Aims: BRCA1-related breast cancer is associated with a basal-like phenotype, and is frequently oestrogen receptor (ER) and HER2 negative. The expression of epidermal growth factor receptor (EGFR) has been considered to be one component of the basal-like phenotype, but no standard criteria exist. This study investigates the relationship between EGFR expression, BRCA1 status and basal markers with respect to clinicopathological associations and prognosis, in addition to evaluating different criteria for EGFR assessment by immunohistochemistry.
Methods: A tissue microarray comprising 230 available cases, from a series of primary invasive breast cancer diagnosed in Ashkenazi Jewish women during 1980–1995, was stained for EGFR using the Dako PharmDX kit, and evaluated by Webslide virtual microscopy.
Results: EGFR was positive in 9–19% according to different criteria. Expression was associated with BRCA1 carrier status and basal-like markers as negative ER, positive cytokeratin 5/6 and positive P-cadherin staining. EGFR was prognostically significant by univariate and multivariate analysis within the group carrying germ-line BRCA1 mutations. Histological grade, axillary lymph node status and P-cadherin status had significant independent value in the final multivariate model including all cases, whereas EGFR was not significant in this model. All five scoring systems gave comparable results concerning clinicopathological associations and patient outcome, although the most restrictive criteria (EGFR-HI) tended to be most sensitive in predicting BRCA1 status, a basal phenotype, and patient prognosis.
Conclusions: EGFR expression, being present in 9–19% of the cases, was prognostically significant among BRCA1 mutated cases only. In multivariate survival analysis of all cases, no independent effect was seen. However, EGFR immunostaining might be relevant to predict the response to targeted therapy, and this should be studied further.
Statistics from Altmetric.com
Recent studies have indicated that invasive breast cancers can be divided into several molecular subgroups with prognostic importance.1–4 The basal-like category, which has been associated with aggressive features and adverse outcome compared with other breast cancers,5–7 is characterised by positive staining for antibodies against basal cytokeratins (CKs) such as CK5/6 and CK14, and is typically negative for oestrogen receptor (ER) and HER2 expression.6 8 In contrast, studies have indicated that the expression of EGFR (also known as HER1) is increased in basal-like carcinomas.6 8 Furthermore, BRCA1-related breast cancers are more likely to be of the basal-like type than non-hereditary tumours.9–11
In the past few years, anticancer treatment targeting EGFR has become available as antireceptor antibodies (cetuximab) or as small molecule tyrosine kinase inhibitors (gefitinib or erlotinib). After initial optimism, therapeutic success in recent studies has been more limited.12–14 Specific point mutations in the tyrosine kinase component are associated with treatment effects in non-small cell lung carcinoma, requiring gene analysis for patient selection.15 16 Treatment with EGFR-inhibiting antibodies has also been studied in colon carcinoma, head and neck cancers, and malignant gliomas. EGFR immunohistochemistry (IHC) as a criterion for treatment selection is controversial, with poor correlation between expression and treatment effects, most notably in colon carcinoma.17–19 A significant problem in this regard has been the lack of standardised scoring criteria in the literature, and the Dako PharmDX kit (Dako, Carpinteria, California, USA) has been promoted to address this issue. More relevant criteria for evaluation of test results and patient selection could reveal subgroups that might benefit from treatment. With this background, we wanted to explore two important issues: (1) what is the significance of EGFR expression in different subgroups of breast cancer, with special reference to BRCA1 mutations, and (2) does changing the criteria for reporting levels of EGFR have a clinically relevant impact? To answer these questions, we studied an established breast cancer cohort with a high frequency of germ-line BRCA1 mutations and long-term follow-up.
The study design is an ethnically restricted single hospital-based retrospective cohort study, as described in previous publications.9 10 20 The study was approved by the institutional review board of the hospital. Of 309 consecutive cases of Ashkenazi Jewish women age 65 years or less and diagnosed with a first primary, non-metastatic, invasive breast cancer between 1 January 1980 and 1 November 1995 at the Sir Mortimer B Davis Jewish General Hospital, Montréal, Québec, Canada, 17 (5.5%) were excluded because of (1) inability to locate pathology blocks, (2) only carcinoma in situ was present on the available pathology blocks, or (3) DNA could not be adequately amplified after repeated attempts, leaving 292 cases for further studies. Breast cancer blocks were identified from each of these women, and clinicopathological and follow-up information was obtained by chart review.
No synchronous cancers were included, and only the first of two metachronous cancers was counted. Development of a second primary is registered as a disease event in our database. In the common data set described below (n = 180), 12/180 cases developed a contralateral breast cancer during the course of the disease, while 11/180 cases were registered as having a new primary cancer. In-depth details concerning chemotherapy regimens given were not available. In the common data set (n = 180), 89 cases did receive chemotherapy at first diagnosis and 44 of these were given an anthracycline-based regimen. Chemotherapy was not given as neoadjuvant treatment.
Specimens were reviewed and re-examined for histological type, histological grade and lymph node status, and were stained using IHC for ER and progesterone receptor (PR) expression, HER2, CK5/6 and P-cadherin, as described in previous publications.10 21–24 Histological grade was determined according to the Nottingham criteria.25 After review, 230 cases had sufficient tumour available in the blocks for arraying. Clinical information and results from previous investigations were available on 223 of these cases in our database. Finally, 180 cases had complete sets of EGFR scores (sufficient quality for all criteria) and follow-up information, and this set was used for further statistical analyses. The distribution and number of missing values in this common data set are specified in table 1.
Tissue microarray (TMA) technology has become a widespread and accepted method for high-throughput analysis of microscopic slides.26–28 Two 0.6 mm cores from representative areas in the donor blocks were transferred to a recipient block using a precision instrument from Beecher Instruments (Beecher Instruments, Sun Prairie, Wisconsin, USA).
IHC staining of 4 µm TMA slides was performed using the Dako EGFR PharmDx kit (K1492), according to the manufacturer’s instructions. Briefly, antigen retrieval was performed using proteinase K, and the pre-diluted EGFR clone 2-18C9 antibody was applied for 30 min at room temperature. The detection system consisted of a polymer–horseradish-peroxidase with diaminobenzidine chromagen. The slides were counterstained with Gill I haematoxylin. Examples of different staining results are shown in fig 1.
Stained TMA slides were digitised using Webslide technology for virtual microscopy (Bacus Laboratories, Lombard, Illinois, USA). The images were stored on the GPEC WebSlide server, and were accessed using the WebSlide Viewer Java Applet and Mozilla Firefox web-browser.
Evaluation of staining
Cases with lost cores (random drop-outs) and cases without sufficient tumour tissue in the cores or insufficient staining quality were excluded from the final analysis, as were cases without sufficient follow-up information, leaving 180 cases available for EGFR analysis using all criteria. As there is presently no consensus on how to evaluate EGFR expression by IHC, the staining was recorded according to five different sets of criteria to explore which of these showed best correlations with clinicopathological characteristics and outcome data in breast carcinoma. IHC scoring was performed blinded for patient characteristics and outcome.
Dako criteria (EGFR-DA)
The DAKO EGFR PharmDX diagnostic kit has obtained US Food and Drug Administration approval, and is supplied with instructions for evaluation. Only membranous staining is considered a positive result. Weak (1+) intensity is defined as faint and incomplete membrane positivity. Moderate (2+) intensity and strong (3+) staining are varying degrees of circumferential staining of membranes. The criteria define a proportion of stained cells above 1% at any intensity as a positive result.
Zymed criteria (EGFR-ZY)
The Zymed 31G7 antibody datasheet (Zymed, Carlsbad, California, USA) outlines a scoring system where membranous staining intensity is graded from weak to strong (0–3), and a proportion of stained cells above 10% is required for positive diagnosis. We used the same criteria for intensity scores as for the staining index (see below). This scoring system also corresponds to the criteria outlined in the articles by Bhargava et al and Reis-Filho et al in their studies of EGFR gene amplification.29–31 For further analysis, grades 2–3 were considered positive, as done by Reis-Filho et al.31
Staining index (EGFR-SI)
Where standardised IHC scoring criteria are lacking, we have previously used a semiquantitative and subjective grading system, considering the intensity of staining and the proportion of tumour cells showing a positive reaction, as previously described.10 24 Intensity was recorded from 0 (no staining) to 3 (strong staining). For evaluation of intensities, we applied similar criteria as suggested by DAKO in the PharmDX kit instructions. The proportion of positive cells was recorded as 1 (<10%), 2 (10–50%) and 3 (>50%). A staining index (SI) was defined as the product of staining intensity and area, and a final score of zero to nine was established. SI = 2 corresponded to weak staining in 10–50% of the area, or moderate staining in any area. To avoid extensive overlap with the EGFR-DA set, and to correspond with more general IHC criteria in our laboratory, cases were categorised as positive (SI = 2–9) or negative (SI = 0–1).
SI, strong staining (EGFR-HI)
Studies of IHC as selection criteria for cetuximab in colon carcinoma,18 32–34 and the association between activating mutations and strong IHC staining in the studies by Bhargava and Reis-Filho,29 30 suggest that the prognostic and predictive value of EGFR staining might increase if the strongest stained cases were selected. Because of these considerations, we wanted to explore the effect of strong staining on associations and outcome. The EGFR-SI scores were dichotomised, comparing SI scores 6–9 with scores 0–5.
Vancouver criteria (EGFR-VA)
The cases were scored on a scale from 0 to 2, as outlined by Nielsen et al.6 Any membranous or cytoplasmic staining in tumour cells was considered positive. Scores: 0, no staining; 1, weak staining and/or <20% stained tumour cells; 2, strong staining of >20% of tumour cells. For further analysis, scores 1–2 were considered positive.
BRCA1 and BRCA2 mutation status
Mutation analysis was carried out as previously described,21 looking specifically for the recurrent mutations in the Ashkenazi Jewish population (BRCA1: 185delAG, 5382insC; BRCA2: 6174delT),
Clinical, pathological, and molecular data were collected in a mutually blinded fashion. The SPSS version 15.0 statistical program was used for analysis. Patient characteristics were compared using Pearson’s χ2 or Fisher’s exact test. Odds ratio (OR) was calculated by the Mantel–Haenszel method or as Peto odds ratio. Two-tailed p values ⩽0.05 were considered significant. Borderline statistical significance was defined as p values between 0.05 and 0.10. Survival rates were calculated from the date of primary surgery until death from breast cancer. The mean follow-up of those who did not die of breast cancer was 9.9 years (n = 180); the mean overall follow-up was 8.3 years. Survival plots and cumulative survival probabilities were estimated using the Kaplan–Meier method and significance was assessed with the two-sided Log rank test. Univariate survival analysis was also done by Cox regression analysis.
To estimate the risk of death from breast cancer, Cox proportional hazards models were used including all factors with p level <0.15 in univariate analysis (ie, age, histological type and grade, tumour diameter, axillary lymph node status, ER, PR, HER2, CK5/6, P-cadherin and BRCA status). Significance was assessed at the p = 0.05 level using two-sided tests. The final most parsimonious model was built using the log likelihood ratio test, employing the SPSS default “forced entry” approach.
To evaluate EGFR as a predictor of BRCA1 mutations, logistic regression models were constructed including different EGFR criteria adjusted for age, histological grade and ER status.
EGFR expression and associations with other markers
Using the EGFR-DA criteria (table 2), 34 cases were positive (18.9%). There were significant associations with higher histological grade (grade 3: OR 10.01, p<0.0005) and negative hormone receptor status (ER and PR: both <0.0005). Further, there were highly significant associations with individual basal markers (CK5/6: OR 8.30, p<0.0005; P-cadherin: OR 9.15, p<0.0005) and BRCA1 mutations (OR 9.16, p<0.0005). In the BRCA1 mutated subgroup, 61% (11/18 cases) were EGFR positive, versus 14% in the BRCA1-negative group (23/162 cases). No association was seen with HER2 status, or with tumour size or lymph node status.
Other criteria for the determination of EGFR expression by IHC (table 3) revealed comparable results. The frequency of positive cases was 11.7% (n = 21) by EGFR Zymed criteria, 16.1% (n = 29) for EGFR by SI, and 14.4% (n = 26) for EGFR by the Vancouver criteria. Strongly staining cases as defined by EGFR-HI, constituted 8.9% (n = 16). All five sets of EGFR criteria were associated with histological grade, ER status, basal markers (CK5/6 and P-cadherin) and BRCA1 status, with comparable results. In contrast, no relationships with tumour size, nodal status or HER2 expression were seen. EGFR expression results by different criteria were all significantly inter-related (all associations p<0.0005). The most highly correlated criteria were EGFR-DA versus EGFR-SI (κ = 0.90) and EGFR-ZY versus EGFR-HI (κ = 0.85). Although all scoring criteria showed highly significant associations, EGFR-HI and EGFR-VA were superior in predicting basal-like features as evaluated by OR, whereas EGFR-HI (OR 10.47) and EGFR-DA (OR 9.16) tended to be better in predicting BRCA1 status.
EGFR as a predictor of BRCA1 mutations was also evaluated by multiple logistic regression, analysing each set of criteria in separate models adjusting for age at diagnosis, histological grade and ER status. Considering EGFR-DA criteria, EGFR expression (hazard ratio (HR) 6.5, p = 0.004) and age (HR 0.94, p = 0.041) were significant, whereas EGFR-HI (HR 4.94, p = 0.015) and ER status (HR 0.21, p = 0.027) were significant in a different model (results for the other models are not shown). Based on this, EGFR-DA and EGFR-HI tended to be the strongest predictors of BRCA1 mutations of the different criteria in these analyses. Age and ER status were also important predictors of BRCA1 positive breast cancer.
EGFR EXPRESSION AND PATIENT SURVIVAL
Four of the grading systems for EGFR were significantly associated with 10 year survival at comparable significance levels by univariate Cox regression analysis, with EGFR-HI criteria having the highest HR (3.27). EGFR-SI showed an association of borderline significance (p = 0.062) (table 4).
EGFR Kaplan–Meier plots were similar, with a more dominant effect on short-term survival (fig 2). The number of cases in the series restricts subgroup analysis. Stratified by BRCA status, Kaplan–Meier plots using the most inclusive EGFR-DA criteria showed significant effects on survival only in the BRCA1-positive subgroup (n = 18; Log rank test: p = 0.034) (fig 3).
EGFR criteria were analysed in separate multivariate Cox models, including variables with p<0.15 in univariate analysis. Missing values were recoded to include the maximum number of cases. In the final parsimonious model, the effect of EGFR positivity was lost by all staining criteria, leaving histological grade (HR 5.16–5.29), axillary lymph node status (HR 1.89) and P-cadherin status (HR 2.70) as independently significant prognostic factors in this series (table 5).
To further examine the effect of EGFR within the group with BRCA1 mutations (n = 18; 10 events), we attempted a multivariate analysis with tumour size, histological grade, and lymph node status in addition to EGFR expression. We first used the most inclusive EGFR-DA criteria. Considering the limited number of cases in subgroups, EGFR was the only remaining variable (EGFR-DA HR 4.73, 95% confidence interval (CI) 0.98 to 22.73, p = 0.052), perhaps indicating a prognostic impact superior to that of tumour size, histological grade and lymph node status in cases with BRCA1 mutations. The other EGFR criteria had no consistent prognostic impact in these models.
Limited information about adjuvant chemotherapy was available, and 89 cases received such treatment. By univariate Cox analysis, chemotherapy was significantly associated with survival (HR 1.75, 95% CI 1.04–2.95; p = 0.036). When EGFR expression and chemotherapy were included in multivariate analysis in addition to tumour size, histological grade, nodal status, and P-cadherin expression (or CK5/6), no consistent prognostic impact of EGFR expression or chemotherapy was observed in these models.
In this series of breast cancers with a high frequency of BRCA1-positive cases and long-term follow-up, we found EGFR expression in 9–19% of all cases, depending on the evaluation criteria. Positive staining was significantly associated with BRCA1 status and individual basal markers; these findings are in line with those of a few recent studies.6 8 11 EGFR expression has been included in the definition of a “core basal profile”,6 although there is presently no consensus on how the basal-like subgroup of breast cancers should be defined, and the relationship between the basal-like category and the “triple negative phenotype” is presently being discussed.35–38 The EGFR expression frequencies observed in our study are close to those reported in a large TMA series of 1500 breast cancers,39 where EGFR was positive in 20% of the cases. The frequencies are, however, lower those found in a meta-analysis of more than 5000 patients from 40 different studies, where Klijn et al40 reported EGFR expression in 48% of the cases (range 14–91%) based on several different techniques. Some of this variation might be explained by the use of different antibodies and IHC protocols.
The prognostic effect of EGFR in breast cancer has been a matter of discussion.40 In the large meta-analysis by Klijn et al, five of nine studies showed significant effect on short-term survival, whereas three of five reports revealed a tendency towards reduced survival by long-term follow-up.40 In our present study, there was no independent prognostic impact of EGFR staining when all cases were examined, since only histological grade, lymph node status, and expression of the basal marker P-cadherin remained significant in the final multivariate model. However, an influence of EGFR expression on patient survival was found among BRCA1-positive cases, even in multivariate analysis. In spite of the limited number of cases in the current report, these findings are interesting and deserve analysis in a larger series of breast cancer cases with long-term follow-up and germ-line mutation status. It is possible that the divergent findings in previous studies on the effect of EGFR on survival are due to certain host factors, such as BRCA1 mutations. It is notable that EGFR mutations are much more likely to be present in lung cancers arising in non-smoking Asian women with bronchoalveolar lung cancer than in other ethnic groups and lung cancer sub-types.13 41 In a similar fashion, there might be an interactive effect on survival which involves tumour type, BRCA1 mutations and EGFR status in breast cancer. We consider that genetic and IHC evaluation of EGFR could be of importance in the selection of patients for targeted therapy, particularly as new therapeutic possibilities have been suggested for the hormone receptor and triple negative subgroups.12 13 This issue needs to be studied in more detail.
There is currently no consensus on how to evaluate EGFR expression in breast cancer by IHC, and we therefore evaluated five sets of criteria using TMA slides and a Webslide set-up. Further, antibody selection and staining protocols are not standardised in the literature. In our series, the frequency of EGFR expression varied from 9 to 19%, with the EGFR-DA and EGFR-HI criteria giving the highest and lowest figures. All EGFR scoring systems showed comparable associations with clinico-pathological variables, BRCA1 status and patient survival. The relationships with basal markers CK5/6 and P-cadherin, as well as ER expression, were slightly stronger using the EGFR-HI and EGFR-VA criteria, whereas EGFR-HI and EGFR-DA scoring showed better prediction of BRCA status, although differences were small.
By multivariate survival analysis, no independent prognostic information was obtained by any of the criteria for EGFR assessment when all cases were included, although EGFR-HI had the strongest impact on survival in univariate analysis. Within the subgroup of BRCA1-positive breast cancers, EGFR expression by the most inclusive DAKO criteria was significant in univariate survival analysis and also tended to show an independent influence on survival (p = 0.052) in the multivariate model. The other criteria did not show significance, although the low number of cases in these subgroups should be pointed out. Last, we conclude that the Webslide set-up might be useful in multicentre quality assurance programmes and research projects, which appear necessary to improve standardisation in this field.
There are no internationally recognised criteria for epidermal growth factor receptor (EGFR) evaluation in breast cancer.
EGFR expression by immunohistochemistry was found in 9–19% of breast cancers using five sets of evaluation criteria.
EGFR expression was significantly associated with BRCA1-positive cases and individual basal markers (CK5/6, P-cadherin).
EGFR expression was a significant prognostic factor by univariate and multivariate analysis of BRCA1-positive cases, although it did not represent an independent prognostic factor among all cases.
In conclusion, our present study confirms a significant association between EGFR expression and BRCA1 mutations in breast cancer as well as with individual basal markers. EGFR expression was a significant prognostic factor by univariate analysis among BRCA1-positive cases, and the only remaining prognostic factor in a multivariate survival analysis of the BRCA1 mutated subset. Although the cases with BRCA1 mutations were few, the impact of these findings for patient care, eventually involving targeted therapy, should be studied in more detail.
This study was sponsored by The Norwegian Cancer Society, The Norwegian Research Council and Helse Vest HF (LAA) and by the Canadian Breast Cancer Research Alliance (WDF).
Competing interests: None.
Ethics approval: Ethics approval was obtained from the Sir Mortimer B Davis Jewish General Hospital.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.