Analysis of estrogen receptor (ER) and progesterone

Transcription

1 Original Articles Comparison of Estrogen and Progesterone Receptor Antibody Reagents Using Proficiency Testing Data Megan L. Troxell, MD, PhD; Thomas Long, MPH; Jason L. Hornick, MD, PhD; Abiy B. Ambaye, MD; Kristin C. Jensen, MD Context. Immunohistochemical analysis of estrogen receptor (ER) and progesterone receptor (PgR) expression in breast cancer is the current standard of care and directly determines therapy. In 2010 the American Society of Clinical Oncology and the College of American Pathologists (ASCO/CAP) published guidelines for ER and PgR predictive testing, encompassing preanalytic, analytic, postanalytic factors; antibody validation; and proficiency testing. Objective. To compare the performance of different antibody reagents for ER and PgR immunohistochemical analysis by using CAP proficiency testing data. Design. The CAP PM2 survey uses tissue microarrays of ten 2-mm cores per slide. We analyzed survey data from 80 ER and 80 PgR cores by antibody clone from more than 1200 laboratories. Results. Laboratories used the ER antibodies SP1 (72%), 6F11 (17%), 1D5 (3%), and the PgR antibodies 1E2 (61%), 16 (12%), PgR-636 (13%), PgR-1294 (8%) in While 63 of 80 ER cores (79%) were scored similarly using each of the 3 antibodies, there were significant differences for others, with SP1 yielding more positive interpretations. Four cores were scored as ER negative by more than half of the laboratories using 1D5 or 6F11, while SP1 produced positive results in more than 70% of laboratories using that antibody. Despite the greater variety of PgR antibody reagents and greater PgR tumor heterogeneity, 61 of 80 cores (76%) were scored similarly across the 4 PgR antibodies. Conclusions. Accurate ER and PgR testing in breast cancer is crucial for appropriate treatment. The CAP proficiency testing data demonstrate differences in staining results by ER clone, with SP1 yielding more positive results. (Arch Pathol Lab Med. 2017;141: ; doi: /arpa OA) Analysis of estrogen receptor (ER) and progesterone (PgR) expression in breast cancer began more than 30 years ago, and indeed these were the first widely applied prognostic and predictive biomarkers in cancer therapy. 1 Estrogen receptor expressing breast cancers tend to have better prognosis than ER-negative breast cancers. Even more importantly, patients with strongly ER-positive tumors derive substantial benefit from therapies inhibiting estrogen signaling. 1 These therapies are referred to as endocrine therapy, hormonal therapy, or antiestrogen therapy and include agents such as the selective ER Accepted for publication January 4, Published as an Early Online Release July 17, Supplemental digital content is available for this article at www. archivesofpathology.org in the October 2017 table of contents. From the Department of Pathology, Stanford University Medical Center, Stanford, California (Drs Troxell and Jensen); Biostatistics, College of American Pathologists, Northfield, Illinois (Mr Long); the Department of Pathology, Brigham and Women s Hospital, Boston, Massachusetts (Dr Hornick); the Department of Pathology, University of Vermont, Burlington (Dr Ambaye); and the Department of Pathology, Veteran s Affairs Medical Center, Palo Alto, California (Dr Jensen). The authors have no relevant financial interest in the products or companies described in this article. A subset of the data was presented in abstract form at the United States and Canadian Academy of Pathology annual meeting; March 15, 2016; Seattle, Washington. Reprints: Megan L. Troxell, MD, PhD, Department of Pathology, Stanford University Medical Center, 300 Pasteur Dr, Stanford, CA ( megant@stanford.edu). modulator tamoxifen, aromatase inhibitors (letrazole, anastrozole, exemestane), or, in young women, upstream inhibition of gonadotropin-releasing factor. 1 Progesterone receptor expression is regulated by estrogen, and PgR positivity indicates a functional ER signaling pathway. 2 Early ER and PgR biochemical assays required fresh tumor tissue and used radiolabeled ligand (so-called ligandbinding assays, dextran-coated charcoal assays); with the advent of monoclonal antibodies some of these biochemical assays were replaced by enzyme-linked immunosorbent assays (ELISAs), still requiring fresh tissue. 1,3 Currently, almost all ER and PgR testing is performed by immunohistochemical analysis using formalin-fixed, paraffin-embedded tissue sections, allowing morphologic correlation and semiquantitative scoring. 1 Polymerase chain reaction analysis of ER and PgR messenger RNA levels has been more recently used as part of multiplexed assays, but it does not allow direct histologic correlation. 4 6 Accurate ER and PgR testing is of utmost importance for correct patient therapy. In 2010 an expert panel convened by the American Society of Clinical Oncology and College of American Pathologists (ASCO/CAP) published guidelines for standardization of ER and PgR testing by immunohistochemistry. 1 These guidelines covered many facets of testing: preanalytic, analytic, and postanalytic, as well as initial antibody validation, ongoing quality assurance, and proficiency testing. 1 The guideline committee extensively reviewed the literature on ER and PgR testing, and established the threshold for positive interpretation of immunohistochemistry as 1% of tumor nuclei with positive 1402 Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al

2 Figure 1. Reported estrogen and progesterone receptor antibody clones use over time (% of laboratories). A, Estrogen receptor (ER). B, Progesterone receptor (PgR). For ER, *other includes EP1, rabbit polyclonal, ER-2-123þ1D5, other; for PgR, *other includes 312, 1A6, Hpra2þHpra3, PR88, rabbit polyclonal, SP2, Y85, other. staining, but additionally required reporting of semiquantitative results. 1 The committee also reviewed antibody reagents that had been well established as clinically relevant predictive markers in breast cancer, based on patient outcomes in published studies. 1,6 8 These monoclonal antibodies included 1D5, 6F11, ER.2.123þ1D5 (cocktail), and SP1 (rabbit monoclonal) for ER-a; and 1294, 1A6, and 312 for PgR. Additional widely applied PgR antibodies include PgR636, 16, and 1E2 (rabbit monoclonal). 1 The few single center studies directly comparing subsets of these antibodies have generally shown comparable results, or in some studies increased sensitivity of rabbit monoclonal antibody reagents. 6,9 24 We analyzed CAP proficiency testing data from more than 1200 laboratories and compared the performance of various ER and PgR antibody reagents on 80 breast cancers in a tissue microarray format. MATERIALS AND METHODS PM2 Survey Data from the CAP Predictive Marker (PM2) survey were retrospectively analyzed. The PM2 survey uses tissue microarray slides (TMAs). Each slide contains ten 2-mm breast cancer tissue cores arrayed in a set configuration, along with 1 tissue core for orientation. Beginning in 2014, all laboratories received unstained slides cut from the same TMA block. Two 10-core slides for ER and 2 10-core slides for PgR are included in each mailing, with 2 mailings per year (total 40 ER and 40 PgR cores per year). For this study, staining results from 2014 and 2015 were analyzed by antibody; data on clone utilization from a longer time period were reviewed ( ). Laboratories are instructed to stain the provided slides by using their clinically validated protocols (assuming a negative control is run and is appropriately negative) and score them according to the ASCO/CAP guidelines. The scoring form asks laboratories to score tissue cores as either negative or positive, and then further specify the percentage of positive tumor nuclei exhibiting staining in the following increments: no staining, less than 1%, 1% to 10%, 11% to 50%, and greater than 50%, with the first 2 categories representing negative interpretations. Intensity of staining is scored by laboratories as strong, intermediate, weak, or not applicable. Laboratories transmit their resultant scores to CAP; stained slides are not submitted for review. For purposes of proficiency testing, ER and PgR tissue cores are considered acceptable for grading if there is consensus of positive or negative interpretation among 80% of participant laboratories. Data on antibody clone, and some parameters of antibody preparation and antigen retrieval, are also queried with each survey mailing. Individual laboratory results as well as complete summary data for each survey mailing are sent to all participant laboratories after analysis. Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al 1403

3 TMA Core Table 1. Results by Estrogen Receptor Antibody Clone, Selected Tissue Cores a,b No. Negative/No. Total (%) 1D5 6F11 SP1 233 P Value 1D5 Versus 6F11 Pairwise P Value 1D5 Versus SP1 6F11 Versus SP1 Cores With.95% Concordance Among Laboratories for All Estrogen Receptor Clones, yet Statistically Significant Difference Between Clones /61 (96.7) 259/261 (99.2) 955/958 (99.7).02 c.16 c.03 c.29 c /61 (96.7) 262/262 (100) 951/955 (99.6).02 c.04 c.046 c.58 c /42 (95.2) 249/250 (99.6) 964/966 (99.8).006 c.06 c.01 c.50 c /35 (97.1) 264/267 (98.9) 1012/1014 (99.8).01 c.39 c.1 c.06 c /35 (0) 7/265 (2.6) 6/1014 (0.6).009 c..99 c..99 c /34 (2.9) 0/265 (0) 1/1009 (0.1),.001 c.11 c.06 c..99 c /35 (97.1) 256/267 (95.9) 1004/1011 (99.3),.001 c..99 c.24 c,.001 Cores With Statistically Significant Difference Between Estrogen Receptor Clones and,95% Concordance /53 (5.7) 6/237 (2.5) 5/849 (0.6).002 c.21 c.009 c.02 c d 2/34 (5.9) 19/146 (13.0) 36/603 (6.0).02 c.37 c..99 c /53 (9.4) 6/237 (2.5) 7/850 (0.8), c,.001 c /53 (9.4) 12/236 (5.1) 18/852 (2.1) /52 (9.6) 3/234 (1.3) 10/841 (1.2).002 c.006 c, c /61 (13.1) 16/258 (6.2) 18/940 (1.9), ,.001, /57 (21.1) 40/257 (15.6) 93/942 (9.9) /61 (86.9) 240/260 (92.3) 65/957 (6.8), ,.001, /60 (33.3) 32/252 (12.7) 27/939 (2.9),.001,.001,.001, /60 (8.3) 2/262 (0.8) 2/955 (0.2),.001 c,.001 c,.001 c.20 c /43 (7.0) 4/253 (1.6) 8/968 (0.8).007 c.07 c.009 c.59 c /43 (53.5) 69/252 (27.4) 80/970 (8.2),.001,.001,.001, /43 (55.8) 177/252 (70.2) 290/967 (30.0), ,.001, /40 (55.0) 109/243 (44.9) 180/930 (19.4), ,.001, /43 (97.7) 245/253 (96.8) 839/970 (86.5), c.03 c, /33 (57.6) 149/264 (56.4) 307/1003 (30.6), , /35 (100) 240/265 (90.5) 1007/1010 (99.7),.001 c.09 c..99 c,.001 c Abbreviation: TMA, tissue microarray. a See Supplemental Table 1 for all 80 cores. b Bolded values: Significant at P ¼.05 for 233, or significant at P ¼.017 for pairwise comparisons. c Fisher exact test. d A total of 378 laboratories reported no invasive carcinoma (histiocytoid lobular carcinoma). Statistical Analysis Statistical tests were performed on both ER and PgR for each unique mailing, specimen, and core. A v 2 test of association was used to compare interpretation (positive or negative) against probe. For ER, the 3 antibody clones tested were 1D5, 6F11, and SP1; and for PgR, the 4 antibody clones tested were 1E2, 16, PgR-636, and PgR If any cell for the 233 tables in ER or 234 tables in PgR had counts below 5, then significance was tested by using a Fisher exact test (denoted in Tables with a c ). A P value of.05 was considered significant. For those specimens and cores across ER and PgR with statistically significant associations between interpretation and antibody clone, multiple pairwise comparisons between the different clones were performed. To assess significance of the pairwise comparisons, a Bonferroni correction was used to correct the type I error level dividing by the total number of comparisons performed. For ER, the level of significance for pairwise comparisons is 0.05/3 and for PgR, the level of significance for pairwise comparisons is 0.05/6. If any cell in the pairwise comparison had counts below 5, then a Fisher exact test was performed (denoted in Tables with a c ). RESULTS PM2 Survey The CAP PM2 survey provides 40 tissues per year in a TMA for testing each of ER and PgR. Beginning in 2014, all laboratories subscribing to the PM2 survey received unstained test slides prepared from the same tissue block, allowing direct comparison of ER and PgR results on the same tissue over many laboratories. In 2014 and 2015, results were received from 1228 to 1440 laboratories in each of the different survey mailings. Changes in Antibody Clone Utilization Over Time Laboratory-reported application of the different ER and PgR clones changed over time from , as shown in Figure 1, A and B. For ER, utilization of clone 1D5 decreased from 70 of 417 laboratories in 2010, to 34 of 1380 in 2015 (from 17% to,3% of laboratories), while 6F11 and SP1 were increasingly used (6F11: from 54 of 417 [13%] in 2010, to 234 of 1380 [17%]; SP1: 283 of 417 [68%] to 989 of 1380 [72%]). Although only captured in the other category, the rabbit monoclonal antibody EP1 has also seen increasing use, from less than 1% to 7%. For PgR, PgR-636 decreased in utilization (from 100 of 418 [24%] to 180 of 1356 [13%]), while clone 16 increased by about 5% (32 of 418 to 170 of 1356), with 1E2 (rabbit monoclonal) and PgR-1294 increasing slightly over the 5- year time span (Figure 1, B). Survey and Study Scoring For the purposes of the study, we analyzed data from all 80 ER and 80 PgR tested cores, even though some did not achieve the 80% consensus among laboratories required for formal proficiency test grading. We reasoned that intratumoral heterogeneity within tissue cores would be randomly distributed among laboratories using different antibody clones, such that tissue cores not reaching consensus owing to heterogeneity should not show differences in results by 1404 Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al

4 Figure 2. Comparison of 3 estrogen receptor antibodies. A through C, Sample core 4. 1D5 and 6F11 show negativity, whereas SP1 shows positivity: (A) 1D5, (B) 6F11, and (C) SP1. D through F, Sample core10. Estrogen receptor expression is heterogeneous with all 3 antibodies: (D) 1D5, (E), 6F11, and (F) SP1 (original magnification 3400 [A through F]). clone, in contrast to lack of consensus due to analytic differences such as antibody clone. For ER, there were 50 graded positive cores and 24 graded negative cores. All 6 ungraded cores were scored as positive by most laboratories. For PgR, 43 positive and 20 negative cores were graded. There were 6 ungraded cores that most laboratories scored as negative, and 10 ungraded cores were scored as mostly positive. The remaining PgR core was scored as positive by 52% of laboratories (2015 PM2-08 core 1, positive in 711 of 1354). Results for Estrogen Receptor by Antibody Of 80 cores, 56 were scored similarly using each of the 3 most commonly used ER antibodies: 1D5, 6F11, and SP1. Twenty-four cores had statistically significant differences in the number (%) of laboratories reporting a negative interpretation across antibody clones. However, 7 of these Arch Pathol Lab Med Vol 141, October 2017 cores had greater than 95% positive or negative consensus among all laboratories for each clone despite the statistically significant differences; these are listed in Table 1 (top), and are not further discussed. Seventeen core samples had significant differences in number of laboratories reporting negative ER interpretations, based on clone applied and lower than 95% consensus, as listed in Table 1 and illustrated in Figure 2, A through F, and Figure 3. Pairwise comparisons between antibodies are summarized in Table 2 and include 14 cores with significant differences in interpretations between SP1 and 1D5, 12 between SP1 and 6F11, and 4 between 1D5 and 6F11. The full ER dataset is provided in Supplemental Table 1 (see supplemental digital content, containing 3 tables at in the October 2017 table of contents). ER and PgR Antibodies: Proficiency Testing Troxell et al 1405

5 Figure 3. Percentage of participant laboratories reporting a negative estrogen receptor interpretation. Tissue microarray cores with statistically significant difference in results between pairs of clones are shown. See supplemental data for all 80 cores. Table 2. Number of Cores With Statistically Significant Different Pairwise Results by Estrogen Receptor Clone (and,95% Consensus a ) No. of Cores With Statistically Different Estrogen Receptor Clone Comparison Results 1D5 versus 6F11 4 1D5 versus SP1 14 6F11 versus SP1 12 SP1 different from both 1D5 and 6F11 9 1D5 different from both SP1 and 6F11 4 Different for all 3 pairs of clones 2 a As in Table 1, bottom section only. We reviewed staining characteristics of ER-positive tumor samples with differences by antibody clone. Of the 17 cores with poor consensus and intended positive results, the rabbit monoclonal SP1 yielded fewer negative interpretations and thus more core tumor samples interpreted as ER positive. Interestingly, for 11 of these 17 cores, most laboratories using SP1 reported positive staining in greater than 50% of tumor nuclei in the tissue core. The staining intensity was generally reported as weak to moderate by most laboratories, across all clones. Strikingly, there were 4 cores scored as ER negative by half of the laboratories using 1D5 or 6F11, while SP1 produced mostly positive results (see Supplemental Table 2 for full scoring details). These include core in which fewer than 7% of laboratories reported negative staining with SP1, but 86% to 92% of laboratories reported negative results with 1D5 or 6F11 (Table 1). For this sample, 73% of SP1 laboratories reported greater than 50% of nuclei staining (Figure 2, A through C). For core , about 8% of laboratories reported negative results with SP1, but nearly one-quarter and onehalf of laboratories found the core to lack staining with 6F11 and 1D5, respectively (Table 1, Supplemental Tables 1 and 2). For this sample and the 2 other SP1-positive and 1D5- negative or 6F11-negative cores, only 4% to 15% of SP1 laboratories reported greater than 50% of nuclei staining, while 26% to 50% of SP1 laboratories reported 10% to 50% of nuclei staining (Supplemental Table 2), and many reported staining near the interpretative threshold (either,1% or 1% 10% positive nuclei, Supplemental Table 2). Results for Progesterone Receptor by Antibody Although PgR expression is more heterogeneous than ER expression in breast cancer, 51 of 80 cancer cores were scored similarly using each of the 4 most commonly used progesterone receptor antibodies: 1E2, 16, PgR-636, and PgR Among the 29 cores with statistically significant differences in the number (%) of laboratories reporting a negative interpretation, 10 cores nevertheless had more than 95% of laboratories reporting either a positive interpretation or a negative interpretation for each clone (Table 3, top); these samples are not further discussed. The full PgR dataset is provided in Supplemental Table 3. The remaining 19 cores had significant differences in number of laboratories reporting negative PgR interpretations by clone and lower than 95% consensus, as listed in Table 3 and illustrated in Figure 4. Many of these 19 samples were PgR-negative cancers or cancers with low percentage positive nuclei. In fact, fully 9 of the 19 cores were reported as being near the 1% interpretative threshold by more than half of the laboratories (laboratories reporting,1% or 1% 10% nuclei positive, across all clones, Table 3). The rabbit monoclonal 1E2 generated more positive interpretations in tissue that was generally negative for PgR with other antibody reagents (eg, cores , , , and ; Table 3, bottom) Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al

6 Table 3. Results by Progesterone Receptor (PgR) Antibody Clone, Selected Tissue Cores a,b Cores With.95% Concordance Among Laboratories for All Progesterone Receptor Clones, yet Statistically Significant Difference Between Clones No. Negative/Total (%) Significant Pairwise TMA Core 1E2 16 PgR636 PgR 1294 v P Value Tests (P,.00833) /710 (1.0) 1/142 (0.7) 6/160 (3.8) 0/132 (0).01 c None /708 (1.1) 2/138 (1.4) 7/159 (4.4) 2/132 (1.5).04 c 1E2v636 ¼ /786 (95.9) 146/147 (99.3) 189/190 (99.5) 136/138 (98.6).01 c None /788 (0) 0/148 (0) 2/192 (1.03) 0/143 (0).01 c 1E2v636 ¼.004 c /800 (99.9) 148/149 (99.3) 190/195 (97.4) 140/142 (98.6).002 c 1E2v636,.001 c /818 (99.0) 156/157 (99.4) 167/175 (95.4) 124/125 (99.2).002 c 1E2v636,.001 c /822 (99.3) 157/157 (100) 171/178 (96.1) 127/127 (100),.001 c 1E2v636,.001 c /815 (0.2) 2/156 (1.3) 3/177 (1.7) 0/126 (0).049 c None /808 (0.6) 4/152 (2.6) 5/173 (2.9) 0/127 (0).01 c 1E2v636 ¼.007 c /849 (98.9) 171/171 (100) 180/187 (96.3) 140/140 (100).003 c 1E2v636 ¼.007 c TMA Core Cores With Statistically Significant Difference Between Progesterone Receptor Clones and,95% Concordance No. Negative/Total (%) 1E2 16 PgR636 PgR 1294 % of Laboratories Scoring Near 1% Threshold (,1% and 1% 10% ) 234 P Value Significant Pairwise Tests (P,.00833) /708 (60.9) 125/142 (88.0) 154/161 (95.7) 120/130 (92.3) 50.0,.001 1E2v all, /704 (35.4) 33/141 (23.4) 53/158 (33.5) 50/132 (37.9) E2v16 ¼ /705 (83.8) 125/139 (89.9) 150/160 (93.8) 126/133 (94.7) 18.6,.001 1E2v636 ¼.001 1E2v1249 ¼ /714 (37.1) 44/141 (31.2) 76/162 (46.9) 47/130 (36.2) v636 ¼ /801 (98.1) 121/149 (81.2) 168/197 (85.3) 141/143 (98.6) 14.2,.001 c 1294v636,.001 c 1294v16,.001 c 1E2v636,.001 1E2v16, /798 (66.0) 122/149 (81.9) 166/197 (84.3) 126/142 (88.7) 50.7,.001 1E2v all, /802 (26.4) 43/150 (28.7) 80/197 (40.6) 43/143 (30.1) v1E2, /778 (83.8) 139/147 (94.6) 179/190 (94.2) 134/141 (95.0) 14.4,.001 1E2v all, /816 (35.7) 78/158 (49.4) 75/177 (42.3) 69/127 (54.3) 60.0,.001 1E2v16 ¼.001 1E2v636, /818 (26.8) 49/155 (31.6) 63/178 (35.4) 23/126 (18.3) v636 ¼ /794 (73.4) 137/152 (90.1) 160/169 (94.7) 119/125 (95.2) 33.0,.001 1E2v all, /820 (71.1) 127/159 (79.9) 129/177 (72.9) 106/127 (83.5) E2v1294 ¼ /815 (2.5) 12/157 (7.6) 16/177 (9.0) 3/127 (2.4) 43.1,.001 c 1E2v16,.001 1E2v636, /778 (1.2) 11/149 (7.4) 2/166 (1.2) 4/118 (3.4) 4.9,.001 c 1E2v16, v636 ¼.006 c /815 (37.3) 50/158 (31.6) 59/175 (33.7) 31/126 (24.6) E2v1249 ¼ /852 (94.4) 170/171 (99.4) 180/186 (96.8) 137/140 (97.9) c 1E2v16 ¼.005 c /843 (72.1) 153/171 (89.5) 173/186 (93.0) 126/138 (91.3) 47.6,.001 1E2v all, /818 (43.3) 75/163 (46.0) 115/179 (64.2) 67/132 (50.8) 75.0, v16, v1E2, /849 (92.9) 152/170 (89.4) 157/186 (84.4) 127/140 (90.7) E2v636,.001 Abbreviation: TMA, tissue microarray. a See Supplemental Table 3 for all 80 cores. b Note: Significant at P ¼.05 for 234, or significant at P ¼ for pairwise comparisons. c Fisher exact test. DISCUSSION High-quality and accurate testing of breast cancers for steroid receptor expression is essential in identifying patients for antihormonal therapy, which provides significant clinical benefit. 1 Immunohistochemical assays for ER and PgR expression are the current standard of care, yet these assays must be appropriately validated, reproducibly performed, correctly interpreted, and reliably reported on specimens that have been properly handled to ideally preserve both antigenicity and morphology. 1 ASCO/CAP guidelines for ER and PgR testing, published in 2010, brought attention to these issues and provided guidelines for ER and PgR testing. 1 Among many other elements, the guideline mandates include assay validation and revalidation with demonstration that positive ER or PgR categories are 90% concordant and negative ER or PgR categories are 95% concordant (in comparison with clinically validated assays), 1,25 as well as ongoing proficiency testing, at least twice a year, with satisfactory performance defined as 90% correct responses on graded challenges. 1 Proficiency testing data allow analysis of assay variables from large numbers of participant laboratories; in fact, the UK Neqas program first drew attention to problematic (false positive) staining with the PgR rabbit monoclonal SP2. 26 Here, we analyzed ER and PgR results by antibody clone from the CAP PM2 survey data from 2014 and 2015, including 80 tissues stained for each of ER and PgR in a tissue microarray format, reported by more than 1200 laboratories (.200,000 interpretations). Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al 1407

7 Figure 4. Percentage of participant laboratories reporting a negative progesterone receptor (PgR) interpretation. Tissue microarray cores with statistically significant difference in results between pairs of clones are shown. See supplemental data for all 80 cores. We found significant differences in the number of laboratories reporting positive results by antibody clone for 17 of 80 tested cores for ER, with the 3 most widely applied antibodies, all of which were considered by ASCO/ CAP to have been validated in prior clinical studies. In general, consensus among antibodies was good for intended negative samples, yet of the 17 ER samples with variance by antibody, 15 were intended as positive. The rabbit monoclonal SP1 resulted in greater numbers of positive interpretations, with concern for disproportionately discordant negative staining with 1D5, and to a lesser degree, 6F11. Prior published studies comparing antibody reagents have generally been performed in a single laboratory, in which titering, validation, staining, and scoring would have been conducted by the same individuals for each of the tested antibodies. 2,9 29 These studies have generally demonstrated either equivalent performance of 1D5, 6F11, and SP1, or have shown greater sensitivity with SP1 (Table 4). 2,9 29 There is some concern that new rabbit monoclonal antibodies could be overly sensitive, resulting in very low-expressing tumors being classified (and treated) as ER-positive rather than triple-negative tumors, especially given the 1% threshold for positive interpretation. 6,30 33 However, we are aware of 5 studies that compared SP1 and other ER antibodies in conjunction with clinical outcome data (Table 4). 13,22 25 Cheang et al 13 found that SP1 produced 8% more positive cases (2861 of 4116 positive with SP1; 2604 of 4124 positive with 1D5), and that the SP1 results correlated better with outcome and dextran-coated charcoal assay results, using previously frozen tissue samples. Welsh et al 23 also found that for cases scored as SP1 positive/1d5 negative, patient outcome correlated best with SP1 results. Two studies found outcome and/or cytosol-based assay (ligand binding or ELISA) results for the discordant cases to be intermediate between concordant-positive and concordantnegative cases. 21,22 Kornaga et al 24 found SP1 to have the poorest specificity as compared to ligand-binding assay in a study also including 6F11 and 1D5þER in ready-touse kits; however, in this study the ligand-binding assay had a poorer ability to predict breast cancer recurrence or death than immunohistochemistry assays. Our study compared data from a large number of different laboratories, using their routine protocols, and showed greater rates of ER positivity with SP1. It is worrisome that 17.5% (14 of 80) of tissues in this study were 1D5 /SP1þ, suggesting that a subset of patients might be falsely reported as ER negative. However, our survey data also reveal that 1D5 is applied by a relatively small proportion of laboratories and its use has decreased over time. We were unable to analyze factors influencing this trend, or informing antibody selection based on the design of the PM2 survey. Initial studies of the rabbit monoclonal antibody SP1 noted an 8-fold greater affinity than 1D5. 12 Others 6 have reported cleaner, more easily interpretable staining with fewer tumors in the low or intermediate positive range with SP1. Our ER findings parallel those of other proficiency testing programs with central slide review, different scoring schemes, and smaller numbers (UK Neqas, NordiQC, Canadian Immunohistochemistry Quality Control). For example, in the NordiQC program reported 0% to 54% optimal or good results for laboratories using 1D5, 37% to 61% with 6F11, 87% to 99% with SP1, and 49% to 88% with EP1 in surveys of up to 345 laboratories. 34,35 Likewise, in the UK Neqas program reported 52% to 86% pass rate for 6F11, 84% to 99% pass rate for SP1, and 88% to 99% pass rate for EP1, with overall participation of up to 270 laboratories. 36 For PgR, there was a greater variety of commonly used antibody clones, and fewer cores officially graded (given the expected PgR heterogeneity 3 ), yet the variety of 1408 Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al

8 Table 4. Literature Review of Hormone Receptor Antibody Clone Comparative Studies Source, y N Results Estrogen receptor Bevitt et al, D5 versus 6F11 concordance 50/55 (90.9%) F11 higher quickscore than 1D5 (8.4 versus 5.5) Kaufmann et al, D5 stains 35/68 (51.4%) breast carcinomas F11 stains 37/68 (54.1%) breast carcinomas Kaplan et al, Huang et al, Cheang et al, Arihiro et al, Brock et al, Rhodes et al, Nassar et al, Rocha et al, Bogina et al, Madeira et al, Bae et al, Grabau et al, Welsh et al, D5 versus 6F11 concordance 577/592 (97.5%) (9 cases ¼ 1D5 /6F11þ; 6 cases ¼ 1D5þ/6F11 ) 6F11 yielded higher stain intensity and less cytoplasmic staining Scoring Comments and Positivity Threshold Quickscore. 0 5% moderate or strong positive cutoff. 6F11 also stains 4/26 (15%) lung and 1/35 (3%) gastric. 10% positive cutoff. 2-lm sections /61 (100%) concordance SP1 versus 1D5 5% positive cutoff. This is initial study describing SP (TMA) 2861/4116 (69.5%) positive with SP1; 11% more 3þ results 2604/4124 (63.1%) positive with 1D5 (337 cases ¼ SP1þ/1D5 ; 77 cases ¼ SP1 /1D5þ) SP1 correlates better with DCC results and better independent prognostic factor 89 70/89 (79%) positive with 1D5 71/89 (80%) positive with 6F11; more 3þ intensity /508 (99.6%) concordance SP1 versus 1D5 (2 cases ¼ SP1þ/1D5 ) 1D5 has cytoplasmic staining in 13 cases SP1 has more intense nuclear staining 51 (TMA) 36/51 (71%) positive with 6F11 34/49 (69%) positive with SP1 Tissue frozen before formalin fixation and paraffin processing. 1% positive cutoff shown. Allred also scored. Allred scoring. Data from optimal conditions listed. 43 (TMA) 98.45% concordance 1D5 versus SP1 5% positive cutoff. Data from visual scoring. 24 (TMA) 15/24 (62.5%) positive with SP1 10% positive cutoff. 15/24 (62.5%) positive with 6F11 Tested different vendor s clones, 13/24 (54.2%) positive with 1D5 best of each listed. (2 cases both SP1þ/1D5 and 6F11þ/1D5 ) 66 a Concordance: SP1 versus 6F11 ¼ 61/66 (92.4%) SP1 versus 1D5 ¼ 58/66 (87.8%) 6F11 versus 1D5¼ 61/66 (92.4%) % positive cells SP1. 6F11. 1D5 2 cases ¼ SP1 /6F11þ and/or 1D5þ 3 cases ¼ 6F11 /SP1þ and/ or 1D5þ 8 cases ¼ 1D5 /SP1þ and/or 6F11þ 61 (TMA) Concordance: 59/61 (96.7%) 1D5 versus SP1 2 cases ¼ 1D5 /SP1þ Stronger intensity with SP1; similar % scores 1077 (TMA) 738/1077 (68.5%) positive with SP1; SP1 with more Allred scores ¼ 8 642/1077 (59.6%) positive with 1D5 634/1077 (58.9%) positive with 6F11 96 cases ¼ SP1þ/1D5 108 cases ¼ SP1þ/6F11 ; 4 cases ¼ SP1 /6F11þ 28 cases ¼ 1D5þ/6F11 ; 20 cases ¼ 1D5 / 6F11þ Survival of discordant cases intermediate between concordantþ and concordant (not statistically significant, and with variable therapy) 390 (TMA) 292/390 (75%) positive with SP1 266/390 (68%) positive with 1D5 (29 cases ¼ SP1þ/1D5 ; 3¼ SP1 /1D5þ) SP1þ/1D5 cases have outcome and LBA/EIA results intermediate between concordantþ and concordant 375 (TMA) 308/375 (82.1%) concordant SP1 versus 1D5 (61 cases ¼ SP1þ/1D5 ; 6 cases ¼ SP1 /1D5þ) Outcome of discordant cases follows SP1 result (non hormonally treated cohort) Study group favored low and intermediate ER expressors. Allred scoring. Allred scoring. 1-mm cores. 1% positive cutoff; same trend holds at 10% threshold. 0.6-mm cores. Data from Figure 2 also QIF data analyzed. Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al 1409

9 Table 4. Continued Source, y N Results Kornaga et al, EP1 studies Badve et al, Diorio et al, Diorio et al, Progesterone receptor Press et al, Arihiro et al, Kornaga et al, (TMA) Versus ligand-binding assay: PPV: 1D5 and ER ¼ 95.9%; 6F11 ¼ 95.9%, SP1 ¼ 95.7% NPV: 1D5 and ER ¼ 83.3%; 6F11 ¼ 52.4%, SP1 ¼ 70% Concordance between antibodies, j statistic: 1D5 and ER versus SP1 ¼ 0.90 (4 cases ¼ 1D5 and ER /SP1þ; 5 cases ¼ 1D5 and ER þ/sp1 ) 6F11 versus SP1 ¼ 0.79 (21 cases ¼ 6F11 /SP1þ) 1D5 and ER versus 6F11 ¼ 0.66 (2 cases ¼ 1D5 and ER /6F11þ; 41 cases ¼ 1D5 and ER-2-123þ/6F11 ) 176 (TMA) 167/176 (94.9%) concordant SP1 versus EP1; EP1 more 3þ intensity (5 cases ¼ SP1þ/EP1 ; 4 ¼ SP1 /EP1þ) 390 (TMA) 377/390 (96.7%) concordant SP1 versus EP1; EP1 more 3þ intensity (4 cases ¼ SP1þ/EP1 ; 9 ¼ SP1 /EP1þ) 130 then /130 (95.4%) concordant SP1 versus EP1 initial validation set 54/54 (100%) concordant SP1 versus EP1 verification set after changing conditions 523 (TMA) 519/523 (99.2%) concordant SP1 versus EP1 (3 cases ¼ SP1þ/EP1 ; 1 case ¼ SP1 /EP1þ) 59 (MTB) Compared to biochemical assay: 90% PgR636 concordance 88% PgR1294 concordance 75% 80% 1A6 concordance (different vendors) 89 52/89 (58%) positive PgR /89 (67%) positive clone 16; more 3þ intensity 532 (TMA) Versus ligand-binding assay: PPV: PgR1294 ¼ 74.4%; 16 and SAN27 ¼ 76.2%, 1E2 ¼ 76.6% NPV: PgR1294 ¼ 71.9%; 16 and SAN27 ¼ 77.4%, 1E2 ¼ 76.9% Concordance between antibodies, j statistic: PgR1294 versus 1E2 ¼ 0.78 (23 cases ¼ PgR1294 /1E2þ; 26 cases ¼ PgR1294þ/1E2 ) PgR1294 versus 16 and SAN27 ¼ 0.81 (19 cases ¼ PgR1294 /16 and SAN27þ; 28 cases ¼ PgR1294þ/16 and SAN27 ) 16 and SAN27 versus 1E2 ¼ 0.82 (23 cases ¼ 16 and SAN27 /1E2þ; 21 cases ¼ 16 and SAN27þ/1E2 ) Scoring Comments and Positivity Threshold Study of ready-to-use assays with different detection chemistry and counterstain; tamoxifen-treated cohort. Allred scoring. More ERþ/PgR cases with 6F11 (1.2%). 1% and Allred scoring by patient. Also compared EP1 to PharmDX and oncotype data (n ¼ 176). 10% positive cutoff. 1% positive cutoff shown. Allred also scored. Study of ready-to-use assays with different detection chemistry and counterstain; tamoxifen-treated cohort. Allred scoring. Detailed Kaplan-Meier survival curves and correlation with ER results by antibody clone. Abbreviations: DCC, dextran coated charcoal; EIA, enzyme immunoassay; ER, estrogen receptor; LBA, ligand-binding assay; MTB, multitumor blocks (strips of tumor); NPV, negative predictive value; PgR, progesterone receptor; PPV, positive predictive value; QIF, quantitative immunofluorescence; TMA, tissue microarray. a Includes 26 with low-intermediate expression. antibodies produced comparable results. Unlike ER, core tissue samples with PgR variance by antibody tended to be PgR-negative samples; further, many of these had reported results near the interpretative positive/negative threshold of 1% positive nuclei. UK Neqas proficiency testing previously drew attention to false-positive staining with thepgrclonesp2. 26 We found that the rabbit monoclonal antibody 1E2 returned fewer negative interpretations than the3otherwidelyusedpgrclones.wedidnotanalyze data for SP2, as it was used by less than 33 (2%) of surveyed laboratories. Anecdotally, there has also been some concern for weak 1E2 positivity in ER-negative highgrade carcinomas. Indeed, NordiQC survey run B20 noted some false-positive staining in tonsil with both 1E2 and SP2, while a prior run B18 reported good performance among clones 16, PgR 636, and 1E2. 35 UK Neqas reported comparable pass rates between 1E2 and PgR636; however, a prior run noted false-positive staining with 1E2, which was diminished by appropriate blocking. 37 Kornaga et al 38 recently extended their study to PgR antibodies, finding differences in 5-year survival prediction based on combinations of ER and PgR antibodies. The tissue microarray format of this survey and this study offer both advantages and limitations. This format allowed all survey subscribers to receive TMA slides prepared from the same tissue block. However, as in any tissue-based 1410 Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al

10 study involving large numbers of sections, it remains difficult to account for intratumoral heterogeneity as might be encountered through the depth of the thick tissue block. We noted that some cores had poor consensus among laboratories that correlated with antibody, while in a few other cores poor consensus appeared to be independent of antibody and could be attributable to sample heterogeneity (eg, 2014 PM2-06 No. 10; Figure 2, D through F). Also, core tumor samples near the interpretative threshold of 1% were often reported differently, especially for PgR. The small tumor sample size (2-mm cores), lack of correlative hematoxylin-eosin stained slides for participant scoring, and lack of central review are also limitations of this study, but are necessities of the large survey subscription. For ER and PgR immunohistochemistry, there is no currently accepted alternative method of analysis to provide a gold standard result. Thus, we analyzed pairwise comparison of antibody performance. However, we believe that these limitations do not alter our conclusions. While we did not further analyze technical variables, it is worth noting from the survey s supplemental questions that only 87% to 89% (1014 of 1161 to 1150 of 1292) of laboratories affirm that their criteria for reporting hormone receptor positive is 1% in In 2010, the ASCO/ CAP guidelines established the 1% threshold for positive interpretation. 1 Most laboratories reporting other scoring schemes use any staining of tumor cells (7% 8% of laboratories; 83 of 1292 to 91 of 1161), which would produce false-positive rather than false-negative results. Nevertheless, it is troubling that after 5 years, more than 10% of laboratories ( laboratories) participating in proficiency testing have not implemented this element of ASCO/ CAP guidelines. In conclusion, analysis of proficiency testing data provides useful comparative data regarding analytic variables in ER and PgR testing. We highlight the striking differences in ER results by antibody clone, with SP1 yielding more positive results than either 6F11 or 1D5. Further, we note that there may not be universal implementation of various elements of the ASCO/CAP 2010 guidelines, even among laboratories enrolled in proficiency testing. The authors thank Immunohistochemistry (IHC) committee members Dennis O Malley, MD, Neogenomics, Aliso Viejo, California, and Jinru Shia, MD, Memorial Sloan-Kettering Cancer Center, New York, New York, for providing additional immunostained slides for figures. Other IHC committee members include Andrew Bellizzi, MD, University of Iowa Hospital & Clinics, Iowa City; Patricia L. Kandalaft, MD, Pacific Pathology Partners, Seattle Washington; Sara Ann McClintock-Treep, MD, MT(ASCP), ThedaCare Regional Medical Center Pathology, Appleton, Wisconsin; Krzysztof Moroz, MD, Tulane University Medical School, New Orleans, Louisiana; Mohamed Salama, MD, University of Utah and ARUP Laboratories, Salt Lake City, Utah; Kurt Schaberg, MD, Stanford University, Stanford, California; Robert A. Schwartz, MD, Departments of Pathology at Waterbury Hospital, Waterbury, Connecticut, and Eastern Connecticut Health Network, Manchester, Connecticut; Qihui Jim Zhai, MD, Mayo Clinic, Jacksonville, Florida. The authors acknowledge the invaluable assistance of College of American Pathologists committee staff, notably Ross Owen, BS, and Jim Dvorak, BS. Norm Cyr, BA, medical and science illustrator, provided expert assistance with the figures. References 1. Hammond ME, Hayes DF, Dowsett M, et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer (unabridged version). Arch Pathol Lab Med. 2010;134(7):e48 e Press M, Spaulding B, Groshen S, et al. Comparison of different antibodies for detection of progesterone receptor in breast cancer. Steroids. 2002;67(9): Dowsett M. Estrogen receptor: methodology matters. J Clin Oncol. 2006; 24(36): Paik S, Shak S, Tang G, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med. 2004;351(27): Wallden B, Storhoff J, Nielsen T, et al. Development and verification of the PAM50-based Prosigna breast cancer gene signature assay. BMC Med Genomics. 2015;8: Calhoun BC, Collins LC. Predictive markers in breast cancer: an update on ER and HER2 testing and reporting. Semin Diagn Pathol. 2015;32(5): Yaziji H, Taylor CR, Goldstein NS, et al. Consensus recommendations on estrogen receptor testing in breast cancer by immunohistochemistry. Appl Immunohistochem Mol Morphol. 2008;16(6): Gown AM. Current issues in ER and HER2 testing by IHC in breast cancer. Mod Pathol. 2008;21(suppl 2):S8 S Bevitt DJ, Milton ID, Piggot N, et al. New monoclonal antibodies to oestrogen and progesterone receptors effective for paraffin section immunohistochemistry. J Pathol. 1997;183(2): Kaufmann O, Köther S, Dietel M. Use of antibodies against estrogen and progesterone receptors to identify metastatic breast and ovarian carcinomas by conventional immunohistochemical and tyramide signal amplification methods. Mod Pathol. 1998;11(4): Kaplan PA, Frazier SR, Loy TS, Diaz-Arias AA, Bradley K, Bickel JT. 1D5 and 6F11: an immunohistochemical comparison of two monoclonal antibodies for the evaluation of estrogen receptor status in primary breast carcinoma. Am J Clin Pathol. 2005;123(2): Huang Z, Zhu W, Szekeres G, Xia H. Development of new rabbit monoclonal antibody to estrogen receptor: immunohistochemical assessment on formalin-fixed, paraffin-embedded tissue sections. Appl Immunohistochem Mol Morphol. 2005;13(1): Cheang MC, Treaba DO, Speers CH, et al. Immunohistochemical detection using the new rabbit monoclonal antibody SP1 of estrogen receptor in breast cancer is superior to mouse monoclonal antibody 1D5 in predicting survival. J Clin Oncol. 2006;24(36): Arihiro K, Umemura S, Kurosumi M et al. Comparison of evaluations for hormone receptors in breast carcinoma using two manual and three automated immunohistochemical assays. Am J Clin Pathol. 2007;127(3): Brock JE, Hornick JL, Richardson AL, Dillon DA, Lester SC. A comparison of estrogen receptor SP1 and 1D5 monoclonal antibodies in routine clinical use reveals similar staining results. Am J Clin Pathol. 2009;132(3): Rhodes A, Sarson J, Assam EE, Dean SJ, Cribb EC, Parker A. The reliability of rabbit monoclonal antibodies in the immunohistochemical assessment of estrogen receptors, progesterone receptors, and HER2 in human breast carcinomas. Am J Clin Pathol. 2010;134(4): Nassar A, Norton CM, Lawson D, Cohen C. Image cytometric validation of breast carcinoma markers (ER, HER2 and MIB-1) using tissue microarrays: rabbit monoclonal vs. FDA-approved antibodies. Anal Quant Cytol Histol. 2010;32(4): Rocha R, Nunes C, Rocha G, Oliveira F, Sanches F, Gobbi H. Rabbit monoclonal antibodies show higher sensitivity than mouse monoclonals for estrogen and progesterone receptor evaluation in breast cancer by immunohistochemistry. Pathol Res Pract. 2008;204(9): Bogina G, Zamboni G, Sapino A, et al. Comparison of anti-estrogen receptor antibodies SP1, 6F11, and 1D5 in breast cancer: lower 1D5 sensitivity but questionable clinical implications. Am J Clin Pathol. 2012;138(5): Madeira KP, Daltoé RD, Sirtoli GM, et al. Comparison of immunohistochemical analysis with estrogen receptor SP1 and 1D5 monoclonal antibodies in breast cancer. Pathol Res Pract. 2012;208(11): Bae YK, Gong G, Kang J, et al. Hormone receptor expression in invasive breast cancer among Korean women and comparison of 3 antiestrogen receptor antibodies: a multi-institutional retrospective study using tissue microarrays. Am J Surg Pathol. 2012;36(12): Grabau DA, Bendahl PO, Rydén L, et al. The prevalence of immunohistochemically determined oestrogen receptor positivity in primary breast cancer is dependent on the choice of antibody and method of heat-induced epitope retrieval prognostic implications? Acta Oncol. 2013;52(8): Welsh AW, Harigopal M, Wimberly H, Prasad M, Rimm DL. Quantitative analysis of estrogen receptor expression shows SP1 antibody is more sensitive than 1D5. Appl Immunohistochem Mol Morphol. 2013;21(2): Kornaga EN, Klimowicz AC, Guggisberg N, et al. A systematic comparison of three commercial estrogen receptor assays in a single clinical outcome breast cancer cohort. Mod Pathol. 2016;29(8): Fitzgibbons PL, Murphy DA, Hammond ME, Allred DC, Valenstein PN. Recommendations for validating estrogen and progesterone receptor immunohistochemistry assays. Arch Pathol Lab Med. 2010;134(6): Ibrahim M, Dodson A, Barnett S, Fish D, Jasani B, Miller K. Potential for false-positive staining with a rabbit monoclonal antibody to progesterone receptor (SP2): findings of the UK National External Quality Assessment Scheme for Immunocytochemistry and FISH highlight the need for correct validation of antibodies on introduction to the laboratory. Am J Clin Pathol. 2008;129(3): Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al 1411

11 27. Badve S, Vladislav IT, Spaulding B, et al. EP1: a novel rabbit monoclonal antibody for detection of oestrogen receptor a. J Clin Pathol. 2013;66(12): Diorio C, Laberge S, Caron C, Provencher L, Hogue JC, Sanschagrin F. Validation of EP1 antibody clone for estrogen receptor immunohistochemistry in breast cancer. Appl Immunohistochem Mol Morphol. 2014;22(8): Diorio C, Furrer D, Michaud A. Validation of EP1 antibody clone for estrogen receptor immunohistochemistry for breast cancer. Anticancer Res. 2016; 36(1): Iwamoto T, Booser D, Valero V, et al. Estrogen receptor (ER) mrna and ERrelated gene expression in breast cancers that are 1% to 10% ER-positive by immunohistochemistry. J Clin Oncol. 2012;30(7): Deyarmin B, Kane JL, Valente AL, et al. Effect of ASCO/CAP guidelines for determining ER status on molecular subtype. Ann Surg Oncol. 2013;20(1): Gloyeske NC, Dabbs DJ, Bhargava R. Low ERþ breast cancer: is this a distinct group? Am J Clin Pathol. 2014;141(5): Sheffield BS, Kos Z, Asleh-Aburaya K, et al. Molecular subtype profiling of invasive breast cancers weakly positive for estrogen receptor. Breast Cancer Res Treat. 2016;155(3): Vyberg M, Nielsen S. Proficiency testing in immunohistochemistry experiences from Nordic Immunohistochemical Quality Control (NordiQC). Virchows Arch. 2016;468(1): NordiQC assessment runs B17 and B19. Accessed February 2, UK Neqas immunocytochemistry runs and www. ukneqas.org.uk/. Accessed February 2, 2016, and July 24, UK Neqas immunocytochemistry run Accessed February 2, Kornaga EN, Klimowicz AC, Guggisberg N et al. Evaluation of three commercial progesterone receptor assays in a single tamoxifen-treated breast cancer cohort. Mod Pathol. 2016;29(12): doi: /modpathol Arch Pathol Lab Med Vol 141, October 2017 ER and PgR Antibodies: Proficiency Testing Troxell et al