Comparison of two antibodies for immunohistochemical evaluation of epidermal growth factor receptor expression in colorectal carcinomas, adenomas, and normal mucosa

Comparison of Two Antibodies forImmunohistochemical Evaluation of Epidermal GrowthFactor Receptor Expression in Colorectal Carcinomas,Adenomas, and Normal Mucosa

Rohit Bhargava, M.D.1

Beiyun Chen, M.D., Ph.D.1

David S. Klimstra, M.D.1

Leonard B. Saltz, M.D.2

Cyrus Hedvat, M.D., Ph.D.1

Laura H. Tang, M.D., Ph.D.1

William Gerald, M.D., Ph.D.1

Julie Teruya-Feldstein, M.D.1

Philip B. Paty, M.D.3

Jing Qin, Ph.D.4

Jinru Shia, M.D.1

1 Department of Pathology, Memorial Sloan-Ket-tering Cancer Center, New York, New York.

2 Department of Medicine, Memorial Sloan-Ketter-ing Cancer Center, New York, New York.

3 Department of and Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York.

4 Biostatistics Research Branch, National Instituteof Allergy and Infectious Disease, National Insti-tutes of Health, Bethesda, Maryland.

Supported in part by Grant 2 P01 CA65930-05A2from the National Institutes of Health.

Rohit Bhargava’s current address: Magee-Wom-en’s Hospital, University of Pittsburgh MedicalCenter, Pittsburgh, Pennsylvania.

Address for reprints: Jinru Shia, M.D., Departmentof Pathology, Memorial Sloan-Kettering CancerCenter, 1275 York Avenue, New York, NY 10021;Fax: (212) 717-3203; E-mail: [email protected]

Received July 14, 2005; revision received October6, 2005; accepted October 26, 2005.

BACKGROUND. Immunohistochemical staining for epidermal growth factor recep-

tor (EGFR) has been used as a criterion for the selection of patients with colon

cancer for anti-EGFR therapy. Two antibodies, the PharmDx kit and the 31G7

clone, are used commonly for immunohistochemistry by various laboratories. No

comparative studies on the performance of these 2 antibodies are available.

METHODS. EGFR status was evaluated in 744 tissue microarray core samples from

primary and metastatic colorectal carcinomas, colorectal adenomas, and normal

colorectal mucosa with both the PharmDx kit and the clone 31G7 monoclonal anti-

bodies. The stains were compared for staining intensity by using an automated

image-analysis system. The intensity of positive staining (brown color) was measured

on a scale from 0 to 255.The staining intensity also was scored manually as 0, 1 �, 2 �,

and 3 �.

RESULTS. Statistically, the median staining intensities scored by the automated system

between the 2 antibodies did not differ significantly, although, within each category of

samples (normal, adenoma, carcinoma, and metastases), the PharmDx antibody

staining was slightly more intense than the clone 31G7 antibody staining. There was a

linear correlation between automated image-analysis and manual scoring categories.

The median automated image-analysis intensity scores for the 4 manual scoring

categories with the PharmDx kit were as follows: 0 staining, 67.5; 1 � staining, 75.5;

2 � staining, 89.6; and 3 � staining, 106.0. The median automated image-analysis

intensity scores for the 4 manual scoring categories with the clone 31G7 antibody were

as follows: 0 staining, 71.3; 1 � staining, 73.6; 2 � staining, 84.6; and 3 � staining, 99.1.

The classification of tumors as EGFR-negative (0 staining) or positive (1 �, 2 �, or

3 � staining) was concordant in 151 of 160 carcinomas (94.4%) with 2 antibodies using

manual scoring. Five samples (3%) that scored 1 � with the PharmDx kit antibody

scored 0 with the clone 31G7 antibody; whereas 4 samples (2.5%) that scored 1 � with

the clone 31G7 antibody scored 0 with the PharmDx kit antibody.

CONCLUSIONS. The EGFR expression results obtained by immunohistochemistry

using both the EGFR PharmaDx kit and the 31G7 clone were comparable. Either

antibody may be used for immunohistochemical detection of EGFR in colorectal

carcinomas. In addition, manual scoring had an excellent correlation with auto-

mated scoring. Cancer 2006;106:1857– 62. © 2006 American Cancer Society.

KEYWORDS: epidermal growth factor receptor, colon carcinoma, immunohisto-chemistry, Ventana clone 31G7 antibody, PharmDx kit Dako antibody.

The epidermal growth factor receptor (EGFR) is a 170-kD receptortyrosine kinase encoded by the c-erb-B (HER-1) protooncogene.1

It is expressed in various solid tumors, including colorectal, prostate,

1857

© 2006 American Cancer SocietyDOI 10.1002/cncr.21782Published online 10 March 2006 in Wiley InterScience (www.interscience.wiley.com).

head and neck, and lung cancers, and in certain nor-mal tissues.2 When it is bound by ligands, such asepidermal growth factor (EGF) and transforminggrowth factor-�, EGFR undergoes conformationalchanges that activate its intracellular tyrosine kinaseactivity, initiating autophosphorylation and down-stream signal transduction pathways.3 Activation canmediate a variety of cellular responses, including geneexpression, cell proliferation, and cell survival. Dys-regulation of the EGFR signaling pathway because ofEGFR overexpression, genetic aberrations, or othercauses leads to malignant transformation. Recentstudies have shown that EGFR expression is present inapproximately 60% to 80% of colorectal carcinomas,4,5

and the receptor has emerged as a rational target foranticancer therapy in these tumors.6,7

Cetuximab is the first EGFR inhibitor to receiveUnited States Food and Drug Administration approvalfor the treatment of colorectal cancer. Cetuximab is ahuman-mouse chimeric monoclonal antibody di-rected against the extracellular ligand-binding domainof EGFR.8 This drug competitively inhibits binding ofEGFR by EGF and transforming growth factor-�,thereby blocking downstream signal transductionpathways and arresting cell growth.9 Cetuximab is in-dicated for the treatment of advanced metastatic colo-rectal cancer that is resistant to irinotecan-based ther-apy and as a single agent in patients who cannottolerate irinotecan.10

Currently, immunohistochemical stain for EGFRis the method of patient selection for cetuximab ther-apy, and only patients with an EGFR positive colorec-tal carcinoma are eligible for the use of this drug.However, recent studies have shown that immunohis-tochemical expression of EGFR does not correlate withclinical response to cetuximab, and response to cetux-imab may be seen in patients who have EGFR-nega-tive colorectal carcinoma.11 Thus, the use of immuno-histochemistry in patient selection for the cetuximabtreatment has to be reevaluated. One important as-pect for the evaluation of immunohistochemistry istechnical, and the type of primary antibody used is ofsignificant technical importance.

The 2 most commonly used EGFR antibodies forimmunohistochemistry are the PharmDx kit antibody(DAKOCytomation, Carpinteria, CA) and the clone31G7 antibody (Ventana Inc., Tucson, AZ). To ourknowledge, no systematic assessment of the perfor-mance of the PharmDx Kit antibody versus the clone31G7 antibody is available in the English literature.Thus, in the current study, we evaluated the EGFRstatus in 744 microarray tissue core samples fromnormal colon, colonic adenomas, primary colon can-cers, and metastases using these 2 antibodies. The

staining intensity in the different colorectal tissueswas detected by 2 methods: an image-analysis systemand manual scoring. Then, a comparative analysis wasperformed on the staining intensity results obtainedfor the 2 antibodies.

MATERIALS AND METHODSTissue Samples and Construction of Tissue MicroarraySeven hundred forty-four tissue cores were derivedfrom 248 samples (3 cores per patient) of normalcolon (n � 49 samples), adenomas (n � 32 sam-ples), primary carcinomas (n � 139 samples), livermetastases (n � 6 samples), and lung metastases(n � 22 samples). The tissue microarrays were con-structed using 0.6-mm tissue cores as described pre-viously.12 A hematoxylin and eosin-stained sectionwas evaluated for the presence of specific diagnosis,and the area to be used for creation of the tissuemicroarray was marked on the slide and the donorblock. Three cores of different areas from a singleformalin fixed, paraffin embedded tissue block weresampled.

ImmunohistochemistryImmunohistochemical staining for EGFR was per-formed on the tissue microarray slides using the EGFRPharmDx kit (DAKO Cytomation) and the 31G7 clone(Vertana, Inc.) according to the manufacturers’ in-structions.

Quantitative Image AnalysisSlides were scanned with the ACIS II system (Chroma-Vision Medical Systems, Inc., San Juan Capistrano,CA), which combines automated slide digitization andimage analysis of immunohistochemically stainedsections. Immunostained tissue microarray sectionswere scanned using the � 10 objective, and a compos-ite digital image was produced. This image was di-vided into individual tissue cores using a grid. Theintensity score represents the maximal intensity foreach tissue core on the tissue microarray. The inten-sity of positive staining (brown color) is measured ona scale from 0 to 255. The minimum intensity score is50.

Manual ScoringAll samples were evaluated by a single pathologist.The stains were scored as 0 when there was nospecific membrane staining within the tumor andwere scored as positive when there was any stainingof tumor cell membrane above background level.The positive samples were classified further into1 �, 2 �, and 3 � staining based on their stainingintensity. The highest staining intensity of all tissue

1858 CANCER April 15, 2006 / Volume 106 / Number 8

cores from the same tumor was used as the finalimmunohistochemical result for that tumor. Exam-ples of negative and positive stains are illustrated inFigure 1.

Statistical AnalysesThe intensity score was taken as the average of 3intensities of positive staining for ACIS scoring. Thehighest staining intensity (of 3 scores) from the man-ual score was used as the final immunohistochemicalresult.

Correlation between the clone 31G7 antibody andthe PharmDx kit antibody was examined by using theSpearman correlation coefficient. The correlation be-tween the ACIS score and the manual score was as-sessed by using the Kruskal–Wallis statistic. The cor-relation between ordered contingency table wasassessed by using a Jonckheere–Terpstra statistic. Pvalues � .05 were considered significant.

RESULTSPharmDx Kit versus Clone 31G7: The ACIS II QuantitativeImage-Analysis SystemOf the 744 tissue cores, there were 147 cores from nor-mal colon, 96 cores from adenomas, 417 cores fromprimary colon carcinomas, 18 cores from liver metasta-ses, and 66 cores from lung metastases. The mean, me-dian, and standard deviation for the staining intensity ofboth antibodies among different diagnostic categoriesare shown in Tables 1 through 4. Staining with the DAKOantibody (PharmDx kit) was slightly more intense thanstaining with the Ventana antibody (clone 31G7) in eachcategory; however, the difference in staining intensitywas not statistically significant.

ACIS II Automated Scoring versus Manual Scoring forBoth AntibodiesThe slides that were stained with both the PharmDxkit antibody and the clone 31G7 antibody also were

FIGURE 1. Examples of epidermal growth factor expression are (A) negative staining, (B) 1 � staining intensity, (C) 2 � staining intensity, (D) 3 � staining

intensity.

EGFR Antibody Comparison in Colon Ca/Bhargava et al. 1859

scored manually. For the PharmDx kit antibody, allcountable tissue cores from adenomas and carcino-mas were included for automated and manual scoringcorrelation. For the clone 31G7 antibody, all countabletissue cores from normal colon, adenomas, and carci-nomas were included for automated and manual scor-ing correlation. Table 5 shows that, with the PharmDxkit antibody, the mean and median intensity scores ofthe tissue cores obtained by the ACIS II quantitativeimage-analysis system increased linearly according tomanual grading from 0 to 3 �. When the tissue coreswere graded 0 manually, the average ACIS II score wasonly 68.9; whereas, when the tissue cores were graded1 �, 2 �, and 3 �, the average ACIS scores increasedto 76.4, 90.6 and 109.4, respectively. Thus, the 2 scor-ing systems had a positive correlation (statistically

significant; P � .0001). A similar significant positivecorrelation between the ACIS II image system and themanual grading system also was observed with theclone 31G7 antibody (P � .0001). The mean and me-dian ACIS II scores for tissues that were graded man-ually for staining from 0 to 3 � are listed in Table 6.

PharmDx Kit versus Clone 31G7: Manual ScoringThe numbers of carcinoma samples that were scorednegative or positive manually by the 2 antibodies werecompared. Of 167 carcinoma samples that were in-cluded in this study, 160 were available for final com-parison between the 2 antibodies. The results wereconcordant in 151 tumors (94%). Five tumors (3%)scored 1 � with the PharmDx kit antibody and scored0 with the clone 31G7 antibody. Conversely, 4 tumors(2.5%) scored 1 � with the clone 31G7 antibody andscored 0 with the PharmDx kit antibody (Table 7).

Among the positively stained carcinoma samples,consistent with the results obtained with the ACIS IIimage system (Table 3), the PharmDx kit antibodyyielded a greater proportion of 2 � and 3 � results (81of 160 samples; 51%) compared with the clone 31G7

TABLE 1Staining Intensity According to the ACIS II Automated Image-Analysis System for Normal Colon Tissue

Variable Clone 31G7 PharmDx Kit

Mean 78.2 88.1Median 78.0 86.0Standard deviation 8.6 15.8

The ACIS II automated image-analysis system is from ChromaVision Medical Systems, Inc. (San Juan

Capistrano, CA), clone 31G7 is from Ventana Inc. (Tucson, AZ), and the PharmDx kit is from DAKO-

Cytomation (Carpinteria, CA).

TABLE 2Staining Intensity According to the ACIS II Automated Image-Analysis System for Adenoma






TABLE 3Staining Intensity According to the ACIS II Automated Image-Analysis System for Primary Colorectal Adenocarcinoma






TABLE 4Staining Intensity According to the ACIS II Automated Image-Analysis System for Metastatic Colorectal Carcinoma






TABLE 5Correlation between Manual Scoring and Automated Scoring with thePharmDx Kit

ACIS Score*

Manual Score (No. of Patients)

0 (n�26) 1� (n�68) 2� (n�72) 3� (n�148)

Mean 68.9 76.4 90.6 109.4Median 67.5 75.5 89.6 106Standard deviation 6.6 8.1 8.7 12.6

ACIS: the ACIS II automated image-analysis system.



Cytomation (Carpinteria, CA). Samples analyzed include all countable tissue cores from adenomas and

carcinomas.

* P � .0001; mean and median intensity scores obtained from the ACIS II automated image-analysis

system correlated positively with the manual scores.


antibody (45 of 160 samples; 28%) by manual grading.Conversely, the clone 31G7 antibody yielded a greaterproportion of 1 � results (93 of 160 samples; 58%)compared with the PharmDx kit antibody (58 of 160samples; 36%). Overall, however, there was significantconcordance between 0 and 3 � categorization per-formed by the 2 antibodies (P � .0001; Table 7).

DISCUSSIONThe primary objective of the current study was tocompare the performance of the 2 commonly usedEGFR antibodies for the assessment of EGFR expres-sion in colorectal tissues: the EGFR PharmDx kit andclone 31G7. Both antibodies identify an epitope in theextracellular domain of the EGFR molecule. Such astudy is necessary to promote our understanding ofthe utility of immunohistochemistry in selecting pa-tients for treatment with cetuximab. It has been rec-ognized that immunohistochemistry is influenced by

a variety of factors, including tissue processing andhandling, the specificity and sensitivity of the anti-body, and the criteria in scoring the stain. The impor-tance of standardized tissue processing and handlingin immunohistochemistry for EGFR has been pointedout by some investigators.13,14 It has been shown thataltered protein expression may result from prolongedstorage time of tissue samples and may allow certaincatalytic degradation of cell surface receptors.13,14 Theimportance of the antibody and the scoring system ofthe stain in immunohistochemistry are obvious. How-ever, such issues remain to be resolved for EGFR im-munohistochemistry. Two different types of EGFR an-tibodies currently are used commonly amongdifferent laboratories. Although it has been indicatedthat positive EGFR staining of any intensity in �1% oftumor cells by the EGFR PharmDx kit is sufficient forthe eligibility of cetuximab therapy, it is not knownwhether the other type of EGFR antibody, the clone31G7 antibody, would have the same indication.

Thus, we systematically analyzed the performanceof the 2 major EGFR antibodies in a large number ofcolorectal tissue samples. We also compared an auto-mated scoring system with manual scoring. Overall,our results showed that the EGFR PharmDx kit fromDAKO consistently stained colorectal tissues more in-tensely than the 31G7 clone antibody from Ventana.However, statistically, the difference in the intensitybetween the 2 antibodies was insignificant. A goodcorrelation was achieved between the staining inten-sities obtained by the 2 antibodies. In addition, ourdata showed that the ACIS II image-analysis systemwas a reliable tool for scoring EGFR staining in colo-rectal tissue.

The slightly more intense EGFR staining with thePharmDx kit was observed in all colorectal tissues andin all positive stains, whether it was 1 �, 2 �, or 3 �.Consequently, it did not affect the linear correlationbetween the staining intensities of the 2 antibodies,and it did not alter the final manual grading in mostinstances. However, we did observe a few colorectalcarcinomas that were scored negative with thePharmDx kit but were scored 1 � positive with theclone 31G7 antibody (3%), or vice versa (2.5%). Theo-retically, such a discrepancy between the 2 antibodiesmay have accounted for the lack of a response tocetuximab in some patients; i.e., approximately 2.5%of patients who had tumors that were classified asEGFR-positive by the PharmDx kit antibody and, thus,received the anti-EGFR drug, in fact, had EGFR-nega-tive tumors and should not have responded to thedrug. However, clinical trials to date have demon-strated a response rate of only 9% to 23% to cetuximabin patients with colorectal cancer.15,16 Thus, the small

TABLE 6Correlation between Manual Scoring and Automated Scoring with theClone 31G7 Antibody

ACIS Score*

Manual Score (No. of Patients)

0 (n�27) 1� (n�148) 2� (n�48) 3� (n�10)

Mean 71.4 75 84.4 105.7Median 71.3 73.6 84.6 99.1Standard deviation 3.8 6.4 6.1 21.1

ACIS: ACIS II automated image-analysis system.



Cytomation (Carpinteria, CA). The samples analyzed included all countable tissue cores from normal

mucosa, adenomas, and carcinomas.

* P � .0001; mean and median intensity scores obtained from the ACIS II automated image-analysis

system correlated positively with the manual scores.

TABLE 7Correlation of Manual Grading Scores between the PharmDx Kit andthe Clone 31G7 Antibody in Colonic Adenocarcinomas

Clone 31G7Score

PharmDx Kit Score

0 1� 2� 3� Total

0 17 5 0 0 221� 4 53 35 1 932� 0 0 24 13 373� 0 0 0 8 8Total 21 58 59 22 160




P � .0001; P values were calculated after combining the 2� and 3� staining categories.

EGFR Antibody Comparison in Colon Ca/Bhargava et al. 1861

frequency of discrepant immunohistochemical stain-ing results between the 2 antibodies does not explainthe lack of a response in most cetuximab nonre-sponders. It also does not explain the presence of aresponse in 25% of patients with EGFR-negative tu-mors.11 In contrast, among the categories of intenselystained tumors, a score of 2 � or 3 � with 1 antibodynever resulted in a score of 0 with the other antibody(Table 7).

In the current study, we used the ACIS II quanti-tative image-analysis system for the evaluation ofEGFR staining and observed a direct correlation be-tween those results and the results from manual scor-ing. A linear increase was observed in the medianstaining intensity with the ACIS system from 0 to 3 �by manual scoring. Thus, although the ACIS softwaredoes not distinguish normal tissues from tumor tis-sues, with the ability to adjust the dynamic range,such a system may provide an efficient modality foranalyzing large quantities of tissue samples in a re-search setting. The main advantages of automatedimage analysis are the improved reproducibility ofresults, which limits interobserver variability,17 andhigh accuracy in measuring intensity.17 For our study,we did not analyze interobserver variations specifi-cally in the manual scoring of EGFR immunohisto-chemistry slides, and all samples were analyzed man-ually by a single pathologist. However, the finding thatthe scores produced by the image system correlatedpositively with the manual results suggests that man-ual interpretation and interobserver variability maynot be significant technical issues in the analysis ofEGFR immunohistochemistry. With regard to accu-racy in measuring intensity, because there is no cur-rent “gold standard” with which to judge the accuracyof EGFR immunohistochemistry results, it will be dif-ficult to compare the superiority of imaging systemversus manual scoring. With regard to HER-2, thereare observations suggesting that imaging systemscores were correlated better with HER-2 amplifica-tion by fluorescent in situ hybridization than manualscores.18

In summary, our results suggest that immunohis-tochemistry using both the EGFR PharmDx kit and the31G7 clone yielded comparable results. We concludethat either antibody may be used to analyze EGFRexpression in colon carcinoma. The semiquantitativemanual scoring correlated well with the results ob-tained through the automated image-analysis system.

REFERENCES1. Cohen S, Ushiro H, Stoscheck C, Chinkers M. A native

170,000 epidermal growth factor receptor-kinase complex

from shed plasma membrane vesicles. J Biol Chem. 1982;257:1523-1531.

2. Salomon DS, Brandt R, Ciardiello F, Normanno N. Epidermalgrowth factor-related peptides and their receptors in humanmalignancies. Crit Rev Oncol Hematol. 1995;19:183-232.

3. McCune BK, Earp HS. The epidermal growth factor receptortyrosine kinase in liver epithelial cells. The effect of ligand-dependent changes in cellular location. J Biol Chem. 1989;264:15501-15507.

4. Goldstein NS, Armin M. Epidermal growth factor receptorimmunohistochemical reactivity in patients with AmericanJoint Committee on Cancer Stage IV colon adenocarcinoma:implications for a standardized scoring system. Cancer.2001;92:1331-1346.

5. Nicholson RI, Gee JM, Harper ME. EGFR and cancer prog-nosis. Eur J Cancer. 2001;37(Suppl 4):S9 –S15.

6. Ciardiello F, Bianco R, Damiano V, et al. Antitumor activityof sequential treatment with topotecan and anti-epidermalgrowth factor receptor monoclonal antibody C225. ClinCancer Res. 1999;5:909-916.

7. Ciardiello F, Tortora G. Epidermal growth factor receptor(EGFR) as a target in cancer therapy: understanding the roleof receptor expression and other molecular determinantsthat could influence the response to anti-EGFR drugs. Eur JCancer. 2003;39:1348-1354.

8. Baselga J. The EGFR as a target for anticancer therapy—focus on cetuximab. Eur J Cancer. 2001;37(Suppl 4):S16 –S22.

9. Harding J, Burtness B. Cetuximab: an epidermal growthfactor receptor chemeric human-murine monoclonal anti-body. Drugs Today (Barcelona). 2005;41:107-127.

10. Starling N, Cunningham D. Second-line therapy for advancedcolorectal carcinoma. Curr Oncol Rep. 2005;7:173-180.

11. Chung KY, Shia J, Kemeny NE, et al. Cetuximab showsactivity in colorectal cancer patients with tumors that do notexpress the epidermal growth factor receptor by immuno-histochemistry. J Clin Oncol. 2005;23:1803-1810.

12. Hoos A, Urist MJ, Stojadinovic A, et al. Validation of tissuemicroarrays for immunohistochemical profiling of cancerspecimens using the example of human fibroblastic tumors.Am J Pathol. 2001;158:1245-1251.

13. Atkins D, Reiffen KA, Tegtmeier CL, Winther H, Bonato MS,Storkel S. Immunohistochemical detection of EGFR in par-affin-embedded tumor tissues: variation in staining inten-sity due to choice of fixative and storage time of tissuesections. J Histochem Cytochem. 2004;52:893-901.

14. Erlichman C, Sargent DJ. New treatment options for colo-rectal cancer. N Engl J Med. 2004;351:391-392.

15. Cunningham D, Humblet Y, Siena S, et al. Cetuximab mono-therapy and cetuximab plus irinotecan in irinotecan-refractorymetastatic colorectal cancer. N Engl J Med. 2004;351:337-345.

16. Saltz LB, Meropol NJ, Loehrer PJ Sr., Needle MN, Kopit J,Mayer RJ. Phase II trial of cetuximab in patients with refrac-tory colorectal cancer that expresses the epidermal growthfactor receptor. J Clin Oncol. 2004;22:1201-1208.

17. Hilbe W, Gachter A, Duba HC, et al. Comparison of auto-mated cellular imaging system and manual microscopy forimmunohistochemically stained cryostat sections of lungcancer specimens applying p53, ki-67 and p120. Oncol Rep.2003;10:15-20.

18. Luftner D, Henschke P, Kafka A, et al. Discordant resultsobtained for different methods of HER-2/neu testing inbreast cancer—a question of standardization, automationand timing. Int J Biol Markers. 2004;19:1-13.


Documents

Comparison of two antibodies for immunohistochemical evaluation of epidermal growth factor receptor expression in colorectal carcinomas, adenomas, and normal mucosa