Validation of immunohistochemical tests performed on cytology cell block material: Practical application of the College of American Pathologists’ guidelines

INTRODUCTION

Cytology cell blocks have assumed increased importance due to the increased demand for specific diagnosis beyond the general categories of benign, malignant, or inflammatory. There is, however, little appreciation of the different methods of creating cytology cell blocks which can lead to significant variation in the performance of immunohistochemical tests. The Clinical Laboratory Improvement Amendments (CLIA) of 1988 requires laboratories in the United States of America (USA) to verify the performance of patient tests. The College of American Pathologists (CAP) in a questionnaire-based survey of 727 laboratories in the USA, found that a significant proportion (63%) of respondents did not have written procedures for validation of immunohistochemical test of cytologic material.[1] Of those laboratories that claimed to validate the immunohistochemical tests, there was wide variation of the validation process. This led to the publication of CAP guidelines for analytic validation of immunohistochemical assays which also includes a brief reference to cytology specimens.[2]

The objective of a cell block is to collect all the cells in the suspension for subsequent processing through paraffin embedding and microtome cut sections stained with hematoxylin and eosin (H and E). The cell concentration is facilitated by centrifugation to form a pellet. The pellet may be collected in a collodion bag which is then tied off before formalin fixation and paraffin embedding (FFPE). Molten agar can be used to bind the pellet of cells before solidifying the mixture by placing it in the refrigerator. This is then processed by FFPE before H and E sections examined. The plasma–thrombin method is the most widely utilized method of generating a cytology cell block. This is due to the relative ease of stirring one drop of plasma to the centrifuged cell pellet before addition of one drop of thrombin (BD Thermo Scientific Cytoblock). The soft clot is then transferred to a cassette for FFPE and sectioning. The Cellient Automated Cell Block System (Hologic) diverges from centrifugation to using vacuum-assisted filtration to deposit the cells directly into the cassette. Table 1 summarizes the pros and cons of these methods.[345678]

It is evident that there are multiple variables in creating a cytology cell block. This includes the time the cytology material spends in the fixative solution, the type of fixative used (ethanol, methanol, isopropanol, formalin, and mixtures of these), protein contaminants in the form of plasma and thrombin, and heat and cooling artifacts on cells in agar cell blocks. The cell block is therefore, potentially, a unique product of the laboratory in which it was created. This is a significant factor pertinent to the performance of validation studies.

COLLEGE OF AMERICAN PATHOLOGISTS GUIDELINES ON ANALYTIC VALIDATION OF IMMUNOHISTOCHEMICAL STUDIES

The CLIA charges laboratories reporting patient results to verify their assays in a stepwise manner through assay optimization, validation, and ongoing monitoring for consistency. The CAP questionnaire-based laboratory survey[1] found “significant interlaboratory variation in validation practices and revealed that many laboratories do not follow consistent procedures when validating immunohistochemical assays.” The CAP, therefore, published guidelines with the specific objective of helping pathology laboratories comply with validation and revalidation requirements. Five recommendations and nine expert consensus opinions were issued to guide the immunohistochemical test validation procedures. These guidelines pertain mainly to surgical material except for one expert consensus opinion which addresses cytology specimens by stating that “… laboratories should test a sufficient number of such cases to ensure that assays consistently achieve expected results. The laboratory medical director is responsible for determining the number of positive and negative cases and the number of predictive and nonpredictive markers to test.”

PROCEDURE FOR APPLICATION OF COLLEGE OF AMERICAN PATHOLOGISTS IMMUNOHISTOCHEMICAL STUDIES VALIDATION GUIDELINES TO CYTOLOGY CELL BLOCKS

Our first priority was to select a subset of antibodies to validate on cytology cell blocks as validation of all antibodies used was impractical due to limited material, time, and expense of the validation process. Table 2 shows a breakdown of specimen types that compose our cytology sample. The lung panel (TTF-1, p40, p63, AE1/3, CK7, CD56, synaptophysin, and chromogranin) was the most frequently utilized diagnostic immunohistochemical study in our laboratory with a frequency of 1016 tests for TTF-1 and 143 tests for chromogranin. In addition, Ki-67 (219 cases) and CK8/18 (188 cases) were also within the frequency of most utilized immunohistochemical stains and hence were included into our antibody selection.

Ten consecutive lung resection specimens were used for validation studies and sampling occurred within 45 minutes of removal from the patient. The lesion in the fresh resection specimen was scraped twice. The first scraping was suspended in alcohol-based fixative (CytoRich Red) and the second scraping was suspended in 10% formalin solution, both were processed as cell block material using the plasma–thrombin method. In addition, a section of the lesion was also procured for processing as FFPE tissue. All materials were handled as per existing standard operating procedures. The cytology cell block material served as both positive and negative control; the negative control being the resident benign population of cells such as macrophages, bronchial cells, lymphoid cells, as well as nonreactive malignant cells.

Immunohistochemical tests on cytology cell block material were performed in the same manner as that validated for the surgical material which had >90% concordance with the morphologic gold standard. The results of the validation study scored percentage of cells staining in the surgical and cytology material as follows (0%, 1%–25%, 26%–50%, 51%–75%, and >75%) except for Ki67 which was scored at decile intervals (<0%, 1%–10%, 11%–20%, etc). Concordance was achieved if the difference in staining was within the same quartile or in case of Ki-67, within the same decile interval. In addition, intensity of staining compared to control FFPE tissue was scored on cytology material using a binary method of equal or greater than control (score of 1) or less than control (score <1).

RESULTS

Table 2 presents the breakdown of cytology specimens received in our department over the 8-month period preceding our validation study. Although exfoliative specimens from body fluids and lymph node aspirates were more numerous, the majority of these specimens did not need immunohistochemical studies (IHC) for characterization of the pathologic process. Specimens from the lungs were the ones that most frequently utilized immunohistochemical tests for diagnosis.

The performance characteristics of the validation tissue panel obtained from ten resected lung specimens are illustrated in Table 3. CK8/18, CD56, chromogranin, synaptophysin, p40, and TTF-1 showed equal intensity and proportional staining of neoplastic cells regardless of whether the cell blocks were fixed in alcohol (CBA) or formalin (CBF). Figure 1 illustrates the TTF-1 staining in the three types of preparations.

Close

Figure 1

Equal intensity and proportional staining of TTF-1 in tumor cells, formalin-fixed paraffin-embedded, ×400 (a), cell blocks fixed in formalin, ×400 (b), and cell blocks fixed in alcohol, ×400 (c)

Export to PPT

AE1/3, p63, and CK7 revealed variation in proportion and intensity of staining of neoplastic cells. Compared with the control FFPE tissue, the proportion of AE1/3-stained tumor cells in one of seven specimens fell into a lower quartile in both CBA and CBF. Compared to the FFPE gold standard, less intense AE1/3 staining of tumor cells was identified in one CBF with no difference in AE1/3 staining intensity identified in any of the CBA [Table 3].

For p63, one out of nine cases showed 1%–25% staining of tumor cells in FFPE tissue compared to no staining of tumor cells in both CBA and CBF. The intensity of the other seven cases was comparable across the preparations with only one CBA specimen showing less intense staining than FFPE tissue [Table 3].

For CK7, in one out of eight specimens, tumor cell staining in the FFPE fell into a lower quartile than tumor cells in the CBF. There was a comparable proportion of staining in all CBA. No variation in intensity of staining was identified across the specimen preparation types [Table 3].

The Ki-67 results were particularly subject to variation in both proportion and intensity of staining [Table 4]. Compared with the control FFPE tissue, the proportion of tumor cell staining was lower in the CBA and CBF of all eight cases. The proportion variation of staining of the cells ranged from 10% to 20%. The intensity of staining of the cell block material was lower (six out of eight) compared to FFPE material. This appeared independent of whether the cell block was from alcohol or formalin-fixed material. Figure 2 illustrates the Ki-67 staining in the three types of preparations.

Close

Figure 2

Difference in intensity and proportion of staining of Ki-67 in tumor cells, formalin-fixed paraffin-embedded, ×200 (a), cell blocks fixed in formalin, ×200 (b) and cell blocks fixed in alcohol, ×200 (c)

Export to PPT

DISCUSSION

With the exception of Ki-67, IHC on cytology cell blocks created using plasma–thrombin method performed comparatively well with reference to the results of previously established immunohistochemical protocols for surgical material. There was broad concordance within the same quartile interval of the proportion of cells staining and intensity of staining across the alcohol-fixed cell block, formalin-fixed cell block, and FFPE material. Intensity of staining (both nuclear and cytoplasmic/membranous) was broadly equivalent across the preparation types. Although there was occasional variation in intensity and proportion of neoplastic cells in cell block compared to FFPE material with reference to AE1/3, CK7, and p63 antibodies these did not preclude interpretation and could be explained by tumor cell heterogeneity rather than processing methods.

The performance of Ki-67 was particularly variable in the cell block material. The degree of variation was interpreted to be too wide for meaningful clinical use. Tumor heterogeneity as well as plasma–thrombin cell block preparation could contribute to the differences observed. We, therefore, cautiously interprete Ki-67 on cell block materials with the realization that it underestimates the proliferative index and that no staining might be due to loss of reactivity of the antibody.

There are various reports whereby the results of nuclear staining are dependent on the clone of antibody used[9] and the fixative. Apart from the Ki-67 antibody clone, all the other antibody clones used in our validation study performed equally well with respect to nuclear positivity of the stain in alcohol-based CytoRich Red and 10% formalin fixatives.

In theory, where there is <90% concordance, the McNemar test may be used to determine whether there is a significant positive or negative bias between the test and reference material. This would then inform mitigation processes such as antigen retrieval procedure and variation in incubation times. The McNemar test requires a large number of samples to run the two by two table and is therefore impractical in our case where only ten specimens comprised the validation set. Changes to the immunohistochemical protocol seem feasible, but the added effort of optimization and revalidation as well as programming of the immunohistochemistry instrument for a varying protocol specifically used for cytology cell block make it a less attractive option.

As noted previously, we did not validate all the antibodies used in histology specimens because of lack of specific tissue type, expense, and time of performing antibody validation tests. Our cost analysis revealed price per slide ranging from $8.34 (majority of antibodies) to $15.20 (p63 antibody) with technical time estimated at $14 per slide based on 30 minutes of time utilized. The total cost amounted to $4000 for the ten cases that were validated. The cost burden would increase significantly if prognostic markers (such as HER2/neu, ER, and PR) are to be validated. This is because the study would require the more stringent criteria of including twice the number of specimens used for diagnostic antibody markers.[2]

FUTURE INNOVATIONS

Efforts toward standardization would very likely lower the cost and increase the range of antibodies that could be validated. Suspension of cells from all needle rinses in formalin would eliminate the duplication of validation procedures for alcohol-fixed cells. Automation of cell block creation would similarly minimize preanalytic variables influencing cell block performance for cytology. The Cellient Automated Cell Block System is an example of such a system; however, it has the distinct disadvantage of processing only one cell block at a time. Validation sets such as cytology cell block microarrays and cell culture pellets would allow laboratories to share validation sets and enable antibody optimization studies before the validation process. Progress toward standardization of cytology cell blocks will have added value of enhancing the validity of other ancillary tests such as in situ hybridization and molecular tests.

SUMMARY AND CONCLUSION

There is increased expectation that a cytology specimen diagnosis would be precise and definitive to allow treatment and prognostication. IHC are central to diagnosis, and it behooves laboratories to validate their tests so as to ensure accuracy.

Our stepwise validation method is based on an academic university cytopathology laboratory practice [Figure 3]. The first step is identification of the most frequent cytology specimens on which diagnostic IHC are performed. This informs on the type of specimen(s) to use to generate tissue for a validation panel. Using standard laboratory procedures, we were able to validate antibodies tested which were in use in the histology laboratory and which demonstrated >90% concordance with morphologic gold standard. Non validated immunohistochemical stains are still used on cell blocks, but these are interpreted cautiously as advised by the CAP guideline. The guidelines also state that a disclaimer should be used in the report to draw attention to this.

Close

Figure 3

Flow diagram of stepwise validation of immunohistochemical studies in cytology cell blocks. *Number of specimens in panel at discretion of Medical Director. **Ki-67 was scored in decile (see main text)

Export to PPT

Innovations of standardizing preanalytical elements for cell blocks would reduce the cost of validation studies, increase the range of validated antibodies, allow for interlaboratory utilization of standard material, and add value for validation of other ancillary testing that is done on cell blocks such as molecular and in situ hybridization studies.

COMPETING INTERESTS STATEMENT BY ALL AUTHORS

The authors have no competing interests.

AUTHORSHIP STATEMENT BY ALL AUTHORS

Dr. Swati Satturwar, Dr. Renuka Malenie, Dr. Ann Sutton, Dr. Diana Dai and Dr. F. Zahra Aly have contributed in collection of cases, analysis of data and writing of manuscript. Dr. Ann Sutton and Dr. F. Zahra Aly edited the manuscript. All authors were involved in conceptualizing the study.

ETHICS STATEMENT BY ALL AUTHORS

Not applicable.

LIST OF ABBREVIATIONS (In alphabetic order)

CAP – College of American Pathologists

CBA – Cell blocks fixed in alcohol

CBF – Cell blocks fixed in formalin

CLIA – Clinical Laboratory Improvement Amendments

FFPE – Formalin-fixed paraffin-embedded

H and E – Hematoxylin and eosin

ICC – Immunocytochemistry

IHC – Immunohistochemical studies

USA – United States of America.

EDITORIAL/PEER-REVIEW STATEMENT

To ensure the integrity and highest quality of CytoJournal publications, the review process of this manuscript was conducted under a double-blind model (authors are blinded for reviewers and vice versa) through automatic online system.