technical problems with the staining procedure or with the interpretation of the results. This highlights the need for training before any technology is used for the first time. However, the excellent inter-laboratory correlation of CISH results highlights its potential as a method for testing HER2-status, even in those laboratories unfamiliar with using CISH.
As can be seen from our results, this study includes cases with a HER2 IHC3+ score but without HER2 gene amplification (as detected by CISH); and cases with a HER2 IHC0/1+ score and with HER2 gene amplification (as detected by CISH). The number of discrepancies was very similar if FISH was used instead of CISH (data not shown). In view of the clinical consequences of accurate HER2 status assessment for many patients, it should be considered to test CISH (or FISH) alongside IHC either as a diagnostic routine or as part of a quality assurance program in each lab.
The pre-treatment of tissue sections, especially the pepsin digestion time, was a critical step in achieving a good CISH result. Like FISH, the optimal pepsin digestion time differs between tumours. For practical reasons, the pepsin digestion time used in the preparation of received slides was calculated in each laboratory on the basis of a pepsin time course performed on one representative tumour provided by the sending laboratory. However, it is possible that this was not the optimal value for all tumour samples, and as a result it was likely that in some cases no HER2 signal would be detected. For such cases, it was determined that the CISH test should be repeated after adjustment of the pepsin digestion time. Using this strategy, a good result was obtained in 99% of the cases (208 out of 211). In all cases, the percentage of tumour cells for which HER2 copy number could be assessed by CISH was recorded; this percentage was >50% for most cases (data not shown). There were no cases in this series where noticeable heterogeneity in the number of HER2 signals per tumour cell was observed.
An advantage of CISH over FISH is that invasive tumour cells can be easily identified using a light microscope, provided that the morphology of the tissue is good following CISH staining. More than 80% of the tumours examined to date in this study had either good or excellent morphology, illustrating this beneficial aspect of CISH testing.