The Contributory Role of Cell Blocks in Salivary Gland Neoplasms Fine Needle Aspirations Classified by the Milan System for Reporting Salivary Gland Cytology

(1) Background: The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) was introduced in 2018, bringing an organ-specific classification system for salivary gland cytopathology. The aim of present study is to evaluate the MSRSGC prospectively, based on a two-year experience in the tertiary care center pathology department, and evaluate the role of routine cell block (CB) preparation in salivary gland cytopathological diagnostics. (2) Methods: In our institution, the Department of Pathology, Fimlab Laboratories, Tampere, Finland, the MSRSGC has been implemented in salivary gland cytopathology since January 2018 and, over a two-year period (January 2018–December 2019), there were 365 fine-needle aspirations, of which 164 had a surgical follow-up. The CB methods used were Plasma-thrombin, the collection of visible fragments, and the Shandon and in-house methods. (3) Results: The MSRSGC diagnostic figures were as follows: accuracy 87.5%, sensitivity 45.8% and specificity 98.9%. For diagnostic categories of MSRSGC (non-neoplastic, benign neoplasm and malignant neoplasm) (n = 63) diagnostic accuracy was 98.4%, and for undetermined categories (atypia of undetermined significance, salivary gland neoplasm of uncertain malignant potential and suspicious for malignancy) (n = 49) diagnostic accuracy was 73.5%. Non-contributory cell blocks resulted more often in a false negative diagnosis (25%, 3/12) than a true negative diagnosis (10%, 7/73, p < 0.001), and is, most likely, an insufficient cytological diagnosis (86%, 18/21, p < 0.001). (4) Conclusion: The application of MSRSGC and CBs are beneficial in salivary gland cytological diagnosis, increasing diagnostic accuracy and, thus, patients’ management and treatment.


Introduction
The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) was introduced in 2018 following other organ specific cytopathological reporting. It aimed for better patient care, bringing a practical, evidence-based, user-friendly classification system with characterization and management algorithms [1,2]. Several international studies have stated that the MSRSGC is a reliable tool for salivary gland cytopathology categorization [3][4][5][6][7]. Cell blocks (CBs) are a collection of sediments and visible pieces of tissue from cytological specimens that are concentrated and processed into paraffin blocks, and stained with hematoxylin-eosinike surgical specimens. Various techniques can be used to prepare CBs, with each method having their advantages and disadvantages [8][9][10]. In our institution, cytological material is triaged, and one of the following CB methods is applied: the plasma-thrombin method, the collection of visible tissue fragments, the in-house method [8,11] or the Shandon method. The risk of malignancy was 22.0% for all salivary gland lesions with surgical follow up. For each MSRSGC category, the risk of malignancy was as follows: non-diagnostic (n = 52) 23.1%, non-neoplastic (n = 4) 25.0%, atypia of undetermined significance (n = 25) 4.0%, benign neoplasm (n = 53) 0.0%, neoplasm of uncertain malignant potential (n = 27) 29.6%, suspicious for malignancy (n = 6) 83.3% and malignant neoplasm (n = 6) 100.0% ( Table 1). The most common benign neoplasm was Warthin's tumor (n = 49) (Figure 1), and the most common malignant neoplasm was carcinoma ex PA (n = 7), forming 36.3% and 5.2% of all salivary gland neoplasms, respectively. All tumors are listed according to MSRSGC categories in Supplementary Table S2.

Results
During a two-year period, 365 salivary gland FNAs were diagnosed according to the MSRSGC, and 164 (44.9%) of them had a surgical follow-up with histopathological diagnosis. The topographical localization was in the parotid gland in 139 (82%) cases and the submandibular gland in 25 (18%) cases. Detailed distribution according to MSRSGC categories can be found in Supplementary Table S1. In total, 86 (52.4%) of the cases were male and 78 (47.6%) were female. The average lesion size was 2.3 cm (0.8-5.0 cm), the average patient age was 58.3 (13-95) years and the clinical characteristics for each MSRSGC category are presented in Table 1. The risk of malignancy was 22.0% for all salivary gland lesions with surgical follow up. For each MSRSGC category, the risk of malignancy was as follows: non-diagnostic (n = 52) 23.1%, non-neoplastic (n = 4) 25.0%, atypia of undetermined significance (n = 25) 4.0%, benign neoplasm (n = 53) 0.0%, neoplasm of uncertain malignant potential (n = 27) 29.6%, suspicious for malignancy (n = 6) 83.3% and malignant neoplasm (n = 6) 100.0% (Table 1). The most common benign neoplasm was Warthin's tumor (n = 49) (Figure 1), and the most common malignant neoplasm was carcinoma ex PA (n = 7), forming 36.3% and 5.2% of all salivary gland neoplasms, respectively. All tumors are listed according to MSRSGC categories in Supplementary Table S2. The FNAs accuracy, sensitivity and specificity in differentiating between benign and malignant neoplasms (n = 112) were 87.5%, 45.8%, and 98.9%, respectively. For diagnostic categories (non-neoplastic, benign neoplasm and malignant neoplasm) (n = 63) the diagnostic accuracy was 98.4%, and for undetermined categories (atypia of undetermined significance, salivary gland neoplasm of uncertain malignant potential and suspicious for malignancy) (n = 49) the diagnostic accuracy was 73.5% (Table 2). The FNAs accuracy, sensitivity and specificity in differentiating between benign and malignant neoplasms (n = 112) were 87.5%, 45.8%, and 98.9%, respectively. For diagnostic categories (non-neoplastic, benign neoplasm and malignant neoplasm) (n = 63) the diagnostic accuracy was 98.4%, and for undetermined categories (atypia of undetermined significance, salivary gland neoplasm of uncertain malignant potential and suspicious for malignancy) (n = 49) the diagnostic accuracy was 73.5% (Table 2). Deviations of non-neoplastic, benign neoplastic and malignant neoplastic cases are illustrated in Figure 2. Deviations of non-neoplastic, benign neoplastic and malignant neoplastic cases are illustrated in Figure 2. The present study material consisted of 111 histologically proven neoplastic cases with the CBs available. CBs with less than 10 cells were all (n = 28) non-contributory for diagnosis, and those with 10-50 cells (n = 24) were crucial for diagnosis in 21%, supported the diagnosis in 75% and were non-contributory in 4% of cases (p < 0.001) ( Table 3).
Out of the four methods used for the preparation of the CBs, the Shandon method was the most likely (4/9 cases) to be non-contributory (p = 0.047). Non-contributory CBs resulted most often in the MSRSGC insufficient category (18/21 of insufficient cases, p < 0.001) and the least in the benign neoplasm category (2/47 cases of benign neoplasm category, p < 0.001), as shown in Table 3 and illustrated in Figure 3. The present study material consisted of 111 histologically proven neoplastic cases with the CBs available. CBs with less than 10 cells were all (n = 28) non-contributory for diagnosis, and those with 10-50 cells (n = 24) were crucial for diagnosis in 21%, supported the diagnosis in 75% and were non-contributory in 4% of cases (p < 0.001) ( Table 3).
In-house method Out of the four methods used for the preparation of the CBs, the Shandon method was the most likely (4/9 cases) to be non-contributory (p = 0.047). Non-contributory CBs resulted most often in the MSRSGC insufficient category (18/21 of insufficient cases, p < 0.001) and the least in the benign neoplasm category (2/47 cases of benign neoplasm category, p < 0.001), as shown in Table 3 and illustrated in Figure 3.    Thirteen false negative cases and one false positive case regarding malignancy were reported during the two-year study period. Only 1/14 of false cases were in the determined category. Extranodal marginal zone B-cell lymphoma of MALT type was diagnosed in the non-neoplastic category of the MSRSGC ( Table 4). The CB, in this case, was Thirteen false negative cases and one false positive case regarding malignancy were reported during the two-year study period. Only 1/14 of false cases were in the determined category. Extranodal marginal zone B-cell lymphoma of MALT type was diagnosed in the non-neoplastic category of the MSRSGC (Table 4). The CB, in this case, was non-contributory. In the AUS category of MSRSGC there were four false-negative cases: Follicular lymphoma (n = 2), Adenoid cystic carcinoma (n = 1) and Myoepithelial carcinoma ex PA (n = 1). Cell blocks were contributory in three cases and supported the diagnosis in one case ( Figure 5A,B). When taking CB performance into account, CBs were crucial for or supported diagnosis most commonly in true-negative (90%, 66/73) and true-positive (80%, 4/5) cases. CBs were less likely to be crucial for diagnosis or support diagnosis in false-negative (75%, 9/12) cases. Moreover, CBs were least likely to be crucial for diagnosis in AUS (1/11) and SUMP (7/25) categories, promoting their uncertain nature. In insufficient cases (n = 21), only three CBs supported diagnosis (only normal tissue) which draws a link between insufficient cytological diagnosis and non-contributory cell blocks (p < 0.001). In a separate analysis, the use of ICC in CBs was also analyzed, and its contribution to the final cytological diagnosis is summarized in Table 5. ICC was contributory in cellular CB samples ( Figure 6) and the SUMP category with p ˂ 0.001.   The majority of the false-negative diagnoses (8/13) were in the SUMP category: Carcinoma ex pleomorphic adenoma (n = 3), Acinic cell carcinoma (n = 2), Adenoid cystic carcinoma (n = 2) and Myoepithelial carcinoma ex pleomorphic adenoma (n = 1). Interestingly, and to highlight the overlapping cytopathological features of salivary gland neoplasms, there were also (n = 9) pleomorphic adenomas in the SUMP category ( Figure 5C,D). In false negative cases, CBs were crucial for diagnosis in four cases (all cases of Carcinoma ex pleomorphic adenoma and Myoepithelial carcinoma ex pleomorphic adenoma) and supported the diagnosis in four cases.
When taking CB performance into account, CBs were crucial for or supported diagnosis most commonly in true-negative (90%, 66/73) and true-positive (80%, 4/5) cases. CBs were less likely to be crucial for diagnosis or support diagnosis in false-negative (75%, 9/12) cases. Moreover, CBs were least likely to be crucial for diagnosis in AUS (1/11) and SUMP (7/25) categories, promoting their uncertain nature. In insufficient cases (n = 21), only three CBs supported diagnosis (only normal tissue) which draws a link between insufficient cytological diagnosis and non-contributory cell blocks (p < 0.001).
In a separate analysis, the use of ICC in CBs was also analyzed, and its contribution to the final cytological diagnosis is summarized in Table 5. ICC was contributory in cellular CB samples ( Figure 6) and the SUMP category with p < 0.001. 5C,D). In false negative cases, CBs were crucial for diagnosis in four cases (all cases of Carcinoma ex pleomorphic adenoma and Myoepithelial carcinoma ex pleomorphic adenoma) and supported the diagnosis in four cases. When taking CB performance into account, CBs were crucial for or supported diagnosis most commonly in true-negative (90%, 66/73) and true-positive (80%, 4/5) cases. CBs were less likely to be crucial for diagnosis or support diagnosis in false-negative (75%, 9/12) cases. Moreover, CBs were least likely to be crucial for diagnosis in AUS (1/11) and SUMP (7/25) categories, promoting their uncertain nature. In insufficient cases (n = 21), only three CBs supported diagnosis (only normal tissue) which draws a link between insufficient cytological diagnosis and non-contributory cell blocks (p < 0.001). In a separate analysis, the use of ICC in CBs was also analyzed, and its contribution to the final cytological diagnosis is summarized in Table 5. ICC was contributory in cellular CB samples ( Figure 6) and the SUMP category with p ˂ 0.001.

Discussion
The cytopathological diagnosis of salivary gland lesions should take into account clinical features such as patient symptoms, physical examination and imaging, as well as lesion site and tumor frequencies [12]. Surgical management is decided in multidisciplinary meetings based on the above lesion characteristics in our hospital.
The MSRSGC, as a unified system, has defined diagnostic categories with predetermined ROMs and management recommendations. Furthermore, it enhances communication between pathologists and clinicians locally, nationally and internationally. [5,13].
During a two-year period of study, the overall accuracy of the MSRSGC in distinguishing between benign and malignant neoplasms in our institution was 87.5%. When only viewing the diagnostic categories of the MSRSGC (non-neoplastic, benign neoplasm and malignant neoplasm), accuracy was as high as 98.4%. In the three MSRSGC undetermined categories (AUS: atypia of undetermined significance, SUMP: neoplasm of uncertain malignant potential, suspicious for malignancy), the accuracy was 73.5%. Sensitivity in all categories was only 45.8%, but as high as 85.7% in diagnostic categories and as low as 29.4% in undetermined categories. Specificity was 98.9% in all categories, 100% in diagnostic categories and 96.9% in undetermined categories. The MSRSGC offers the undetermined categories between defined benign and malignant categories necessary for better management of cases with overlapping cytopathological features, due to heterogeneity of neoplastic lesions, neoplastic cell resemblance to normal salivary gland cells, metaplastic and/or cystic changes, and capsular invasion being impossible to evaluate in FNA material [13][14][15].
CBs are complimentary to conventional cytology, and are beneficial in several scenarios: (1) adding architecture dimensions to cytological material; (2) immunohistochemical characterization of cytological material; and (3) molecular techniques for further characterization of cytological material that may have a central role in targeted therapy in the future [8,[16][17][18]. Nevertheless, CBs should not be a substitute for cytology [17].
CBs were reported as essential in head and neck cystic and metastatic lesions [19][20][21], but less data are available in salivary gland cytology [22], particularly in the salivary gland lesions classified by the MSRSGC [23,24]. In some MSRSGC institutional analyses, where CBs are routinely performed, their role in the diagnostic work-up was not evaluated [25,26]. In our present study, CBs were crucial for or supported the diagnosis in 90% of truenegative cases and in 80% of true-positive cases. Behaeghe et al. retrospectively analyzed 359 salivary gland samples processed only as a Cellient CB in view of the MSRSGC, with an overall accuracy of 92.9%, sensitivity of 75.9%, specificity of 97.9%, PPV of 91.7% and NPV of 95%, respective to the diagnostic categories (excluding non-diagnostic, AUS and SUMP categories) [23].
The limited cellularity of FNA specimens is a common limitation for CBs [10,27]. In our setting, radiologists in training are widely involved in FNAs and, due to the learning curve, there are high rates within the insufficient category. Pre-analytical issues, including triaging of cytological material, are crucial steps in the cytological diagnostic work-up [8,28,29]. In thyroid gland AUS/FLUS categorized specimens, cytological material triaging had a diagnostic input [28]. In the present study, CBs were contributory to (crucial for or supporting) the diagnosis in 72.7% of AUS cases, 95.7% of benign neoplasm cases, 88% of SUMP cases, 66.7% of suspicious for malignancy cases and 50% of malignant neoplasm cases, respectively. The CB contribution was statistically significant in the benign neoplasm category, where the CB was contributory to the diagnoses. On the other hand, in the non-diagnostic category, the non-contributory CB role was also statistically significant. Cellient CB processed in view of the MSRSGC, with a 33% non-diagnostic rate and 13.8% ROM in this category [23], supports the importance of triaging the cytological material to decrease non-diagnostic rates.
The need of ancillary techniques is increasing with the growing amount of known genetic mutations and rearrangements in salivary gland tumors [30]. This will further increase the diagnostic accuracy of salivary gland FNA and the MSRSGC. CBs are feasible as a material for ancillary molecular techniques [29]. In the European Federation of Cytology Societies survey, CBs were used for immunohistochemistry in 38% of laboratories, and 60% used a combination of material for immunohistochemistry [31]. Ancillary techniques were not deeply analyzed in the light of MSRSGC. In a Dubucs et al. study of 328 cases, immunohistochemistry contributed to a definitive diagnosis in 23.7% of cases and FISH in 33% of cases. Both techniques were applied on smears [32].
In practice, lymphoproliferative diseases are commonly misinterpreted in the FNA material of salivary glands. In our series, in addition to a false-positive case of extranodal marginal zone B-cell lymphoma of MALT type diagnosed in a non-neoplastic category, there were two false-negative follicular lymphomas in an AUS category. In all three cases, the CBs were non-contributory and contained less than 10 cells. In a series of 6249 reviewed cases performed by the College of American Pathologists, the lymphoma cases featured the highest false-negative rate at 57% [33]. In recent MSRSGC analyses, 10 lymphoma cases were misinterpreted in a non-neoplastic category and 9 lymphoma cases in an AUS category [4,23,25,26,34].
Another cytological pitfall is carcinoma ex pleomorphic adenoma (PA), which was the most common false-positive entity classified in undetermined categories in our series: one case of myoepithelial carcinoma ex PA was placed in a AUS category and another myoepithelial carcinoma ex PA in a SUMP category. In addition, the SUMP category contained three cases of carcinoma ex PA. The CBs in these cases were cellular, and either supported the neoplastic diagnosis or were crucial for the neoplastic diagnosis. To summarize other MSRSGC analyses, carcinoma ex PA cases were misinterpreted in a non-neoplastic category in three cases [6,23,35]: in an AUS category in one case [36] and in a SUMP category in only one case [4], but placement in a SUMP category is not generally rated as misinterpretation. Notably, nine cases were in a benign neoplasm IVA category [4,[23][24][25]. On the other hand, 4 cytological diagnoses of carcinoma ex PA in a study by Pujani et al. were, histologically, a salivary duct carcinoma, squamous cell carcinoma, mucoepidermoid carcinoma and carcinoma NOS [36].
Interestingly, in one false-positive case of adenoid cystic carcinoma categorized as AUS, the CB was non-contributory, with cellularity less than 10 cells. In contrast, two cases each of adenoid cystic carcinoma and acinic cell carcinoma, categorized as SUMP, had cellular CBs which supported the neoplastic diagnoses in the present study. Of note, 43% of adenoid cystic carcinomas, acinic cell carcinomas and mucoepidermoid carcinomas were in a SUMP category in a large international multi-institutional study [37]. In an institutional study from the Memorial Sloan Kettering Cancer Centre, no acinic cell carcinoma nor adenoid cystic carcinoma were categorized as either a non-neoplastic or AUS category, but 33.3% of acinic cell carcinomas and 50% of adenoid were in a SUMP category [38].

Conclusions
The application of the MSRSGC and CBs are beneficial in salivary gland cytological diagnosis, increasing diagnostic accuracy and, thus, patients' management and treatment. This study is limited by the short follow-ups of cases, as the MSRSGC was introduced in 2018 [1,7]. To increase communication with clinicians, false-positive result rates were recently recommended to be part of cytopathology reporting. In general, the false-positive rate is 10% in salivary gland FNAs [39].
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/diagnostics11101778/s1, Table S1: Topography of Salivary Gland Fine-Needle Aspirations, Table S2: MSRSGC: Histopathology Diagnoses of Neoplasms. Data Availability Statement: All data generated or analysed during this study are included in this article and its Supplementary Material files. Further enquiries can be directed to the corresponding author (I.K.).

Conflicts of Interest:
The authors have no conflict of interest.