Risk Assessment of Salivary Gland Cytological Categories of the Milan System: A Retrospective Cytomorphological and Immunocytochemical Institutional Study

Objective: The Milan System for Reporting Salivary Gland Cytology (MSRSGC) has been recently published to help communication between cytopathologists and clinicians. The aim was to assess our institutional experience with salivary gland fine needle aspiration cytology (FNAC) and the potential applicability of the MSRSGC for the estimation of the risk of neoplasm (RON) and risk of malignancy (ROM) for each category. Material and Method: Salivary gland FNAC procedures performed at NCI, Cairo University in a three-year period from 2016 to 2018 and had a corresponding histopathological diagnosis were included in the current study. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy were estimated. Histopathological final diagnosis was the gold standard. Cytological diagnoses were re-stratified according to MSRSGC with estimation of RON and ROM for each category. Results: A total of 118 cases were included in the current work. Sensitivity, specificity, PPV, NPV and accuracy were 84.6%, 88.2%, 78.6%, 91.8% and 87%, respectively. Cytological diagnoses were re-classified as non-diagnostic (2.5%), non-neoplastic (14.4%), atypia of undetermined significance (AUS) (6.8%), benign neoplasm (40.7%), salivary gland neoplasm of uncertain malignant potential (SUMP) (7.6%), suspicious for malignancy (8.5%), and malignancy (19.5%). The RON and ROM for each category were as follows: non-diagnostic (100%, 33.3%), non-neoplastic (17.6%, 11.8%), AUS (50%, 37.5%), benign neoplasm (97.9%, 2.1%), SUMP (88.9%, 44.4%), suspicious (90%, 60%), and malignancy (100% for each). Conclusion: The Milan System for Reporting Salivary Gland Cytology is a helpful classification system. The calculated ROM for each category of the studied cases was slightly above the published MSRSGC rates but still supported the recommended management for the patient.


INTRODUCTION
Salivary gland neoplasms represent 4% to 6.5% of all head and neck lesions.Malignancy rates of 82%, 43% and 25% have been recorded for the minor salivary gland, submandibular gland and parotid gland, respectively (1).Appropriate therapy of salivary gland tumors necessitates a precise preoperative diagnosis.Non-neoplastic lesions can be handled conservatively with medical therapy and followup, while neoplastic lesions require surgical intervention with major surgery for high grade malignancies (2).
The initial diagnostic workup of salivary gland lesions is based on ultrasound, computed tomography (CT) or magnetic resonance imaging (MRI) that determines the exact location of the lesion within the salivary gland and the imaging features of the nodules.They provide crucial information that aid in the surgical planning (3).However, imaging modalities failed to differentiate between benign and malignant lesions with confidence in most cases (4).
Fine-needle aspiration cytology (FNAC) has gained a wide acceptance among clinicians for the preoperative evaluation of salivary gland lesions over incisional biopsy which has the risk of fistula formation, tumor implantation and facial nerve damage in the parotid region (5).FNAC is a quick, economical and less invasive test that is easily applied in an outpatient setting (4).It is the favored diagnostic technique to differentiate non-tumorous lesions from tumors and to identify the malignant potential of the tumors with an accuracy ranging from 81% to 98%.This rate falls to 60-75% when specific tumor subtypes are considered (6).Unnecessary surgery could be obviated in about 33% of cases based on preoperative cytological diagnosis and thus can decrease the overall management cost of salivary gland tumors (7).The accuracy of salivary gland FNAC relies on various factors such as the aspiration technique (whether with image guidance or free handed), cytological preparation (whether conventional or liquid based), intra-tumor heterogeneity and experience of the cytopathologists (2).The considerable diversity of salivary gland tumors with overlapping morphological features as well as the rarity of these tumors create a major cytological interpretation challenge in some cases (3).Therefore, cytological interpretation that is descriptive without a definitive diagnosis could confuse the clinicians in the management choices (2).
Until recently, there was no uniform reporting system for interpretation of salivary gland lesions, which made it hard for clinicians to understand the reports and led to management dilemmas (7).To create a standardized practical reporting system that aids in the communication between clinicians and cytopathologists, advance patient care, as well as allow exchange of study data between various laboratories, the American Society of Cytopathology (ASC) and the International Academy of Cytology (IAC) have suggested a classification system: The Milan System for Reporting Salivary Gland Cytology (MSRSGC) (8).This system is similar to the cytological reporting systems of the thyroid, cervix, and the pancreaticobiliary, respiratory and urinary systems (5).MSRSGC contains six diagnostic categories that are associated with proposed risk of malignancy (ROM) and recommendations for clinical intervention (2).As this system is still novel, further studies are required to determine its effectiveness and ROM of each category of the system (2,4).
The aim of the current work was to determine the cytological variety of salivary gland lesions presented to the Egyptian National Cancer Institute (NCI) over a period of 3 years, determine the diagnostic accuracy of salivary gland FNAC for distinguishing benign from malignant lesions, re-classify the salivary gland cases based on MSRSGC criteria, and define the risk of neoplasm (RON) and risk of malignancy (ROM) for each category of MSRSGC in cases which had histopathological follow-up.

MATERIAL and METHODS
This was a retrospective three-year study (2016)(2017)(2018).Review of the registry of the Cytology Unit, Pathology Department, NCI, Cairo University in this period revealed 245 cases with cytological diagnoses of salivary gland lesions.Of these, only 118 cases (48.2%) had corresponding histopathological diagnosis and these were the cases included in the current study.Unavailability of corresponding histopathological diagnosis for any case may be due to non-surgical management of non-neoplastic cases and some benign neoplasms or due to escape of some patients from therapy.Informed consents were initially obtained from all patients for the cytological and surgical procedures and for the use of tissues for research purposes following the regulations of the Ethical Committee of the National Cancer Institute.
Relevant patients' demographic data including age, sex and the anatomical location of the lesion were recorded from patient files.Slides of the cases including cytology smear and cell block slides were retrieved from the archive of the Cytology Unit.All cytological aspirations were carried out using 23-gauge needles with an average of 2 to 3 passes depending on the size and yield of the lesion.Rapid onsite evaluation (ROSE) for the adequacy of smears was carried out at the time of aspiration using May-Grünwald Geimsa (MGG) stain on one slide.The remaining slides were immediately fixed in 95% ethyl alcohol.The retrieved smear slides were stained using modified Papanicolau stain and MGG, whereas the cell block sections were stained with H&E.In addition, immunocytochemical slides were available for some cases that required immunohistochemical verification to reach a definitive diagnosis at the time of the initial diagnosis.The slides of each case were reviewed to confirm the diagnoses and the cytological diagnoses were re-stratified according to the previously published criteria of MSRSGC (3,8) by the two authors without knowing the initial cytological interpretation or the final surgical diagnoses.
For statistical analysis, categories II (non-neoplastic), III (atypia of undetermined significance, AUS) and IVA (benign neoplasm) were combined in a negative group whereas categories IVB (salivary gland neoplasm of uncertain malignant potential, SUMP), V (suspicious for malignancy) and VI (malignancy) were combined in a positive group.This classification is based on the fact that the categories in each group have almost similar therapeutic management of their cases with minor discrepancies that might have little clinical implication.
Upon comparing cytological diagnosis with its histopathological counterpart, the cytological cases were additionally sub-classified into true positives, true negatives, false positives (interpreted inaccurately on cytology as positive and proved to be benign on excision) and false negatives (misdiagnosed cytologically as negative and turned out to be malignancy on histopathology).Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy of the cytological interpretation were estimated.Histopathological final diagnosis was considered as the gold standard.Furthermore, the corresponding histological follow-up of cytological cases was further differentiated into non neoplastic lesions, benign neoplasm and malignant tumors.The risk of neoplasm (RON) and the risk of malignancy (ROM) for each of the six categories of the MSRSGC were calculated.

RESULTS
This retrospective study included a total of 118 cases that underwent FNAC of salivary gland lesions during the study period and had corresponding histopathological follow-up (specimen or biopsy).Accordingly, the cytohistopathological association could be assessed in all studied cases.One hundred two cases (86.4%) were aspirated blindly and 16 cases (13.6%) were aspirated under image guidance.The age of the studied 118 cases ranged from 6 to 86 years with a mean of 48.2±14.71.Of the studied cases, 66 (55.9%) were male and 52 (44.1%) were female.The male to female ratio was 1.3:1.The most frequently involved salivary gland was the parotid gland (109 cases, 92.4%) followed by the submandibular gland (9 cases, 7.6%).No cases with minor salivary gland affection were identified in the current work.Most cases had left salivary gland involvement (65 cases, 55.1%).Forty seven cases (39.8%) had right gland lesions.Bilateral salivary gland lesions were noticed in 6 cases making up 5.1% of the cases.
Re-categorization of the salivary gland cytological diagnoses according to MSRSGC recommendations was performed.The distribution of the preoperative cytological interpretation in each diagnostic category of MSRSGC is identified in Table I and Table II.Overall, the rate of the non-diagnostic category (category I) was 2.5% (3/118).Non-neoplastic category (category II) was reported in 14.4% of cases (17/118) and chronic sialadenitis represented the most common cytological diagnosis (9/17).AUS category (category III) was noted in 6.8% of cases (8/118); most of these cases had a descriptive cytology report revealing various degrees of atypical lymphoid or epithelial cells but a definite diagnosis could not be made.The benign neoplasm category (category IVA) constituted the maximum number of cases with 48/118 cases (40.7%) and the most common diagnosis was pleomorphic adenoma (26/48) followed by Warthin's tumor (19/48) (Figure 1A,B-2).Category IVB (SUMP) accounted for 9/118 cases (7.6%)where it was

A B
hard to define the exact subtype of the neoplasm with the associated list of differential diagnoses.Suspicious of malignancy (category V) was reported in 10/118 cases (8.5%), while 23/118 cases (19.5%) were identified under the malignant category (category VI) (Figure 3;4A,B;5A-C; 6A-C).The most common interpreted malignant tumor was mucoepidermoid carcinoma (11/23).
After exclusion of non-diagnostic inadequate samples, concordance and discordance between cytologic and histopathologic diagnoses were calculated for the detection of critical cytologically diagnosed cases that required more serious and urgent therapy.Concordance was detected

Categories of Milan System
No.

Categories of the Milan System
No.

Metastatic squamous cell carcinoma (1)
MSRSGC: Milan System for Reporting Salivary Gland Cytopathology, SUMP: Salivary gland neoplasm of uncertain malignant potential.were estimated for each diagnostic category of MSRSGC (Table IV).The three cases of non-diagnostic interpretation on cytology proved to be Warthin's tumor, pleomorphic adenoma and metastatic papillary thyroid carcinoma on histopathological follow-up (Table I).IV).In the non-neoplastic category, one case was histopathologically diagnosed as basal cell adenoma and two were diagnosed as MALT lymphoma and mucoepidermoid carcinoma (Table I).Therefore, nonneoplastic category had RON of 17.6% and ROM of 11.8% (Table IV).For the category of AUS, the RON was 50% and the ROM was 37.5% (Table IV).Cases of the cytological AUS category were diagnosed histopathologically as one benign case of Warthin's tumor, and three malignant cases of mucoepidermoid carcinoma, acinic cell carcinoma and MALT lymphoma (Table I).Among cases in the cytological benign neoplasm category, one case was diagnosed histopathologically as non-neoplastic and one case as malignant tumor (epithelial/myoepithelial carcinoma) (Table I).Benign neoplasm category (IVA) revealed 97.9% RON and 2.1% ROM (Table IV).In the SUMP category, the RON was 88.9% and ROM was 44.4%.On resection, 4 of these SUMP cases were benign, 4 cases were malignant and one case was non-neoplastic (Table II).The category suspicious for malignancy had RON and ROM of 90% and 60% respectively.The RON and ROM for the malignant category were 100% each.

DISCUSSION
In an effort to standardize cytological terms and organize the therapeutic management of salivary gland lesions, the MSRSGC was proposed with detection of ROM for each diagnostic category and suggestions for management (3,10,11).In order to support the available published data of the MSRSGC and to report our institutional experience, we retrospectively assessed the accuracy of salivary gland FNAC over a three-year period, re-classified the cytological materials based on the criteria of the MSRSGC, and calculated RON and ROM for each category of MSRSGC.
In the present work, the diagnostic sensitivity, specificity, PPV and NPV of salivary gland FNAC were 84.6%, 88.2%,A C B 78.6% and 91.8% respectively.These were in accordance with a previous work in which authors reported sensitivity, specificity, PPV and NPV of 84%, 84%, 64% and 94%, respectively (1).In the literature, overall sensitivity of salivary gland cytology in most series was between 69.1% and 98% and specificity of approximately 88% to 100% was detected (3,4).The overall diagnostic accuracy achieved in our study was 87% which fell near the lower limit of the previously published range of 86% to 98% (4).
In the current work, three cases were categorized as nondiagnostic due to inadequate cellularity.The suggested MSRSGC management for cases in this category is to repeat FNA or use clinical and radiological correlation (8).Our cases were referred to excision without FNAC repetition; probably due to suspicious clinical or radiological findings that necessitated surgical intervention.One turned out to be metastatic malignancy and two were benign tumors.The rate of the non-diagnostic category (2.5%) in our work was similar to the rate of less than 10% proposed by the MSRSGC (8).Our result was in the range of 1.1% to 7.8% that was detected in a prior review study (10).Rossi et al. (7) reported an allowed range of 10% to 15% for this category's incidence.However in some reports, the rate reached up to 44% (12) or 50 % (9).The cause of these wide variations in the incidence of the non-diagnostic category between different studies might be related to the fact that the adequacy of FNA materials of the salivary gland lesions had no definite criteria and was widely related to variable quantitative and qualitative restrictions for a long period of time.Poor cellularity, non-mucinous cyst contents, needle positioning outside of the target nodule or improperly prepared and stained smears can be the reasons of nondiagnostic reports (2).Lack of clinical and radiological findings was also identified to be in this category (3).Recently, sixty cells representative of the target lesion with clinico-radiological correlation were suggested to be the key for adequacy based on MSRSGC regulations (2,8).The possible cause of the lower incidence rate in our work was related to the application of rapid on site evaluation (ROSE) of the yield to check the adequacy before discharging the cases.In this study, the non-diagnostic category had 100% RON and 33.3% ROM.A ROM of 25% was recommended by MSRSGC (8).Our ROM was much higher than that reported in another study (6.7%), but nearly similar to its RON (95.6%) (1).In a previous comparable study, a much lower RON of 64.5% and ROM of 16.1% were detected (2).Maleki et al., 2019 (9) noticed a RON of 34% and ROM of 10.6%.Some authors reported a ROM of 0% as they found no malignancy on surgical follow-up of these cases (3).The probable cause of the elevated ROM in the current study might be the small number of our studied cases.This could also be attributed to the fact that our institution is considered a referral center for malignant cases in Egypt.
Regarding the non-neoplastic category of the MSRSGC, 14.4% of our studied cases were found to be non-neoplastic on FNAC.In the literature, this value ranged from 5.1% to 53.4% (1,10  (8).Within the non-neoplastic category, the discordant cases with major clinical discrepancy on surgical follow-up included a case of MALT lymphoma and a case of mucoepidermoid carcinoma that were misdiagnosed cytologically as chronic sialadenitis and retention/benign cyst, respectively (false negative cases).Chronic sialadenitis as well as nonspecific sialadenosis are recognized pitfalls in salivary gland cytology as an associated malignancy might be not aspirated leading to a false negative diagnosis or presence of associated reactive cellular atypia leading to an over-diagnosis with possible recommendation of unnecessary surgery (1,13,14).A review of the smear of the first false negative case revealed polymorphous lymphoid cell population, frequent epithelioid histiocytic cells and scattered epithelial cells.The features were still favoring chronic sialadenitis.It is advised to use ancillary techniques for any lymphoid-rich aspirate to rule lymphoid neoplasm out and confirm the benign nature of the lesions (10).Mucoepidermoid carcinoma is the commonest malignant tumor and one of the most problematic neoplasms in cytological interpretation (13).Recognition of mucinsecreting, intermediate and squamous cells in smears is essential for a precise diagnosis.However, all these features are not clearly present in most cases.Cyst fluid aspiration with only mucinous background with scattered lymphocytes and rare mucus cells may cause this underdiagnosis (2).On reviewing the smears of our second false negative cases, there was mucin-like material in the background with scattered chronic inflammatory cells and debris.Based on the published recommendations, the possibility of low grade mucoepidermoid carcinoma cannot be ruled out in such situation (10).
In the present study, the bulk of the salivary gland lesions (40.7%) were in the benign neoplasm category.This figure was similar to that reported in previous studies (1)(2)(3).When the benign neoplastic cytological diagnoses were correlated with the corresponding final histopathological diagnoses, the calculated RON and ROM were 97.9% and 2.1%.This estimated ROM was consistent with the suggested rate of MSRSGC (less than 5%), which recommended conservative surgery or clinical follow-up as management for cases in such category (8) and ROMs in our study and other studies was related to the fact that the cytomorphological features of benign salivary gland tumors have been well described in the literature and are highly reproducible and also because of the fact that benign tumors are relatively common (12).These high RON and low ROM values could enable the clinicians to trust the cytological diagnosis and manage these cases confidently.In the present work, the cause of ROM in the benign neoplasm category was attributed to the presence of a case of low grade epithelial myoepithelial carcinoma on surgical follow-up that was cytologically misinterpreted as pleomorphic adenoma (false negative case).The reported ROM in other studies is caused by false negative interpretation predominantly of carcinoma ex pleomorphic adenoma (1,2) followed by low grade mucoepidermoid carcinoma (13,14), adenoid cystic carcinoma (9,14), epithelial myoepithelial carcinoma (14) and oncocytic carcinoma (13).Epithelial myoepithelial carcinoma can pose a diagnostic complexity to the cytopathologist.It shows a bimodal pattern of epithelial and myoepithelial cells.The appearance is usually dependent on the dominant cellular population (14).In our case, myoepithelial cells were the main cell population with hyaline basement material in the background giving the appearance of pleomorphic adenoma.At the same time, cellular atypia was mild.
In the present work, 19.5% of our studied cases were identified within the malignant category.An incidence rate of 13.8% was reported by others (2,9  2).In the current study, there was no false positive case in this category.
The possible cause might be related to the fact that the cytopathologists at our institution abide by the malignant morphological characteristic.Any case with uncertain malignant criteria was interpreted indeterminately with a descriptive report.Thus, our malignant results had high validity and reliability.
In spite of the adequacy of cytological smears, a definitive diagnosis is not possible even in experienced hands in some cases.It was noted that this indeterminate cytological interpretation accounted for more than 30% of salivary gland cytological diagnoses and fell into the MSRSGC categories of "AUS", "SUMP" or "suspicious for malignancy" (16).These categories form a major problem for the clinicians.
In the current work, the AUS category was noted in 6.8% of our cases and was similar to that reported by others (3).The MSRSGC has recommended that the "AUS" category be used wisely and that more effort is needed to decrease this category to less than 10% (8) 14) reported 10%, 10.8% and 11% occurrence rates for the AUS category, respectively.Among different institutions from the USA, Europe and China, the reported frequency varies from 0.7% to 17.0% (16).A much lower incidence of 0.6% was observed in another work (12).Among our studied cases, the RON was 50% and the ROM was 37.5% for the AUS category.Our results fell in the RON range of 41.7% to 100% and the ROM range of 0% to 75.1% that were documented in a previous work done on five different institutions (16).Our ROM was higher than the ROM of 20% proposed by MSRSGC (8).A ROM of 100% was demonstrated by Rohilla et al. (13).Like others (9), we noticed that the ROM of this category was found to be between the ROMs of the non-neoplastic/benign and malignant categories, which was in accordance with the category description.The relatively high ROM rate of this category in the present study supported the management suggestion of surgery in the clinical-radiological worrisome lesions and FNAC repetition in radiological non annoying cases (3).Careful assessment of the smears and paying attention to any specific features before reporting "AUS" could reduce the frequency of this category and probably lower the RON and ROM (2).In the present study, one case was reported as "atypical epithelial cells indefinite for neoplasm" and diagnosed as mucoepidermoid carcinoma on histopathology (false negative case).The only significant cytological feature in addition to the atypical epithelial cells after meticulous re-examination of smears was the presence of small amount of mucin in the background which was not mentioned in the initial report.Some authors have

Figure 1 :
Figure 1: Warthin's tumor of the parotid gland.A) Smear featuring monolayered sheets of oncocytic cells in a lymphoid background (Papanicolaou; x400).B) Cell block of the same case showing papillary structure lined by oncocytic cells with its core filled with lymphoid cells (H&E; x400).

Figure 2 :
Figure 2: Smear from a case of schwannoma of parotid gland formed of proliferating benign-looking spindle-shaped tumor cells embedded in an afibrillary eosinophilic matrix (Papanicolaou; x400).

Figure 3 :
Figure 3: Smear from a case of mucoepidermoid carcinoma showing sheets of squamous cells with dense scanty cytoplasm and scattered glandular cells having abundant fine vacuolated cytoplasm (Papanicolaou; x400).

Figure 4 :Figure 5 :B
Figure 4: Case of acinic cell carcinoma.A) Smear featuring sheet of acinic cells with round nuclei and abundant granular cytoplasm (Papanicolaou; x400).B) Positive immunologic reaction of tumor cells to CK7 on cell block section (IHC; x400).A B

Figure 6 :
Figure 6: Case of non Hodgkin's lymphoma of the parotid gland.A) Smear showing scattered atypical large round cells admixed with small reactive lymphocytes (Papanicolaou; x400).B) Cell block section showing tumor cells positively stained for CD20 (IHC; x400).C) Cell block with positive immunocytochemical staining of reactive small lymphocytes to CD5 with large tumor cells totally negative (IHC; x400).

Table I :
The distribution of preoperative cytological interpretation and final histopathological follow-up according to the categories I, II, II and IVA of MSRSGC.

Table II :
The distribution of preoperative cytological interpretation and final histopathological follow-up according to the categories IVB, V and VI of MSRSGC.

Table III :
Cyto-histopathological correlation of the 15 discordance cases.

Table IV :
The risk of neoplasm and risk of malignancy across categories of MSRSGC.
MSRSGC:Milan System for Reporting Salivary Gland Cytopathology, AUS: Atypia of undetermined significance, SUMP: Salivary gland neoplasm of uncertain malignant potential, RON: Risk of neoplasm, ROM: Risk of malignancy.