Quality Control in Immunocytochemistry: Experiences with the Oestrogen Receptor Assay J Clin Pathol: First Published As 10.1136/Jcp.45.2.120 on 1 February 1992

Home , Enzyme assay

120 J Clin Pathol 1992;45:120-124 Quality control in immunocytochemistry: Experiences with the oestrogen receptor assay J Clin Pathol: first published as 10.1136/jcp.45.2.120 on 1 February 1992. Downloaded from

F T Bosman, A F P M de Goeij, M Rousch

Abstract ted specificity to be produced, and the develop- Aims: To evaluate the feasibility of an ment of very sensitive second step reagents. As interlaboratory quality control pro- a consequence, almost every surgical pathology gramme in immunohistochemistry. laboratory routinely uses immunocytochemistry on a daily basis. As recently Methods: Several pathology laboratories ' were asked to carry out immunohisto- indicated by Elias et al, however, an important chemical oestrogen receptor staining on a source of variability in the performance of set of freeze dried cryostat sections of immunocytochemistry techniques is the vari- breast cancer tissue. The sections and ation among laboratories in tissue processing protocols for staining and semi- and histochemical procedures. This could be quantitative scoring were mailed to the improved by standardisation, monitored by participating laboratories in two trials. quality control programmes. The oestrogen receptor content of the A typical example of an immunocyto- breast cancer samples was determined by chemical test which requires such an approach radioligand binding assay on the tumour is the evaluation of the oestrogen receptor cytosol. content in breast cancer. About 60% of Results: In the first trial 11 laboratories patients with an oestrogen receptor positive (response rate 60%) participated. Eight tumour will respond to some form ofendocrine (73%) of the participants scored within a treatment; less than 10% of oestrogen receptor 95% confidence interval and all but one negative neoplasms respond.23 Expression of correctlyclassified thetumour as receptor oestrogen receptor protein may also be regar- positive. In the second trial all 20 partici- ded as a hallmark of a differentiated state,4 and pating laboratories (response rate 55%) many studies have shown that the prognosis in correctly scored one tumour sample as terms of disease free interval and survival in negative and 18 of them (90% of respon- oestrogen receptor positive breast cancer is dents) correctly classified the two other more favorable than that of oestrogen receptor http://jcp.bmj.com/ tumour samples as receptor positive. In a negative breast cancer.5 quantitative evaluation a histochemical In order for it to be clinically useful a reliable score within 95% confidence interval technique for oestrogen receptor determina- limits was provided by eight (40%) and 12 tion is required. For two decades now, the (60%) of the participants. standard steroid receptor assay has been the of dextran coated charcoal (DCC) assay, used on Conclusions: Semiquantitative scoring on September 28, 2021 by guest. Protected copyright. immunocytochemical staining is valu- cytosols of tumour tissue. Although reliable in able for performing correlative inter- principle, this ligand binding assay has been laboratory studies, although this scoring shown to lead to high inter- and intra-assay protocol may not be required for diag- variability. Only the introduction of quality nosis or prognosis. Significant inter- control programmes in the clinical chemistry laboratory variability exists, leading to laboratories in Europe, under the auspicies of qualitatively correct receptor classifica- the EORTC has led to a steroid receptor assay tion in 100% ofreceptor negative and 80% with reasonable widescale reproducibility.67 of receptor positive cases, and quanti- Replacement of this technique with an tative agreement in only about half of the immunocytochemical assay, which would be cases. The perceived variability is not preferable in view of the lesser quantity of caused by systematic differences in the tissue required, avoidance of the use of choice of the immunocytochemical tech- radiolabelled ligands, the possibility oflocating of freeze dried the receptor protein specifically in tumour cells, nique, or the mailing in Department of sections. Quality control programmes and the ability to take tumour heterogeneity Pathology, should be included in the standard pro- receptor content into account have all been University of immunohisto- emphasised repeatedly."'0 The development of Limburg, cedures of each diagnostic allowed steroid Medical Faculty, chemistry laboratory. monoclonal antibodies has PO Box 616, receptor assays based on antigen-antibody 6200 MD Maastricht, interactions to be introduced." It is clear, The Netherlands that immunocytochemical receptor F T Bosman Over the past decade immunocytochemistry however, A F P M de Goeij has become essential in surgical pathology. The determination can only be justified if: (a) the M Rousch methodological requirements of repro- results of the immunocytochemical assay have Correspondence to: ducibility and specificity have been met with the been validated against those of the ligand Dr A F P M de Goeij technique, which allows large binding assay; and (b) similar quality control Accepted for publication hybridoma validate the intra- and inter- 3 February 1991 quantities ofmonoclonal antibodies with selec- programmes Quality control in oestrogen receptor immunocytochemistry 121

laboratory reproducibility of the technique. The sections were weakly counterstained in The first requirement has repeatedly been diluted haematoxylin and mounted in Entellan.

met.41213 This study was designed to test the After the sections had been returned 11 of J Clin Pathol: first published as 10.1136/jcp.45.2.120 on 1 February 1992. Downloaded from feasibility of a quality control programme to the 20 laboratories indicated that they had used meet the second requirement. the ERICA staining protocol, as outlined in the instructions supplied with the commercial kit (ERICA, Abbott Laboratories). One partici- Methods pant indicated that the protocol used was a local TISSUES AND MAILING OF THE SECTIONS modification ofthe ERICA protocol. In the first Samples of breast cancer tissue were snap- pilot study two participants used the ERICA frozen in isopentane, cooled with dry ice. (Abbott) protocol. Tumours with a high, intermediate, or zero receptor content, as determined by radio- SEMIQUANTITATIVE ANALYSIS chemical (DCC) assay on cytosol, were selec- The stained sections were scored semiquanti- ted. Frozen sections (6.um) were cut on a tatively, according to a modification of the cryostat microtome and mounted on gelatin scoring by McCarty et al,15 as described chromealum coated slides. The slides were earlier.'6 The resulting histochemical score was freeze dried and subsequently stored in a obtained as follows: plastic airtight container over silica gel at room i=4 Histochemical score = I P(i) x i temperature. Each participating laboratory i=O received four sections of each tumour and one where i = staining intensity, which may vary section of a rabbit uterus which served as a between 0 (no staining) and 4 (strongest stain- receptor positive standard tissue. This was ing), and Pi = percentage of stained tumour found to be an acceptable approach after it was cell nuclei in category i (0-100). The maximum shown that when kept absolutely dry, the attainable score is 400 by definition. For reliable sections could be kept at ambient temperatures visual scoring at least three cohorts of 100 for up to several weeks without deterioration of tumour cells had to be scored in different high the immunoreactivity of oestrogen receptor power fields (objective x 40). protein. A receptor score ofless than 35 was regarded In the first trial 11 pathology laboratories as negative, based on earlier correlative participated, 20 in the second trial. This studies'6 between immunohistochemical recep- represented a response rate of 60% and 55%, tor assays and radioligand binding assays. In respectively. these studies a cut-offvalue of 10 fmol oestrogen receptor/mg cytosol protein was used. IMMUNOHISTOCHEMISTRY Together with the mailing of the sections a OESTROGEN RECEPTOR ANALYSIS IN CYTOSOL staining protocol was provided that differed TISSUE from the protocol included in the ERICA kit Oestrogen receptors were quantitated in http://jcp.bmj.com/ with respect to the working dilution ofthe anti- tumour cytosols, prepared from a tumour sam- oestrogen receptor serum (1 in 4 instead of ple directly adjacent to the sample used for undiluted) and to the conditions of incubation with this antibody (overnight at 4°C instead of one hour at room temperature). Before immunostaining, the sections were fixed (10 minutes at 4°C) in picric acid-para- 400- on September 28, 2021 by guest. Protected copyright. formaldehyde'4 or in phosphate buffered 4% paraformaldehyde (pH 7 4). Subsequently the sections were rinsed (three times for five I minutes at room temperature) in phosphate 300 buffered saline (PBS). Before incubation the sections were exposed to 20% normal goat I serum minutes at room .o0 (15 temperature). After i- T U blotting the section, the primary antibody 0 7nSW-I U_ _ Uw T (ERICA kit, kindly provided by Abbott 0 -w-B---- U) I1 Laboratories, Diagnostic Division, The I Netherlands) was applied (diluted 1 in 4 in 1% U bovine serum albumin in PBS) and the sections 100- were incubated overnight in a humid chamber at 4°C. Subsequently the sections were washed in PBS (three times for five minutes each) and 0 S incubated (30 minutes at room temperature) I I I I with goat-anti rat IgG (included in the ERICA 0 5 10 kit or from Dako). After washing in PBS (three Participant times for five minutes each) the sections were Figure I Semiquantitative immunohistochemical scoring incubated (30 minutes at room temperature) ofoestrogen receptor staining in cryostat sectionsfrom a human breast cancer sample distributed to 11 with undiluted rat peroxidase-antiperoxidase participating laboratories. complex (included in the ERICA kit or from Scores ofthisfirst trial are presented as mean (SD) score Dako). After a final wash the immunoreactivity calculated over three tumour cell cohorts. The dashed was lines indicate the 95% confidence interval. Participants 4 visualised in a diaminobenzidine-H202 and 5 used the Abbott protocol, the other laboratories used mixture (seven minutes at room temperature). the Maastricht protocol. 122 Bosman, de Goeij, Rousch

Table I Overall results (second test)

Sections histochemical score* Cytosol (fmol receptor/mg protein) J Clin Pathol: first published as 10.1136/jcp.45.2.120 on 1 February 1992. Downloaded from Total Maastricht Abbott Sample (n = 20) (n = 9) (n = 10) DCC EIA 1 0 0 0 Negative Negative 2 145 (67) 163 (64) 148 (74) 86 72 3 180(90) 211 (100) 164(71) 103 128 *Results include one laboratory that used a different staining protocol.

cryostat sectioning, using both radiochemical The results of the immunohistochemical, and immunochemical assays. The radio- radiochemical, and immunochemical oestrogen chemical receptor determination was done with receptor assays of the second test are summar- a standard DCC assay according to the ised in table 1. One tumour sample was guidelines of the EORTC. 7 The immuno- oestrogen receptor negative by DCC assay as chemical oestrogen receptor assay was carried well as by enzyme immunoassay and was out with the enzyme immunoassay kit obtained correctly scored negative by all laboratories. from Abbott Laboratories according to the The other two tumour samples contained manufacturer's instructions.'3 Both types of oestrogen receptor by both cytosol techniques assay were performed in the laboratory of Dr and both immunohistochemistry protocols. On Th Benraad of the University of Nijmegen, average, sample 3 yielded higher histochemical Academic Hospital, while participating in a scores than sample 2, which corresponded with steroid receptor assay quality control the results of the cytosol assay. The slides programme in Europe under the auspices ofthe stained with the Maastricht staining protocol EORTC.6 tended to show somewhat higher histochemical scores than those of the Abbott protocol, although the differences were not significant Results (p < 0-0001). One participant (15) scored sig- In the first pilot study a single sample ofbreast nificantly lower than the other using a modified cancer tissue was distributed to 11 pathology staining protocol. laboratories. The mean histochemical score Strikingdifferences in immunohistochemical (with standard deviations), calculated over the results occurred (figs 2 and 3). Figure 2 shows three scored tumour cell cohorts for each photomicrographs of the two tumour samples, participant is shown in fig 1. An average histo- stained intensely in one laboratory (2A and C) chemical score of210 was obtained on a sample and only weakly in another (2B and D). This is

with a cytosolic oestrogen receptor content of quantitatively represented in fig 3, which shows http://jcp.bmj.com/ 519 fmol/mg protein. Eight of the 11 partici- the plots of the histochemical scores. For pants scored within a 95% confidence interval. sample 2, the histochemical scores varied bet- All except one laboratory correctly classified the ween 0 and 280 (fig 3A) and for sample 3 tumour as receptor positive. between 10 and 330 (fig 3B). Table 2 shows that on September 28, 2021 by guest. Protected copyright. Figure 2 Illustration of immunohistochemical staining ofhuman breast cancer samples by different pathology laboratories. (A) and (C): laboratory 1, samples 2 and 3; (B) and (D): laboratory 2, samples 2 and 3.

.%;.....rs i O... sl "I 4t

s* Quality control in oestrogen receptor immunocytochemistry 123

Figure 3 400n 400 Semiquantitative immunohistochemical scoring of oestrogen T J Clin Pathol: first published as 10.1136/jcp.45.2.120 on 1 February 1992. Downloaded from receptor staining in I cryostat sectionsfrom two 300- 300 human breast cancer .a. T samples, distributed to 20 1 0 participating laboratories. 0 * IT * T Scores of this second trial In 0 ____ T are presented as mean it(a -or T (SD) score. (A) sample 2 *ff 200- 200 a T n and (B) sample 3. 0) * T Participants 2, 5, 6, 8, 10, I ~ 11, 13, 16, 17, 18 and 19 s~~~ used the Abbott protocol, - i participant 15 its own I______procedure, the others 100 100* applied the Maastricht 5 10 15T 2 protocol. U S~~~~ I T0~~~~

0- 0- 5 10 15 20 5 10 15 20 Participant Participant

samples 2 and 3 were correctly classified as study on receptor immunocytochemistry a oestrogen receptor positive by 18 (90%) of the reliable scoring procedure for the immunoreac- 20 laboratories by obtaining a histochemical tion should be established. Several approaches score ofmore than 35. A quantitative classifica- have been described for the semiquantitative tion based on a histochemical score within 95% evaluation of the receptor staining, such as the confidence interval limits was provided by 40% assessment of the percentage of positive and 60% of the participants for samples 2 and cells,2 18-20 classification of varying numbers of 3, respectively. To test the hypothesis that staining categories,15 1-21 and various ways of deterioration of oestrogen receptor immuno- mathematical analysis.'51612' For this study reactivity over time might have influenced the we chose a quantitation method based on results, the participating laboratories were counting a fixed number of cells, classified asked to supply the date that the sections were according to staining intensity, which resulted stained, which allowed us to calculate the in low intra- and interobserver vari- elapsed time interval. In fig 4 the histochemical abilities. 51622 scores are plotted against time. It is clear that Our results show that very significant dif-

there is no correlation between immunostain- ferences exist among laboratories in the results http://jcp.bmj.com/ ing and time interval. ofthe immunocytochemical oestrogen receptor assay. Comparison ofthe stained sections (fig 2) shows that this is largely due to variability in Discussion staining intensity and not to differences in the Our results clearly indicate that the chosen semiquantitative evaluation as reflected in the approach, circulation of the freeze dried cryo- histochemical score.

stat sections by normal post is adequate, given The interlaboratory variability does not on September 28, 2021 by guest. Protected copyright. the fact that immunoreactivity did not seem to be attributable to the choice of tech- deteriorate over time. A problem with the nique: no significant differences were found circulation of tissue sections obtained from a 400] block of (tumour) tissue is that tissue samples might be heterogeneous, which may lead to differences among laboratories because of biological var'iation rather than technical 300- A inadequacies. This problem can be overcome A by monitoring the composition of the tissue 0 0 0 sample by examining sections at regular A 0 A intervals. It does, however, limit the number of U.) A 200 laboratories which might be included in a *8 single quality control programme, probably A A A 0 0 A not more than 50 and optimally, about 25. [cP A AA To perform an interlaboratry quality control D 100

Table 2 Tumour classification (second test) 6 0

A Quantitative* Qualitativet o0 I I~I-- 11 Sample Correct Incorrect Correct Incorrect 1 10 16 26 35 45 Interval (days) 2 8 12 18 2 3 12 8 18 2 Figure 4 Semiquantitative immunohistochemical scoring ofoestrogen receptor staining as afunction of time *Within 95% confidence interval. interval between sectioning and staining. tHistochemical score of more than 35. D sample 2; A sample 4. 124 Bosman, de Goeij, Rousch

between the two main staining protocols used. 4 Reiner A, Reiner G, Spona J, Schemper M, Holzner JH. Histopathologic characterization of human breast cancer In terms of overall assignment of the tumours in correlation with estrogen receptor status. A comparison to oestrogen receptor positive or negative ofimmunocytochemical and biochemical analysis. Cancer 1988;61:1 149-54. J Clin Pathol: first published as 10.1136/jcp.45.2.120 on 1 February 1992. Downloaded from categories a fair score was reached (correct 5 Howat JMT, Harris M, Swindell R, Barnes DM. The effect assignment in 100% for oestrogen receptor ofoestrogen and progesterone receptors on recurrence and survival in patients with carcinoma of the breast. Br J negative and of 80% for oestrogen receptor Cancer 1985;51:263-8. positive samples). In terms of the quantitative 6 Koenders A, Thorpe SM. Standardization of steroid receptor assays in human breast cancer. Long-term within evaluation, however, only 40% and 60%, res- laboratory and between laboratory variation of estrogen pectively of the oestrogen receptor positive and progesterone receptor assays. Eur J Cancer Clin Oncol 1986;22:945-52. samples were rated within a 95% confidence 7 Jordan VC, Zava DT, Eppenburger E et al. Reliability of interval. steroid hormone receptor assays-an international study. Eur J Cancer Clin Oncol 1983;19:357-63. Our results lead to the following conclusions: 8 De Goeij AFPM, Bosman FT, Berns EMJJ. Determination 1 Circulation of freeze dried sections is an ofsteroid hormone dependency oftumours utilizing tissue sections. Survey of histochemical techniques and their adequate approach for the quality control of application in surgical pathology. J Pathol 1986;149: oestrogen receptor immunohistochemistry. 163-72. 9 McCarty KSJr,ReintgenDS, SeiglerHF. Cytochemistryof 2 Semiquantitative scoring of immunocyto- sex steroid receptors: a critique. Br J Cancer Res Treat chemical staining is valuable for performing 1982;1:315-25. 10 Underwood JC. Oestrogen receptors in human breast can- correlative interlaboratory studies, although cer: a review ofhistopathological correlations and critique this scoring protocol may not be required for ofhistochemical methods. Diagn Histopathol 1983;6:1-22. 11 King WJ, Greene GL. Monoclonal antibodies localize diagnosis or prognosis. oestrogen receptor in the nuclei of target cells. Nature 3 Significant interlaboratory variability exists, 1984;307:745-7. 12 Heubner A, Beck T, Grill H-J, Pollow K. Comparison of leading to a qualitatively correct oestrogen immunocytochemical estrogen receptor assay, estrogen receptor classification in 100% negative but receptor enzyme immunoassay, and radioligand-labeled estrogen receptor assay in human breast cancer and 80% in positive oestrogen receptor cases and uterine tissue. Cancer Res 1986;46(suppl):4291 S-5S. quantitative agreement of only 45% (95% 13 Anderson J, Bentzen SM, Poulsen HS. Relationship between radioligand binding assay, immuno enzyme assay confidence interval). and immunohistochemical assay for estorgen receptors in 4 The perceived variability is not caused by human breast cancer and association with tumor differen- tiation. Eur J Cancer Clin Oncol 1988;24:377-84. systematic differences in the choice of the 14 Stefanini M, De Martini C, Zamboni L. Fixation of immunocytochemical technique. ejaculated spermatozoa for electron microscopy. Nature 1967;216:173-4. 5 Quality control programmes should be 15 McCarty KS Jr, Miller LS, Cox EB, Konrath J, McCarty included in the standard procedures of every KS Sr. Estrogen receptor analyses. Correlation of biochemical and immunohistochemical methods using mono- diagnostic immunohistochemistry laboratory. clonal antireceptor antibodies. Arch Pathol Lab Med 1985;109:716-21. 16 Scheres HME, De Goeij AFPM, Rousch MJM, et al. We are much indebted to Dr Th Benraad and Mrs Anneke Estrogen receptor quantification in human breast cancer: Geurts of the Laboratory for Experimental and Chemical radiochemical assay on cytosol and cryostat sections Endocrinology, Sint Radboud Hospital, University of compared with semiquantitative immunocytochemical Nijmegen, for performing the radiochemical and immunocyto- analysis. J Clin Pathol 1988;41:623-32. chemical oestrogen receptor assay on the two tumour cytosols 17 EORTIC Breast Cancer Cooperative Group. Revision ofthe used in the second trial. standards for assessement ofhormone in human receptors http://jcp.bmj.com/ We thank the pathologists who participated in the quality breast cancer. Eur JCancer 1980;16:1513-15. control programme. A Th Ariens, Sittard; M Beckers-Bloem- 18 King WJ, De Sombre ER, Jensen EV, Greene GL. Compar- kolk, Zaandam; B Beerends, Dordrecht; APR Blok, Den Haag; ison of immunocytochemical and steroid-binding assays H Ebels, Groningen; S Hensen-Logmans, Rotterdam; R van der for estrogen in human breast tumors. Cancer Res Heul, Rotterdam; D Israel, Bergen op Zoom; J Koudstaal, 1985;45:293-304. Heerlen; J Misere, Tilburg; F van de Molengraft, Alkmaar; W 19 De Sombre ER, Thorpe SM, Rose C, et al. Prognostic Mooi, Amsterdam; J Peters, Breda; R Schapers, Venlo; W usefulness of estrogen receptor immunocytochemical Schoneman, Amsterdam; JW Steffelaar, Den Haag; P van der assays for human breast cancer. Cancer Res 1986;46 Valk and Dr C Meijer, Amsterdam; GP Vooijs, Nijmegen; Th (suppl):4256S-64S. Vroom, Amsterdam; Sj S Wagenaar, Nieuwegein. 20 GiriDD, Dangerfield VJM, Lonsdale R, Rogers K, Under- This study was partly supported by grant No RUL 85-3 from wood JCE. Immunohistology of oestrogen receptor and on September 28, 2021 by guest. Protected copyright. The Netherlands Cancer Foundation KWF. D5 antigen in breast cancer: correlation with oestrogen receptor content of adjacent cryostat sections assayed by radioligand binding and enzyme immunoassay. J Clin Pathol 1987;40:734-40. 22 Remmele W, Hildebrand U, Hienz HA, et al. Comparative histological, histochemical, immunohistochemical and 1 Elias JM, Gown AM, Nakamura RM, et al. Quality control biochemical studies on oestrogen receptors, lectin recep- in immunohistochemistry. Report of a workshop spon- tors, and Barr bodies in human breast cancer. Virchows sored by the Biological Stain Commission. Am J Clin Arch (Pathol Anat) 1986;409:127-47. Pathol 1989;92:836-43. 23 Snijders MP, De Goeij AF, Koudstaal J, et al. Is immuno- 2 McGuire WL. Prognostic factors in primary breast cancer. histochemical analysis of oestrogen and progesterone Cancer Surveys 1986;5:527-36. receptors in endometrial carcinoma superior to the 3 De Sombre ER. Breast cancer: hormone receptors, prog- radioligand binding assay? JPathol 1990;161:129-35. nosis and therapy. Clin Oncol 1982;1:191-213.