Fluorescence in situ hybridization (FISH) and brightfield in situ hybridization (ISH) are 2 clinically approved laboratory methods for detecting ERBB2 (HER2) amplification in breast cancer.
To compare the performance of FISH and brightfield ISH on proficiency testing administered by the College of American Pathologists Laboratory Accreditation Program.
Retrospective review was performed on 70 tissue core samples in 7 separate proficiency testing surveys conducted between 2009 and 2013.
The samples included 13 consensus-amplified tissue cores, 53 consensus-nonamplified cores, and 4 cores that did not reach consensus for FISH and/or brightfield ISH. There were 2552 individual responses for FISH and 1871 individual responses for brightfield ISH. Consensus response rates were comparable for FISH (2474 of 2524; 98.0%) and brightfield ISH (2135 of 2189; 97.5%). The FISH analysis yielded an average HER2 copy number per cell that was significantly higher (by 2.86; P = .02) compared with brightfield ISH for amplified cores. For nonamplified cores, FISH yielded slightly, but not significantly, higher (by 0.17; P = .10) HER2 copy numbers per cell. There was no significant difference in the average HER2 to control ratio for either consensus-amplified or consensus-nonamplified cores. Participants reported “unable to analyze” more frequently for brightfield ISH (244 of 2453; 9.9%) than they did for FISH (160 of 2684; 6.0%).
Our study indicates a high concordance rate in proficiency testing surveys, with some significant differences noted in the technical performance of these assays. In borderline cases, updated American Society of Clinical Oncology/College of American Pathologists cutoff thresholds that place greater emphasis on HER2 copy number per cell could accentuate those differences between FISH and brightfield ISH.
Amplification or overexpression of the human epidermal growth factor receptor 2 (ERBB2, HER2) gene is an important oncogenic driver in up to 20% to 25% of breast cancers. Breast cancers with HER2 amplification or HER2 overexpression tend to have a higher histologic grade, and before the use of HER2-targeted therapy, HER2+ cancers had a poor prognosis.1 The first HER2-targeted therapy, trastuzumab (Genentech, South San Francisco, California), was introduced in 1998, and early clinical trials demonstrated the efficacy of trastuzumab for treating HER2+ breast cancers.2–9 Most HER2+ breast cancers demonstrate overexpression of the HER2 tyrosine kinase receptor protein on the cell surface; that overexpression is driven at the DNA level by amplification of the HER2 gene.10 HER2 overexpression by immunohistochemistry was found to be closely correlated with a high copy number (amplification, usually >10 copies of HER2 per cell) in several fluorescence in situ hybridization (FISH) studies in the 1990s.11,12 Early clinical trials from the late 1990s, which led to the US Food and Drug Administration (FDA) approval of trastuzumab in 2006, used immunohistochemistry and FISH testing to determine eligibility for targeted therapy.
In an effort to provide guideline recommendations for when and how to test for overexpression and/or amplification of HER2, a joint committee of the College of American Pathologists (CAP) and the American Society of Clinical Oncology (ASCO) convened and published a guideline for HER2 testing in breast cancer in 2007.13 New test methods have emerged since the 2007 guideline was released. One of those new methods is brightfield in situ hybridization (ISH), and several FDA-approved brightfield ISH assays are now available. In the 2013 update to the original ASCO/CAP guideline, the joint committee approved brightfield ISH for evaluating HER2 status in breast.14 Moreover, FISH and brightfield ISH are interpreted using the same criteria, and several studies have indicated similar performance with these methods.15–29 However, some studies suggest that brightfield ISH might be less sensitive for detecting low-level amplification and might have a lower concordance than FISH in challenging samples.15,19,30,31
Since 2007, the CAP Laboratory Accreditation Program has required that laboratories performing HER2 testing on breast cancer participate in external proficiency testing, according to ASCO/CAP guidelines. The Cytogenetics Resource Committee, comprising members of CAP and the American College of Medical Genetics and Genomics, has administered proficiency testing surveys for HER2 FISH on formalin-fixed, paraffin-embedded breast cancer tissue since 2000.32 The CAP Molecular Oncology Committee has offered HER2 proficiency testing for brightfield ISH since 2009. We compared the performance of FISH and brightfield ISH on samples that were shared between these proficiency testing surveys. The samples used for proficiency testing during the study period were selected to represent routine cases that had unambiguously negative or positive results for HER2 amplification.
MATERIALS AND METHODS
This study includes data from CAP HER2 proficiency testing surveys from 2009 to 2013 for FISH (survey designation, CYH) and brightfield ISH (survey designation, ISH2). The CAP offered the CYH and ISH2 surveys biannually in the form of 2 tissue-microarray slides; each slide had 5 cores of formalin-fixed, paraffin-embedded breast cancer tissue and 1 negative control (nonneoplastic liver tissue). The CAP randomly assigned tissues from 1 of 4 material groups (A–D) to CYH survey participants, and the CAP assigned tissue from material group D to all ISH2 survey participants. Of 9 CYH and 9 ISH2 surveys from 2009 to 2013 conducted before the release of the 2013 updated ASCO/CAP guideline, 7 (78%) shared material between CYH (Material Group D) and ISH2 and were compared. Survey participants provided the mean number of HER2 signals per nucleus, the ratio of HER2 to the chromosome 17 control (centromere) signals (if applicable), and an interpretation of not amplified, equivocal, or amplified according to the 2007 ASCO/CAP guideline. Consensus for each tissue core required a minimum 80% participant agreement on the interpretation of nonamplified, equivocal, or amplified. This study focused on survey performance for FDA-approved probes, excluding results from 2 participating laboratories that used laboratory-developed probe sets.
Stepwise logistic-regression models were used to analyze survey performance according to the consensus interpretation. The initial model included a length-of-participation factor, but because that factor was not significant, it was excluded from the final stepwise logistic-regression model. The model was fit with 4 factors: (1) method (FISH or brightfield ISH), (2) annual test volume (5 levels), (3) amplification status (amplified or not amplified), and (4) probe (single or dual color). The model also included a strata specification to compare the specimen/core results between CYH and ISH2. In addition, the CYH and ISH2 average HER2 copies/cell and the average HER2 to control ratios were tested for systematic positive or negative differences with the sign test. A significance level of α = 0.05 was used for those analyses. Statistical analyses were performed using SAS 9.3 software (SAS Institute, Cary, North Carolina).
RESULTS
The analysis included a total of 70 breast cancer tissue samples from 14 surveys (CYH [material group D] and ISH2 09B, 10B, 11A, 11B, 12A, 12B, 13A). There were 13 consensus-amplified tissue samples, 53 consensus-nonamplified tissue samples, and 4 tissue samples that did not reach consensus for CYH and/or ISH2 surveys. One tissue sample did not reach consensus for CYH, 2 did not reach consensus for ISH2, and 1 did not reach consensus for either survey. The cumulative results of the study period included 2552 individual responses to FISH (CYH) challenges and 1871 individual responses to brightfield ISH (ISH2) challenges. The difference in the number of responses for FISH versus brightfield ISH was primarily due to differences in participant enrollment. During the 5-year period, 245 to 300 laboratories participated in the CYH surveys, 32 to 56 of which received material group D in any given survey. Participation in the ISH2 survey grew from 20 to 63 laboratories during that 5-year period (Figure 1).
Number of laboratories participating in College of American Pathologists (CAP) surveys for HER2 fluorescence in situ hybridization (FISH) and brightfield in situ hybridization (ISH) for surveys included in the study. The tissue cores included in the study were from material group D and were shared between FISH and brightfield ISH surveys.
Number of laboratories participating in College of American Pathologists (CAP) surveys for HER2 fluorescence in situ hybridization (FISH) and brightfield in situ hybridization (ISH) for surveys included in the study. The tissue cores included in the study were from material group D and were shared between FISH and brightfield ISH surveys.
The overall rate of consensus interpretations was comparable for FISH and brightfield ISH (Table 1). Likewise, there was no significant difference in consensus rates based on dual-color versus single-color probe strategy. The duration of participation in CAP surveys was included in the logistic-regression models, and there were no statistically significant performance differences based on that factor.
Although the overall performance of FISH and brightfield ISH was similar, FISH tended to yield higher average HER2 copies per cell for consensus-amplified cores; on average, the HER2 copy number per cell was higher for FISH by 2.86 (P = .02). For nonamplified cores, the average HER2 copy number was slightly, but not significantly, greater for FISH than brightfield ISH; on average, the HER2 copy number per cell was higher for FISH by 0.17 (P = .10) (Figure 2; Table 2). For the average HER2 to control ratios, there was no significant difference between FISH and brightfield ISH for either amplified or nonamplified cores (Figure 3; Table 2). Only 4 cores failed to reach consensus; in those cases, FISH assays showed a higher average number of HER2 signals per cell and 3 of 4 (75%) showed a higher average HER2 to control ratio than did brightfield ISH on the same sample.
Bias chart demonstrating the difference between the mean participant value for HER2 copies per cell for fluorescence in situ hybridization (FISH) and the mean participant value for HER2 copies per cell for brightfield in situ hybridization (ISH) on 70 tissue cores as a function of the overall consensus number of HER2 copies per cell.
Bias chart demonstrating the difference between the mean participant value for HER2 copies per cell for fluorescence in situ hybridization (FISH) and the mean participant value for HER2 copies per cell for brightfield in situ hybridization (ISH) on 70 tissue cores as a function of the overall consensus number of HER2 copies per cell.
Bias chart demonstrating the difference between the mean participant HER2 to control ratio value for fluorescence in situ hybridization (FISH) and the mean participant HER2 to control ratio value for brightfield in situ hybridization (ISH) on 70 tissue cores as a function of the overall consensus HER2 to control ratio value.
Bias chart demonstrating the difference between the mean participant HER2 to control ratio value for fluorescence in situ hybridization (FISH) and the mean participant HER2 to control ratio value for brightfield in situ hybridization (ISH) on 70 tissue cores as a function of the overall consensus HER2 to control ratio value.
Participant responses of “unable to analyze” were more frequent for brightfield ISH (244 of 2453; 9.9%) than they were for FISH (160 of 2684; 6.0%). Data obtained from participant results indicated that an “absent or weak probe signal” was the most common reason specimens examined by either method could not be analyzed. The percentage of unsatisfactory samples affected by that issue was greater for brightfield ISH (124 of 239; 51.9%) versus FISH (62 of 159; 39.0%) for those participants who provided a reason why a sample could not be analyzed.
The proficiency testing summarized here was performed from 2009 to 2013; during which time, the 2007 ASCO/CAP guideline was in effect. To compare the performance of FISH and brightfield ISH according to the 2013 ASCO/CAP guideline, we performed a scoring simulation based on the participant data for the HER2 to control ratio and on average HER2 copies per cell for all tissue samples, including both consensus and nonconsensus samples (Table 3). In cases in which the classification changed, most changes were from not amplified to either equivocal or amplified, or from equivocal to amplified, reflecting the changes in the cutoff values instituted in the 2013 ASCO/CAP guideline update.
DISCUSSION
This study evaluated CAP proficiency testing data generated by more than 300 accredited laboratories participating in FISH and brightfield ISH surveys for HER2 in breast cancer. The breast cancer tissue samples used for proficiency testing were selected to represent routine, nonamplified or amplified cases typically encountered in clinical laboratory practice.
We noted comparable performance for both methods in the interpretation of those nonamplified and amplified specimens that reached consensus. However, some differences between methods were apparent. The FISH analyses showed a statistically significant greater number of HER2 copy numbers per cell in consensus-amplified cores. For consensus-nonamplified cases, there was a trend toward greater HER2 copy numbers per cell for FISH, which did not reach statistical significance. For the few samples that failed to reach 80% consensus, the mean HER2 copy number per cell and HER2 to control ratio by FISH trended higher compared with brightfield ISH. There was no significant difference in the average HER2 to control ratios for consensus responses for either amplified or nonamplified cores. Similar findings have been demonstrated previously in a cohort of challenging cases.30
When we analyzed the participant data according to the updated guideline, the appropriate interpretation changed from nonamplified to equivocal HER2 status in 4.3% (100 of 2327) of the FISH cases and from nonamplified to equivocal status in 1.2% (25 of 2126) of the brightfield ISH cases, reflecting the tendency of FISH to generate higher average HER2 copy numbers per cell, which sometimes fell in the equivocal range (4.0–5.9 average HER2 copies per cell). HER2 copy number has become a more important factor following the release of the 2013 updated ASCO/CAP guidelines for HER2 testing.14 In the original 2007 guideline, an equivocal result for ISH was defined as a HER2 to chromosome 17 control ratio of 1.8 to 2.2 for dual-color probe assays or an average HER2 copy number per cell of 4 to 6 for single-color probe assays. In the updated 2013 guideline, an equivocal result was defined as a HER2 copy number of 4.0 or greater and less than 6.0 signals/cell and, for dual-color probes, a HER2 to CEP17 ratio of less than 2.0. Additionally, cases with 6.0 or more average HER2 copies per cell were reported as positive for HER2 amplification regardless of the HER2 to CEP17 ratio. These changes caused a shift toward more equivocal and positive results in HER2 testing since 2013,33–38 particularly with the use of alternative probes for reflex testing on equivocal cases.39,40
Several technical differences between FISH and brightfield ISH might contribute to the differences in the average HER2 to control ratio and HER2 copy number per cell. At the level of the cell nucleus, HER2 amplification frequently occurs intrachromosomally in homogenously staining regions of 1 or more chromosomes, typically generating 1 or more clusters of signals.41,42 Although FISH generates discrete fluorescent HER2 signals for each locus, brightfield ISH can generate aggregates of silver particles that coalesce in a manner that makes enumeration more difficult.43 Moreover, FISH is usually scored at a higher magnification (×60–×100) than brightfield ISH (×40–×60), and the use of a single-pass filter to enumerate HER2 and control signals individually in dual-color probe assays is possible for FISH but not for brightfield ISH. These factors can generate differences in the estimated HER2 copy number, as was shown in a recent study of a cohort enriched for equivocal cases.31
Another difference we observed between the 2 methods was the rate of assay failure; laboratories using brightfield ISH responded “unable to analyze” at a greater rate (9.6%) than did laboratories using FISH (6.0%). That might be related to technical challenges that are unique to brightfield ISH, such as the presence of black precipitate (“black silver dust”) that can obscure HER2 signals and large chromosome 17 centromere (control) signals that can mask smaller HER2 signals.43
There are some limitations to this study. First, the study was not enriched for breast cancers considered problematic or challenging, such as samples that are close to the equivocal or amplified thresholds. Second, the samples used for proficiency testing were small tissue cores and would not be reflective of larger tissue samples that are also frequently encountered in clinical practice. Third, result entry for proficiency testing differs from clinical test reporting and can result in clerical errors.
In conclusion, we found no significant differences in the overall assignment of HER2 status for FISH and brightfield ISH in routine breast cancer tissue core samples used for proficiency testing. However, FISH yielded a significantly higher average HER2 copy number per cell compared with brightfield ISH for amplified tissue cores, and FISH had a somewhat lower rate of technical failures. These differences could affect the evaluation of cases with values near the thresholds for HER2 equivocal or amplified status and, thus, possibly affect the choice of therapy.