Information on hormone receptor status, especially estrogen receptor (ER) status, is important for therapeutic decision making in early and advanced breast carcinomas. Approximately 70% of primary breast cancers are positive for ER, which predicts a favorable initial tumor response to ER-targeted endocrine therapy and is associated with better tumor differentiation and a more indolent natural history.1, 2 For ER-positive, non-life-threatening metastatic breast cancer, endocrine manipulation represents a major option for initial treatment.3 ER status is usually determined in an invasive breast carcinoma at the time of initial diagnosis by immunohistochemical (IHC) staining, and the information is used mainly for selection of adjuvant therapy. However, the ER status obtained from the primary breast carcinoma is also used frequently to select candidates for endocrine therapy in the metastatic setting. This is because metastatic tumors are not routinely biopsied in many hospitals, largely because of the difficulty of sampling metastatic tumors and the lack of a standard testing method. Given the fact that metastatic breast carcinoma often occurs several years after the initial diagnosis of primary carcinoma, and many patients have already received a variety of systemic therapies by the time of relapse, controversy remains as to whether ER status obtained from a primary tumor can reliably guide therapeutic decisions for metastatic disease of the same patient and whether ER testing needs to be repeated for the metastatic carcinoma. In earlier studies, the reported concordance between primary and metastatic tumors varied widely from 55% to 95%.4-15 It is notable that the majority of these studies included relatively small cohorts of patients and used the ligand-binding assay to test ER, which is less reliable than IHC.16
In the current study, we used IHC to evaluate stability of ER status in paired primary and metastatic tumor samples from 227 patients, and to determine the effect of previous disease course and intervening systemic therapy on ER status of metastatic tumors.
MATERIALS AND METHODS
Patients and Samples
We retrospectively searched our institution's pathology database for women with invasive breast carcinoma for whom ER status of both the primary tumor and a subsequently sampled metastatic tumor was known. We first identified metastatic breast carcinomas that were sampled at The University of Texas M. D. Anderson Cancer Center with available ER status between the period of March 2003 and August 2008. We then recorded the ER status of the paired primary breast carcinomas using information in our institutional patient-care database (ie, pathology reports and/or clinical notes). These primary tumors were diagnosed and tested for ER status at either an outside referral hospital or M.D. Anderson Cancer Center between September 1984 and June 2006. As a result, we identified a total of 227 patients with the matched ER status data; each patient had one metastatic carcinoma for which ER status was available, although some patients developed more than one metastasis. We retrospectively reviewed and recorded each patient's demographic information, the systemic treatment received, and tumor characteristics. Table 1 summarizes the characteristics of patients and their primary tumors, and Table 2 lists the sites of the metastatic tumors studied. The mean age of the patients at diagnosis of primary carcinoma was 51 years (range, 21-86 years). The median time interval between diagnosis of primary tumor and sampling/testing of the paired metastatic tumor was 61 months (range, 1.5-276 months).
Table 1. Characteristics of 227 Patients and Their Primary Breast Carcinomas
| <50||131 (57)|
| ≥50||96 (43)|
| Caucasian||144 (63)|
| Others||83 (37)|
| <2 cm||81 (36)|
| 2-5 cm||103 (45)|
| >5 cm||22 (10)|
| Not determined||21 (9)|
| Ductal||202 (89)|
| Lobular||12 (5)|
| Mixed ductal and lobular||8 (4)|
| Others||5 (2)|
|Modified Black nuclear grade|
| I||14 (6)|
| II||71 (31)|
| III||128 (57)|
| Not determined||14 (6)|
| Positive||157 (69)|
| Negative||70 (31)|
| Positive||105 (46)|
| Negative||111 (49)|
| Not determined||11 (5)|
| Positive||41 (18)|
| Negative||134 (59)|
| Not determined||52 (23)|
Table 2. Location of 227 Metastatic Tumors
|Locoregional metastases||124 (55)|
| Axillary lymph node||47 (21)|
| Supraclavicular lymph node||37 (16)|
| Infraclavicular lymph node||6 (3)|
| Ipsilateral anterior chest wall||34 (15)|
|Distant metastases||103 (45)|
| Lung||11 (5)|
| Liver||24 (10)|
| Effusion fluid||14 (6)|
| Bone||22 (10)|
| Distant lymph node||20 (9)|
| Distant soft tissue||8 (3)|
| Other visceral organs||4 (2)|
Most patients received systemic therapy during the intervening period; in some patients, however, the metastasis was discovered and sampled before completion of the required therapy. Specifically, 159 (70%) of the 227 patients had received chemotherapy (mostly anthracycline-based regimens with or without taxane): preoperative chemotherapy alone in 23 patients, preoperative and postoperative therapy in 39 patients, and postoperative chemotherapy alone in 97 patients. Twenty-four (11%) patients whose primary cancer was positive for HER2 received trastuzumab. One hundred twenty-eight (56%) patients received adjuvant endocrine therapy, including tamoxifen, aromatase inhibitors, and/or other selective ER modulators such as raloxifene; 48 of these 128 patients completed 5 years of endocrine therapy before the metastatic tumor was sampled. Semiquantitative comparison of ER values was conducted for 92 tumor pairs for which ER expression value (as a percentage) was available for both tumors.
This study was conducted with the approval of the M.D. Anderson Institutional Review Board.
ER Staining and Scoring
For all 227 primary breast carcinomas, ER status was tested on formalin-fixed, paraffin-embedded tissue sections by IHC staining. Of these, 56 tumors were tested at M. D. Anderson on paraffin tissues obtained at this hospital or from outside referral hospitals. Another 55 were tested at outside hospitals, but staining results were reviewed by pathologists at M. D. Anderson. Information on ER status of the remaining 116 primary tumors was obtained either from outside pathology reports (n = 31) or clinical notes (n = 85).
All 227 metastatic breast carcinomas were sampled at M. D. Anderson. Four were sampled via tissue biopsy (core needle or excision) and 223 via fine-needle aspiration (FNA). The methodology of ER testing on cytologic specimens is described elsewhere.17 ER status was defined as positive if greater than10% of tumor cells demonstrated nuclear staining.
ER status of metastatic tumors was compared with that of matched primary tumors. Agreement of ER status between the paired tumors was expressed by both concordance and the Cohen kappa (κ) coefficient with a corresponding 95% confidence interval (CI). The relationship between κ and level of agreement was proposed by Landis and Koch:18 0.00-0.20, slight agreement; 0.21-0.40, fair agreement; 0.41-0.60, moderate agreement; 0.61-0.80, substantial agreement; and 0.81-1.00, almost perfect agreement. Comparison of ER status change between paired conditions was analyzed using the Fisher exact assay or chi-square test.
The overall concordance of ER status between the primary and paired metastatic tumors was 92.5% (210/227, 95% CI, 89.1-95.9%), including 147 of positive status and 63 of negative status, with a Cohen κ coefficient of 0.8265 (95% CI, 0.7474-0.9056) (Table 3). ER status was discordant in 17 (7.5%) patients: 7 with negative status in the primary tumor and positive status in the metastatic tumor, and 10 with positive status in the primary tumor but negative status in the metastatic tumor (Table 3). In the 17 discordant pairs (Table 4), 15 primary tumors were sampled and tested at an outside hospital. In 7 patients, the discordance was between marginal positive status (ie, 10%-15%) in one tumor, but negative status in its counterpart. Four of the 17 patients had 2 tumor nodules in the breast, but in each case ER status was assessed in only 1 of the nodules.
Table 3. Effect of Various Factors on ER Concordance
| Locoregional||124||90.3||0.8013||6||6|| |
|Interval between assays|
| <5 years||129||93.8||0.8738||4||4|| |
| ≥5 years||98||90.8||0.6557||6||3||.3978|
| Yes||159||92.5||0.8268||7||5|| |
| Yes||128||91.4||0.5754||9||2|| |
|Sample type (metastases)|
| FNA, smear||154||92.2||0.8241||7||5|| |
| FNA, cell block||69||92.8||0.8265||3||2||.8871|
| Tissue biopsy||4||100||1.00||0||0|| |
Table 4. Information on 17 Patients With Discordant ER Status in Primary and Paired Metastatic Tumors
|Negative to Positive Conversion|
| Case 1||Negative||1||Tissue||OS||10||LN, axillary||FNA||MDA|
| Case 2||Negative||2||Tissue||OS||10||LN, supraclavicular||FNA||MDA|
| Case 3||5||1||Tissue||MDA||10||bone||FNA||MDA|
| Case 4||Negative||1||Tissue||OS||50||LN, supraclavicular||FNA||MDA|
| Case 5||0||2||Tissue||OS||12||chest wall||FNA||MDA|
| Case 6||Negative||1||Tissue||OS||10||breast||FNA||MDA|
| Case 7||5.5||1||Tissue||OS||100||LN, supraclavicular||FNA||MDA|
|Positive to Negative Conversion|
| Case 8||Positive||1||Tissue||OS||0||LN, pelvic||FNA||MDA|
| Case 9||Positive||2||Tissue||OS||0||chest wall||FNA||MDA|
| Case 10||Positive||2||Tissue||OS||0||LN, axillary||FNA||MDA|
| Case 11||85||1||Tissue||OS||<5||LN, axillary||FNA||MDA|
| Case 12||80||1||Tissue||OS||0||Breast||FNA||MDA|
| Case 13||Positive||1||Tissue||OS||0||Liver||FNA||MDA|
| Case 14||15||1||Tissue||MDA||Negative||LN, infraclavicular||FNA||MDA|
| Case 15||Positive||1||Tissue||OS||Negative||LN, supraclavicular||FNA||MDA|
| Case 16||90||1||Tissue||OS||Negative||Liver||FNA||MDA|
| Case 17||10||1||Tissue||OS||Negative||Lung||FNA||MDA|
Effect of metastatic site on ER concordance
Of the 227 metastatic tumor samples, 124 (55%) were from a locoregional metastasis and 103 (45%) from a distant metastasis (Table 2). The concordance between primary tumors and the paired locoregional metastases was 90.3% (112/124, 95% CI, 85.1-95.5%) with a Cohen κ coefficient of 0.8013 (95% CI, 0.6945-0.9081), whereas the concordance between primary tumors and the paired distant metastases was 95.1% (98/103, 95% CI, 91.0-99.3%) with a Cohen κ coefficient of 0.8421 (95% CI, 0.7081-0.9761). The difference between the two groups, however, was not statistically significant (P = .1693) (Table 3).
Effect of time interval between measurements on ER concordance
We stratified patients on the basis of the time interval from initial diagnosis of the primary carcinoma to sampling/testing of the paired metastatic tumor. The interval was less than 5 years for 129 patients and 5 years or longer for 98 patients. The concordance of ER status between primary tumor and the paired metastasis in the former group was 93.8% (121/129, 95% CI, 89.6-98.0%) with Cohen κ coefficient of 0.8738 (95% CI, 0.7891-0.9585). In the latter group, the concordance was 90.8% (89/98, 95% CI, 85.1-96.5%) and the Cohen κ coefficient 0.6557 (95% CI, 0.4487-0.8627). The differences between the groups was not statistically significant (P = .3978) (Table 3).
Effect of intervening chemotherapy on ER concordance
Of the 227 patients, 159 (70%) had received chemotherapy before their metastatic tumor was sampled; the ER status between primary and paired metastatic tumors agreed in 92.5% of these patients (147/159, 95% CI, 88.4-96.6%) with a Cohen κ coefficient of 0.8268 (95% CI, 0.7328-0.9208) (Table 3). ER status changed in 12 of these patients: 7 converted from positive to negative and 5 from negative to positive. Of the 68 patients who received no intervening chemotherapy, tumor ER status agreed in 92.7% (63/68, 95% CI, 86.4-98.9%) with a Cohen κ coefficient of 0.8255 (95% CI, 0.6787-0.9723) (Table 3). ER status changed in 5 of these patients: 3 from positive to negative and 2 from negative to positive. The difference in ER concordance between these 2 groups was not statistically significant (P = .9594) (Table 3).
Effect of intervening endocrine therapy on ER concordance
A total of 128 (56.4%) patients had received intervening endocrine therapy and 99 (43.6%) patients had not. Of the 128 patients who received endocrine therapy, 117 had an ER-positive primary carcinoma and the other 11 an ER-negative, but progesterone receptor (PR)-positive primary tumor. ER status was concordant in 91.4% of these patients (117/128, 95% CI, 86.6-96.3%) with a Cohen κ coefficient of 0.5754 (95% CI, 0.3533-0.7975). ER status changed in 11 of these patients: 9 from positive to negative and 2 from negative to positive. Of the 99 patients who received no intervening endocrine therapy, 40 had an ER-positive primary carcinoma, 4 an ER-negative/PR-positive primary, and 55 an ER-negative/PR-negative primary. ER status was concordant in 93.9% of these patients (93/99, 95% CI, 89.2%-98.6 %) with a Cohen κ coefficient of 0.8761 (95% CI, 0.7804-0.9719). ER status changed in 6 of these patients: 1 from positive to negative and 5 from negative to positive. No significant difference between the endocrine treatment and no endocrine treatment groups was observed (P = .4724).
Of the 117 patients with an ER-positive primary breast carcinoma who had received endocrine therapy, 45 received tamoxifen alone, 51 received tamoxifen and an aromatase inhibitor, 19 received an aromatase inhibitor alone, and 2 patients received raloxifene alone. A change of ER status from positive in primary tumor to negative in metastatic tumor was observed in 7.7% (9/117) of patients: 13.3% (6/45) of those in the tamoxifen group, 3.9% (2/51) of those in the tamoxifen and aromatase inhibitor group, and 5.3% (1/19) of those in the aromatase inhibitor group (Table 5). The differences among the treatment groups had no statistical significance (P > .05).
Table 5. ER Status Change on 117 Patients With ER-Positive Primary Tumors Treated With Endocrine Therapy
| Tamoxifen (n=45)||6||13.3|
| Tamoxifen + aromatase inhibitor (n=51)||2||3.9|
| Aromatase inhibitor (n=19)||1||5.3|
| Other (n=2)||0||0|
|Duration of treatment|
| ≥5 years (n=45)||5||11.1|
| <5 years (n=72)||4||5.6|
When patients were stratified on the basis of duration of endocrine therapy, we found that 45 of the 117 patients completed a 5-year course of endocrine therapy and 72 patients did not. The positive to negative conversion of ER status occurred in 11.1% (5/45) of the former group and 5.6% (4/72) of the latter group. However, the difference between the two groups did not reach statistical significance (P = .3026).
Effect of metastatic tumor sample type on ER concordance
Four of the 227 metastatic tumors were sampled via tissue biopsy (core needle or excision), and ER status between primary and metastatic tumors agreed in all 4. In the remaining 223 patients, the metastatic tumors were sampled via FNA: 69 of them had cell block sections and 154 of them had direct smears available for ER test. The concordance of ER status in the cell block and smear groups was nearly identical: 92.8% (64/69, κ = 0.8265) and 92.2% (142/154, κ = 0.8241) (P = .8871) (Table 3).
Semiquantitative comparison of ER expression value between primary and metastatic tumors
Of the 92 tumor pairs available for semiquantitative comparison of ER expression, the ER status was concordant in 94.6% (87/92, 95% CI, 89.9-99.2%) with a Cohen κ coefficient of 0.8373 (95% CI, 0.6994-0.9753). We then arbitrarily stratified ER expression into 4 ranges (0%, 1-9%, 10-49%, ≥50%) to evaluate quantitative change (Table 6). Compared with ER in primary tumors, the ER expression level in metastatic tumors remained in the same range in 71 of the 92 (77.2%) patients, increased in 13 (14.1%) patients (which led to a negative-to-positive status change in 3 patients), and decreased in 8 (8.7%) patients (which led to a positive-to-negative status conversion in 2 patients). Of the 21 pairs in which ER expression level in the metastatic tumor differed from that in its primary tumor counterpart, 17 (81%) showed only a 1-step change in ER expression. The distribution of the ER-positive cells in the 4 ranges was similar: 15%, 5%, 11%, and 61% for the primary tumors and 13%, 6%, 11%, and 62% for the metastatic tumors.
Table 6. Semiquantitative Comparison of ER Expression Value in 92 Pairs of Primary and Metastatic Tumors
Although testing biomarkers in metastatic breast carcinoma has become increasingly popular with the advent of high-resolution imaging and radiological techniques for tissue sampling, it is not a standard procedure in many hospitals. If ER status from a primary breast carcinoma is used to guide therapeutic decisions in a metastatic setting, it is essential that the ER status remains stable with time and is not influenced by disease course or intervening systemic therapy. In this retrospective study, we observed a high overall concordance (92.5%, 210/227) in ER status between primary and paired metastatic carcinomas. ER status changed in both directions through disease progression, either from positive to negative or vice versa.
Tumor cells that metastasize locally seem biologically distinct from cells that invade and penetrate through blood vessels with subsequent clonal overgrowths at distant sites.19 Distant metastasis likely represents a more aggressive disease course than locoregional metastasis and often occurs years after diagnosis of the primary cancer. With respect to the influence of metastatic site on ER status change, published data showed contradictory results. A study by Kamby et al suggested that distant metastases tend to have lower concordance with the corresponding primary tumor than locoregional metastases.20 In that study, ER status of primary breast carcinomas agreed with that of regional lymph node metastases in 90% of the patients, but the concordance rates were 75% with liver metastases and 58% with bone metastases. Holdaway et al reported a significant reduction of ER expression in chest wall recurrences and visceral metastases compared with locoregional lymph node metastases, raising a possibility that tumor cells with different ER expression may be selected and adapted to grow in different anatomical sites.10 More recent studies, however, demonstrated similar rates of discordance for local recurrences and distant metastases.6, 13, 21, 22 Consistent with this, our data did not show a significant influence of metastatic site on ER concordance rate. In fact, the concordance rate for distant metastases was slightly higher than that for locoregional metastases (95.2% and 90.3%, respectively). Our data also showed that the time interval from diagnosis of primary carcinoma to measurement of ER in the paired metastatic tumor did not significantly affect ER concordance, although the concordance rate was slightly higher with intervals shorter than 5 years (93.8%) than with longer intervals (90.8%). Studies from other investigators also indicated that time interval alone does not significantly affect ER status.6, 14, 22
Effect of chemotherapy on hormone receptor expression is controversial in the literature. Considering that breast cancer is a biologically heterogeneous disease with variability of ER expression among cells in the same tumor, and that ER-negative tumor cells generally respond better to chemotherapy, it is conceivable that chemotherapy may lead to “enrichment” of ER-positive tumor cells by preferentially killing poorly differentiated ER-negative tumor cells. A few studies based on relatively small series have reported the effect of chemotherapy on ER status. The majority of these studies examined the influence of chemotherapy administered in the neoadjuvant setting, and ER expression was compared between pretherapy and posttherapy specimens from the same primary tumors. Some studies reported a significant change of ER status (mostly decreased ER expression) after neoadjuvant chemotherapy,23-29 whereas others demonstrated a generally stable ER status or a status change that was not of statistical significance compared with the no chemotherapy group.11, 30-35 Unlike these studies, we focused on comparison between primary and metastatic tumors. Our data demonstrate that both the chemotherapy group and the no chemotherapy group showed very high and nearly identical concordance rates (92.5% and 92.7%, respectively), indicating that intervening chemotherapy is not responsible for discordance in ER status. A similar result has been reported in a study with 75 patients.22
Endocrine therapies that patients with an ER-positive and/or PR-positive breast carcinoma may receive include tamoxifen, aromatase inhibitors, other selective ER modulators, and the selective ER downregulator fulvestrant, depending on menopausal status and other medical factors. Tamoxifen is a nonsteroidal estrogen antagonist that interrupts the stimulatory effects of estrogen by binding to and blocking its receptor. Several earlier studies that used the ligand-binding assay to measure ER expression reported a significant fall in ER expression after endocrine manipulation, especially tamoxifen.11, 27, 33, 36, 37 This was initially attributed to selective elimination of ER-positive cells by endocrine therapy. Later it became clear that decreased ER content was primarily because of occupancy of ER by tamoxifen, which interfered with the ligand-binding assay. This phenomenon did not seem to occur with immunostaining methods.38 Several research groups then used immunostaining methods to examine ER expression stability in primary tumors treated with endocrine therapy and reported no significant difference in ER expression between pretherapy and posttherapy samples of the same tumors.39-41 However, Dawsett et al observed that, after 12 weeks of treatment with tamoxifen alone or tamoxifen plus anastrozole, ER H-score was significantly lower than baseline score.42
Data comparing ER status between primary and paired metastatic tumors with regard to endocrine therapy are limited and inconsistent. Johnston et al reported a significant reduction of ER expression in metastatic tumors of patients receiving prior tamoxifen treatment,38 whereas others found no significant association between endocrine therapy and ER discordance rate.6, 13, 22 Our data demonstrate that neither endocrine therapy nor the duration of endocrine therapy significantly affects ER discordance rate. We noticed that, in the patients with an ER-positive tumor who received endocrine therapy, ER status underwent a positive to negative conversion at a slightly higher rate in the tamoxifen-treated group than in the other groups. It is uncertain, however, whether this phenomenon is truly related to tamoxifen-induced biologic alteration or is a consequence of other factors. Lower et al reported that the change from ER positive to negative in metastasis occurred in approximately one-third of patients regardless of tamoxifen usage, indicating that tamoxifen alone did not play an important role.13
The mechanisms responsible for ER status discordance between primary and metastatic tumors could be multifactorial and complicated. Possible contributory factors include genetic drift during tumor progression and treatment-related clone selection.43, 44 Intratumoral heterogeneity may lead to a false-positive or false-negative result in a small biopsy sample.45 Furthermore, interlaboratory variation in testing method (ie, sample processing and staining) and interpretation variability may account for ER discordance in some patients.46-48 In our study, of the 17 patients with discordant ER status (Table 4), 15 had their primary tumor tested at an outside hospital, whereas the paired metastasis was tested at M. D. Anderson. In 7 patients, marginal positive status was found in 1 site of each tumor pair, which might be due to interpretation difficulty as such cases are often associated with high interobserver variability. Moreover, 4 of the 17 patients had 2 tumor nodules in the breast, but ER status was assessed only in 1 of the nodules. We cannot exclude the possibility that ER status in the 2 nodules differed and the metastasis was derived from the unsampled primary tumor nodule.
With the advantage of advanced imaging and radiological techniques, and our validated technique for ER testing on cytology samples,17 we were able to sample the majority of metastatic breast carcinomas via FNA with subsequent biomarker testing. Our data demonstrate that the ER status of the metastatic tumors measured on FNA samples was highly concordant with the status of the corresponding primary tumors, and direct smears performed equally well as cell block sections when used for ER staining.
ER expression level has been shown to correlate with response to endocrine therapy and prognosis.49, 50 Decreased ER expression in metastatic carcinoma is associated with shorter overall survival.13, 51 To determine whether the quantity of ER expression could change through disease progression, we compared the ER percentage between primary and metastatic tumors in 92 patients. By using 4 arbitrarily defined ranges, we found that the ER level of primary and metastatic tumors remained in the same range in 71 (77.2%) patients. Of the 21 pairs in which ER expression levels fell into different ranges, 17 (81%) pairs showed only a 1-step difference and only 5 had a true ER status change, with an ER status concordance rate of 94.6%. Some studies indicated that ER staining was essentially bimodal, either completely ER-negative or diffuse and uniformly ER-positive.52, 53 This “all or none” phenomenon was not evident in our study, however, because 16 (17.4%) primary tumors and 17 (18.5%) metastatic tumors had an ER value between 1%-49%.
This study demonstrates that ER status is generally stable through disease progression and is not significantly influenced by metastatic site, intervening interval, prior chemotherapy, or endocrine therapy. Variability of ER status between primary and metastatic tumors occurs only in a small proportion of patients, although the mechanisms responsible for the discordance of ER status are yet to be elucidated. Nevertheless, given the importance of ER status for clinical management, an effort should be made to retesting ER status in metastatic breast carcinoma. The recently published American Society of Clinical Oncology/College of American Pathologists guidelines for immunohistochemical testing of ER and PR in breast cancer also made this recommendation.54