Feasibility Study for Collection of HER2 Data by National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) Program Central Cancer Registries

Over the past several years, the prognostic and clinical importance of human epidermal growth factor receptor-2 (HER2) has increased considerably. Guidelines from the American Society of Clinical Oncology now recommend that all primary invasive breast cancers be evaluated for HER2 overexpression and/or amplification (1). This stems primarily from the fact that HER2 status will not only determine whether trastuzumab, a targeted adjuvant therapy, is a treatment option, but also reflects growing evidence that HER2 status is an independent prognostic indicator (2) and a marker of sensitivity to anthracycline-based chemotherapy regimens (3). While the importance of HER2 testing is widely agreed upon, there are few population-based studies of the frequency of HER2 testing in clinical settings, types of tests being used, results of these tests, and how they vary across factors such as age, race/ethnicity, and treatment facilities (4, 5). Contributing to these gaps in knowledge is the fact that few population-based cancer registries routinely collect HER2 data. Such data collection would add greatly to information on HER2 testing and resultant patient treatment. Ultimately, availability of population-based data on HER2 testing will be critical for determining dissemination of treatment protocols in the community and comparative effectiveness of various protocols.

A feasibility study was conducted to assess the ability of cancer registries in the Surveillance Epidemiology and End Results (SEER) Program to collect HER2 data and to identify barriers to data collection. The overall availability of HER2 data as well as characteristics associated with data availability are described.

Three SEER registries (Los Angeles, Louisiana, and Seattle-Puget Sound) participated in this feasibility study focused on invasive first primary breast cancer patients diagnosed between June and December 2007. At each registry facilities diagnosing breast cancer cases were divided into tertiles of facility size (large, medium, and small) based on the number of new cancer cases submitted to each registry in the last complete year of ascertainment. Numbers of cancer cases seen at large, medium, and small facilities may differ by registry. Facilities were selected at random from each tertile. A target number of patients for each registry was set at 30 from each of four large facilities, 30 from each of four medium facilities, and 10 from each of five small facilities for a total of 290 cases. A run-in phase was established to test data collection procedures. The accrual goal for the run-in period was 50 cases from each registry, with 20 from a total of at least three large facilities, 20 from a total of at least three medium facilities, and 10 from at least three small facilities. After the run-in period, the Los Angeles registry experienced a drastic change in resources, necessitating modification of some registry processes. The registry ceased further study participation during this transition period.

Certified tumor registrars from the registry's staff used a standard data collection form to extract the HER2 data from pathology reports and medical records. These forms, which included a generated number for patient identification, were submitted to the SEER Program at the NCI. Data were collected on the type of HER2 test performed [i.e., immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), and/or other/unknown], quantitative results of test(s) performed, interpretation of test(s) performed, and the source of data for each data item (i.e., pathology report, addendum to pathology report, laboratory report in chart, other electronic data, and other). In addition, the type of FISH test performed (ratio or copy number) and the cut point of the FISH test for positivity was obtained.

We defined HER2 positivity based on current clinical practice for interpreting the results of IHC and/or FISH testing, the two most commonly used approaches for evaluating HER2 status. FISH is generally considered as the “gold-standard” assay and is less susceptible to adverse effects of tissue handling than IHC. However, FISH is also more expensive and time consuming, and requires equipment and training that may not be available in all clinical settings. FISH testing can be reported in two ways: as a ratio of HER2 gene copy number to an internal control probe or as the number of gene copies per nucleus when an internal probe is not used. CEP17, a centromeric probe for chromosome 17, on which the HER2 gene is located, is used as an internal control. American Society of Clinical Oncology/College of American Pathologists guidelines (1) suggest a FISH test be deemed negative for a ratio of <1.8 or HER2 gene copy number <4.0, positive for a ratio >2.2 or a copy number >6.0, and equivocal for a ratio between 1.8 and 2.2 or a copy number of 4.0 to 6.0. This is estimated to result in an equivocal group of not more than 3%. Because studies have documented close to 95% concordance between IHC scores of 0/1+ and FISH nonamplification (i.e., HER2−) and >90% concordance between IHC scores of 3+ and FISH amplification (i.e., HER2+; refs. 6, 7), a common clinical protocol calls for the use of FISH testing only when IHC results are equivocal (i.e., 2+). Previous studies have shown that approximately 17% to 23% of breast tumors scored as 2+ by IHC are classified as positive by FISH (6-8). Thus, HER2 testing in individual cases may be performed by IHC alone, FISH alone, or a combination of IHC and FISH. In this study, HER2 status was derived by giving priority to FISH results over IHC, and to IHC results over a test of unknown type. Cases with a 2+ IHC result and no FISH testing were categorized as equivocal.

Information regarding patient age at diagnosis, race/ethnicity, tumor stage, and grade were obtained from registries as available. Data were collected for this study much closer to the date of diagnosis than is standard for the SEER Program; thus items, such as race and ethnicity, which would be complete when the case is officially reported to SEER, may not have been complete at the time of this study. Deidentified data from all three registries was combined in a database for analysis using the SAS statistical package. Percents of cases with HER2-related data items were calculated by various characteristics and statistical associations determined by χ² tests.

Among 541 observations obtained for analysis, HER2 data were either missing or it was unknown if a HER2 test was performed for 3.5% (n = 19); an additional 1.3% (n = 7) of records indicated that HER2 testing had been performed but results were unknown. Thus, a HER2 test was known to be performed on 96.5% (n = 522) of observations in this feasibility study, and HER2 results were available for 95.2% (n = 515). The completeness of these data varied somewhat by registry, with results from a HER2 test available for 95.9%, 95.5%, and 89.8% of cases from the Seattle-Puget Sound, Louisiana, and Los Angeles registries, respectively (Table 1). The proportion of women with available HER2 data only varied somewhat by age, with data available from 96.5% of women <50 years of age at diagnosis and available for 92.9% of women ≥70 years of age at diagnosis. This proportion also did not vary much by either disease stage or race/ethnicity. In contrast, HER2 data were available from 97.7% of women treated at large facilities compared with only 88.6% of those treated in small facilities. The only statistically significant association with percent of breast cancer cases with known HER2 results was for facility size (Table 1).

Overall, 17.7% (n = 91) of cases were HER2 positive, 2.5% (n = 13) were equivocal, and 79.8% (n = 411) were HER2 negative (Table 2). The majority of cases with available HER2 data were tested by IHC alone (50.9%), 35.3% were tested by both IHC and FISH, and 11.8% of cases were tested by FISH only (for seven cases the type of test performed was unknown, and for three cases a test of unknown type was performed along with IHC and/or FISH). For cases tested by FISH, the vast majority (88.2%) was reported as a ratio test, 3.3% reported copy number results, and the test type for the remaining 8.5% was unknown.

For 77.3% of cases with data, HER2 data were obtained from either the pathology report or an addendum to the pathology report. This increases to 93.2% if laboratory reports in the medical chart are also included. These percentages include the conservative requirement that if both IHC and FISH tests are performed, the results of both test types needed to be in the path report, addendum, or laboratory report.

This feasibility study shows that SEER central cancer registries can, with limited additional resources (i.e., using only documents currently obtained by registries), collect information on whether a HER2 test was performed, the type of test methodology used, and the result of that test for ∼95% of the population, based on a sample of breast cancer cases diagnosed in 2007. If sources of information are limited to pathology reports, their addenda, and laboratory reports in chart, this percentage drops to 88.7% and to 73.6% if only pathology report and addenda are used. The distribution of HER2 testing practices (i.e., IHC alone, FISH alone, and FISH and IHC) are likely to vary both on an institutional level, due to differences in institutional protocols, and according to patient population characteristics. Although FISH is generally considered the as gold-standard assay, the use of the more costly and involved FISH testing is likely to be lower in resource-limited facilities and in patient populations that are most likely to be HER2 negative on IHC (e.g., non-Hispanic white women, women diagnosed at an advanced age). Further work is needed to fully evaluate these potential associations.

The data source was collected because central cancer registry resources are limited and unlikely to increase in the near future, making it very important to assess the level of completeness that can be reached without excessive effort. The source of information on the HER2 test is critical to the ability of central cancer registries to collect these data since it determines the resources that need to be allocated. Sources not routinely accessed, including other laboratory reports, physician offices, additional electronic data, or obtaining the data through telephone calls, are much more labor intensive and would require increased resources. Thus, it is quite important that this feasibility study indicates that SEER central cancer registries can obtain fairly complete information on HER2 without expanding data sources. In addition, abstraction of HER2 data will also be facilitated by applications that highlight these items in pathology reports.

These analyses found that the only statistically significant association with percent of breast cancer cases with known HER2 results was facility size (Table 1). Further investigation is needed to determine the reasons behind this difference. It is important to note that this feasibility study was not designed or powered to assess differences in HER2 status by patient, tumor, facility, or registry characteristics. Larger scale population-based data collection is needed to better describe how the distribution of HER2 status varies with these factors. Expanded collection of HER2 data by SEER Program registries should be able to address these important issues.

As a result of the findings in this study, the majority of SEER central cancer registries will be collecting HER2 data for breast cancer cases diagnosed in 2009 on an optional basis with the intent to require these data for cases diagnosed in 2010. This will enable analyses to be performed that examine variation in HER2 results by patient and tumor characteristics and associations with survival outcome in population-based samples. Examination of associations between HER2 test results and treatment may require treatment data beyond that currently found in the SEER limited-use data file.

In general, it is difficult to determine when a biomarker “matures” to the point where it should be collected as part of population-based surveillance. The degree to which testing for that biomarker is part of standard medical practice, and to which assay methodology is stable contribute to this decision. HER2 likely represents a marker where its routine collection by population-based cancer registries is warranted, given the frequency of its testing, the stability of assay methodology, and its proven clinical and prognostic utility.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.