Articles Nordic Immunohistochemical Quality Control (NordiQC) - An Organization for External Quality Assurance Mogens Vyberg, MD Institute of Pathology. Aalborg Hospital Aarhus University Hospital Denmark Søren Nielsen, HT, CT Institute of Pathology Aalborg Hospital Aalborg University Hospital Denmark Graduated in medicine from University of Copenhagen, Denmark, 1973. Senior Pathology Consultant, Institute of Pathology, Aalborg Hospital, Aarhus University Hospital from 1988. The main fields are gastrointestinal pathology, liver pathology and neuropathology as well as immunohistochemistry. Author and Co-author of more than 50 scientific papers and several book chapters in these fields. Author of the text book Applied Immunohistochemistry [Danish]. Author of the Dako Antibody Algorithm for Tumor Diagnostics. Co-founder and Scheme Manager of Nordic Immunohistochemical Quality Control (NordiQC) established in 2003 with Aalborg Hospital as its domicile. Senior Histotechnologist. International lecturer and organizer of immunohistochemical workshops, particularly in the field of protocol optimization. Co-author of more than 10 scientific papers based on immunohistochemical studies. Scheme Organizer of NordiQC since 2003. Is has long been recognized that standardization is vital for reliable and reproducible results of immunohistochemistry (IHC), but in spite of this, laboratory protocols are not regulated and the staining quality varies greatly between different laboratories depending on the individual selection of methods and the technical expertise (1-6). IHC is influenced by multiple biological and technical parameters and the final result is highly related to a number of choices made to set up the analysis. Even though internal quality control (IQC) procedures, essential for reproducibility of the IHC performance in the individual laboratory, are carried out, they will often not identify a poorly calibrated IHC system giving insufficient staining results. External quality assessment (EQA) is a system, which retrospectively and objectively compares staining results from many laboratories by means of an external agency. In contrast to IQC, EQA allows the identification of insufficient stains, inappropriate protocols, interpretation problems and lack of appropriate use of control tissue (2,5). The multiple parameters influencing the results of IHC can largely be ascribed to 1) Pre-analytic parameters (i.e. tissue material, tissue handling before submission to the laboratory, tissue processing in the laboratory); 2) Analytic parameters (i.e. antigen retrieval procedure, staining protocol, control selection, validation); and 3) Post-analytic 45 Connection 2008
parameters (i.e. results interpretation and reporting) (4, 5). In an EQA setting, by circulating serial sections to a number of laboratories to be stained for specific antigenic markers and assessing them in a standardized and objective way, the pre- and post-analytical parameters influencing the results of IHC can be taken into consideration, allowing a direct comparison of the laboratory performance and analytic parameters potentially influencing the staining quality. Establishment of NordiQC. A Brief History In 1999, Dako Nordic organized a meeting of pathologists from Denmark, Finland, Norway and Sweden to find a suitable means of optimizing methods and improving results of clinical IHC. Based on these meetings and a set of pilot runs, NordiQC was established on January 1, 2003, as an independent, scientific, non-profit organization for Nordic laboratories. The following year, the capacity was expanded to include a limited number of laboratories outside the Nordic countries, and currently, 130 laboratories from 20 countries are enrolled. The NordiQC EQA scheme currently comprises of a general module with three annual runs catering for a total of 15-18 markers selected for diagnostic purposes in pathology departments, particularly in the analysis of neoplastic lesions. Recently a breast module with two annual runs catering mainly for estrogen and progesterone receptors and HER- 2 has been initiated. Laboratories can only participate by submitting protocols detailing the technical variables in a web-based questionnaire. For tests, multi-tissue blocks made from several (usually 5-7) normal and tumor tissue samples selected to include cells with varying content of epitopes are used. Included are tissues containing critical stain quality indicators (CSQI); defined as cell or tissue epitopes that are weakly expressed, due to a low concentration and, therefore, easily lost in suboptimal protocols. For each marker to be demonstrated, a pair of unstained slides is circulated to the participating laboratories, which are requested to perform IHC stains using their standard protocols and return one of the slides. All slides are assessed anonymously by pathologists and technicians experienced in analyzing IHC slides. Each stain is by consensus marked as optimal, good, borderline or poor. The general results of each testing are presented on the Web site www.nordiqc.org in an aggregate fashion together with an analysis of the protocol data pointing out variables that are found to be of importance for the staining quality. Individual scores are sent to all participating laboratories confidentially using e-mail. In the case of borderline or poor marks, individual explanations and specifically tailored recommendations for protocol improvement are provided. Results of five years NordiQC work Between 2003-2007, NordiQC tested the immunohistochemical staining for 65 clinically relevant epitopes at least one to five times (refer to details at www.nordiqc.org). The over-all scores were almost evenly distributed between optimal, good and insufficient (i.e. borderline or poor). For some important markers like CD5, CD15, chromogranin A, cytokeratins, and immunoglobulin light chains, as many as 50-70% of the tests were insufficient in the first runs. The main causes of insufficient stains have been identified as: 1. less successful primary Abs, including ready-to-use (RTU) products, 2. insufficient, inappropriate or missing epitope retrieval, 3. improper calibration of primary Ab concentration, and 4. endogenous biotin reaction. Examples of optimal and insufficient staining results are illustrated in Figs. 1-3. Often a combination of several of the above-mentioned factors have been found. In about 90% of the insufficient results, the sensitivity of the applied protocols was too low resulting in too weak or false negative reactions. False-positive tests were found in about 10% of all insufficient results (giving a false-positive reaction or marked diffuse background). For tests carried out several times, the proportion of insufficient results declined in almost all cases. Among the laboratories following the NordiQC recommendations, improvement in a subsequent run was seen in about 70%, while among those who did not change their protocols, improvement was seen in less than 20%. In the following, some details from selected examples of test results are described. Estrogen receptor (Fig. 1) The rate of sufficient test results increased considerably during five runs, from 32 of 71 laboratories (45%) in the first run to 61 of 73 (84%) in the latest. Optimal staining results could be obtained with all the three mab clones SP1, 6F11 and 1D5. However, with the latter a generally higher proportion of suboptimal results occurred, indicating that this clone may be less robust, if other protocol parameters are not optimal (frequently insufficient Heat-Induced Epitope Retrieval (HIER)). This is in accordance with the results obtained by United Kingdom National External Quality Assessment Scheme for Immunocytochemistry (UK NEQAS-ICC). (http://www.ukneqasicc.ucl.ac.uk/neqasicc.shtml) and the recently published results by Phillips et al. (7). These evaluations are based on staining intensity and are not clinically correlated. Laboratories that changed their estrogen receptor protocol more or less according to the NordiQC recommendations, improved their staining result much more frequently than those not changing their protocol (Table 1). An example of optimal vs. insufficient estrogen receptor staining is shown in Fig 1. Laboratory Recommendation Recommendation followed not followed Number of labs advised 39 24 Number of labs improved 32 (82%) 7 (29%) Table 1. Effect of NordiQC recommendations to laboratories with an insufficient estrogen receptor staining Connection 2008 46
CD79a (Fig. 2) Among 112 laboratories submitting stains in the latest run, most used mab clone JCB117 and obtained sufficient results in 85%. In contrast, all of the six laboratories using mab clone HM57 obtained poor marks, due to insufficient staining of neoplasms (Fig. 2) Even the Dako Web site suggests that the clone JCB117 be preferred over HM57, but the datasheet does not outline the problems with the use of clone HM57 on human material*. Cytokeratins, low molecular weight (CK-LMW) (Fig. 3) Looking at 5 mab clones used by at least six laboratories in the latest runs, the proportion of sufficient stains was: 35betaH11: 6/26 (23%), 5D3: 9/15 (60%), C51: 13/14 (93%), CAM 5.2: 21/47 (45%), and DC10: 33/37 (89%). The prevalent feature of an insufficient staining was a very weak or false - negative staining of liver cells or cells of the renal cell carcinoma (Fig. 3). Virtually all the laboratories were able to detect CK-LMW in columnar epithelial cells, emphasizing that these cells cannot be used as CSQI. While suboptimal results were in part due to insufficient or inappropriate epitope demasking, it was also confirmed by retesting in the NordiQC laboratory that differences in the performance of the antibodies was an important parameter (Fig. 3). Consequently, a change in clone was recommended to laboratories using one of the less successful clones Discussion For the majority of the IHC tests, optimal staining results can be obtained by following different protocols for antigen retrieval techniques, primary antibodies or visualization systems. Although individual choices of protocols are necessary for new developments of reagents, systems and instruments, these procedures complicate the approach to standardization. It is in this context that the identification of tissues containing CSQI becomes important. Unfortunately, neither the scientific literature nor the vendors datasheets are focused on this issue. NordiQC puts special emphasis on the identification of robust control tissues enabling the laboratories to reliably validate their own staining results. It also encourages the laboratories to join the efforts of standardization. NordiQC is careful to advice laboratories how to improve their performance based on readily accessible solutions. Basically, improvement of the protocols can be readily done by the participating laboratory that uses so-called open staining systems, when the analytical problems are identified and the laboratory has become aware of the CSQI. However, when insufficiencies are detected in so-called closed staining systems, there is very little that the participating laboratory can do, and the producers of such closed systems and RTU products should change test parameters. With regard to proper HIER, the advantage of the alkaline buffers in terms of enhanced sensitivity and analytical robustness has been confirmed in the NordiQC reference laboratories for almost all the markers tested. When an efficient HIER is applied, endogenous biotin is also revealed (8) inevitably giving false-positive staining reactions in many tissues and tumors in biotin-based detection systems. The best solution to overcome this problem is to implement a non-biotin based visualization system. NordiQC is cautious in suggesting change of antibody unless the evidence that this should be done is strong. In some cases, tests in the NordiQC reference laboratory sustain the suspicion obtained from the results of many participating laboratories that an optimal staining with a certain primary antibody is difficult or impossible, as is the case with several cyclin D1 antibodies (9) and cytokeratin antibodies as illustrated in this paper. When confirmatory tests have not been carried out, it cannot be excluded that an unsuccessful antibody would eventually produce optimal results by adjustments of the protocols. In calibration of the dilution of the concentrated primary antibodies, the CSQI in the control tissues will determine the optimal dilution. Such indicators should enable laboratories to adjust the concentrations of the primary antibodies to produce the required staining results irrespective of their protocol platform. In contrast, RTU antibodies (pre-diluted antibodies), which have not been properly developed to meet the present diagnostic criteria or to fit into various protocol settings, often result in insufficient stains**. NordiQC has demonstrated that significant improvement may be achieved as a direct consequence of EQA that provides feedback to the less successful laboratories with an explanation of the probable cause of the insufficiency and individually-adjusted recommendations for protocol improvement. The strong correlation between protocol adjustments (according to NordiQC advice) and improved laboratory performance is even more encouraging as it directly reflects the benefit of quality assurance Conclusion EQA should be implemented in all IHC laboratories in the healthcare system as well as in any company guided by the same goals and principles that apply to other clinical laboratory testing. While standardization of methods does not appear possible, the emphasis should be focused on standardization of controls and staining results. It has to be taken into account that such standards are not invariable, but must be adjusted in parallel with the development of knowledge as well as future technical developments. Consequently, EQA organizations that continuously survey IHC performance in the pathology laboratories must be accessible and considered to be obligatory for diagnostic anatomic pathology. *Since the editing of this article, Dako has made the necessary changes to correct this information in production lot of the CD79, clone HM57 antibody. **A new pathology quality endorsed RTU product line will be launched by Dako in March 2008 47 Connection 2008
References Fig. 1. Estrogen receptor 1. Rhodes A, Jasani B, Barnes DM, Bobrow LG, Miller KD. Reliability of immuno-histochemical demonstration of oestrogen receptors in routine practice: interlaboratory variance in the sensitivity of detection and evaluation of scoring systems. Clin Pathol 2000;53(2):125-30 2. Varma M, Berney DM, Jasani B, Rhodes A. Technical variations in prostatic immunohistochemistry: need for standardisation and stringent quality assurance in PSA and PSAP immunostaining. J Clin Pathol. 2004 Jul;57(7):687-90. 3. Taylor CR. The total test approach to standardization of immunohistochemistry. Arch Pathol Lab Med. 2000;124(7):945-51. 4. O Leary TJ. Standardization in immunohistochemistry. Appl Immunohistochem Mol Morphol. 2001;9(1):3-8. 5. Taylor CR. Standardization in immunohistochemistry: the role of antigen retrieval in molecular morphology. Biotech Histochem. 2006 Jan-Feb;81(1):3-12. 6. Taylor CR. Standardization in immunohistochemistry: the role of antigen retrieval in molecular morphology. Biotech Histochem. 2006 Jan-Feb;81(1):3-12. 7. Phillips T, Murray G, Wakamiya K, Askaa J, Huang D, Welcher R, Pii K, Allred DC. Development of standard estrogen and progesterone receptor immunohistochemical assays for selection of patients for antihormonal therapy. Appl Immunohistochem Mol Morphol. 2007 Sep;15(3):325-31. 8. Bussolati G, Gugliotta P, Volante M, Pace M, Papotti M. Retrieved endogenous biotin: a novel marker and a potential pitfall in diagnostic immunohistochemistry. Histopathology. 1997 Nov;31(5):400-7. 9. Torlakovic E, Nielsen S, Vyberg M. Antibody selection in immunohistochemical detection of cyclin D1 in mantle cell lymphoma. Am J Clin Pathol. 2005 Nov;124(5):782-9. A. Optimal estrogen receptor (ER) staining of the uterine cervix, which is appropriate for control tissue. Virtually all columnar and squamous epithelial cells show a distinct nuclear staining. The smooth muscle cells are also stained, while endothelial and lymphoid cells are negative. C. Optimal estrogen receptor (ER) staining of a lobular breast carcinoma. The majority of the tumor cell nuclei are strongly stained. B. Insufficient ER staining of the uterine cervix (same field as in Fig. 1A). Only a subset of epithelial cells shows a nuclear reaction. With this protocol, some breast carcinomas will show a false negative reaction (see D). D. Insufficient ER staining of the lobular breast carcinoma shown in (C). The tumor cell nuclei are virtually negative. The reason was a too dilute antibody. Insert: A ductal breast carcinoma positively stained with the same protocol. High expression of epitopes may give a positive reaction even with an insensitive protocol misguiding the laboratory. E. This staining of the lobular breast carcinoma was scored as good. The nuclear staining reaction is moderate. There is a weak cytoplasmic reaction due to endogenous biotin. F. This staining of the lobular breast carcinoma was scored as poor. There is a weak nuclear staining reaction and a moderate cytoplasmic reaction due to endogenous biotin. Connection 2008 48
Fig. 2. CD79a A. Optimal CD79a staining of the tonsil using the monoclonal antibody (mab) clone JCB117. The mantle zone B-cells show an intense staining, while the germinal centre B-cells show a moderate staining. B. CD79a staining of the tonsil same field as in (A) using an insufficient protocol based on the mab clone JCB117 in a too low concentration. The mantle zone B-cells and the late stage germinal centre B-cells show a moderate staining, while the rest of the germinal centre B-cells are only weakly demonstrated. C. High magnification of the border between the mantle zone and germinal centre in (A) showing the difference in staining intensity between the two cell types. D. High magnification of the border between the mantle zone and germinal centre in (B) showing the weak staining in the two cell types. Compare with (C). E. Optimal staining of a testicular precursor B-acute lymphoblastic lymphoma using the mab clone JCB117. Virtually all the neoplastic cells show a distinct reaction, while a seminiferous tubule (upper right) is unstained. F. Poor CD79a staining of the precursor B-acute lymphoblastic lymphoma using the same protocol as in (B). The neoplastic cells are virtually negative same field as in (E). G. Optimal staining of appendix using the mab clone JCB117. The B-cells show a strong and distinct reaction while the epithelial and smooth muscle cells are negative. H. Staining of appendix using the mab clone HM57. The mantle zone B-cells show a strong reaction, while the germinal centre B-cells are weakly stained. Cross reaction - inherent to this clone - is seen in epithelial and smooth muscle cells. Fig. 3. Cytokeratin, low molecular weight (CK-LMW) A. Optimal staining for CK-LMW of the liver using mab clone DC10. The same result can be obtained with clones C51, 5D3 and Ks-B17.2.The majority of the hepatocytes show a distinct, moderate to strong, predominantly membranous reaction. B. Insufficient staining for CK-LMW of the liver using mab clone 35 BH11- same field as in (A). Only the bile duct epithelial cells are positive, while the hepatocytes are virtually negative. Even after optimization in our own laboratory no significant staining of the hepatocytes was seen. C. Optimal staining for CK-LMW of the renal cell carcinoma using mab clone DC10. The majority of the tumor cells show a strong reaction. D. Insufficient staining for CK-LMW of the renal cell carcinoma using mab clone 35BH11, same field as in (C). The tumor cells are weakly stained or unstained. 49 Connection 2008