Histopathological Review of Diagnostic Categories of the Milan System for Reporting Salivary Gland Cytopathology - An Institutional Experience of 6 Years.

INTRODUCTION: Salivary gland malignancies account for 2 to 4% of head and neck cancers. Fine needle aspiration cytology (FNAC) is used in preoperative diagnosis of salivary gland lesions. Although FNAC is a highly reliable technique for preoperative diagnosis, there were no consensus on salivary gland cytopathology reporting. Recently, an international group has recommended a classification system for salivary gland FNAC reporting titled "Milan System for Reporting Salivary Gland Cytopathology" (MSRSGC). In this study, we aimed to evaluate the usability of the Milan System, its ability to determine the risk of malignancy for each category, with comparisons of inital cytologic and final histopathological diagnosis.
MATERIALS AND METHODS: We performed a retrospective analysis of salivary gland lesion FNAC in our department from 2013 to 2019. A total of 578 FNACs were performed in 514 patients. Of these, 85 cases had surgical follow-up (parotid gland, n = 73, submandibular gland, n = 12). The cytological samples were categorized according to the MSRSGC into six categories by two pathologists. The risk of malignancy (ROM) and diagnostic accuracy values were calculated for each diagnostic categories.
RESULTS: A total of 85 aspirates of the patients with follow-up, the MSRSGC diagnostic categories were as follows: non-diagnostic in 7 aspirates (8.2%), non-neoplastic in 3 (3.5%), atypia of undetermined significance (AUS) in 9 (10.5%), benign neoplasm in 43 (50.5%), salivary gland neoplasm of undetermined malignant potential in 7 (8.2%), suspicious for malignancy in 10 (11.7%), and malignant in 6 (7%). The ROM for each category was 28, 5%, 0%, 33%, 0%, 28.5%, 90%, and 100%, respectively.
CONCLUSION: FNAC plays a critical role in the evaluation of patients with salivary gland lesions. The MSRSGC helps in the standardization of the process of diagnosis and clinical management of salivary gland lesions, especially of AUS and SUMP categories that are indeterminate categories in nature. Copyright:

INTRODUCTION

Malignant salivary gland neoplasms account for approximately 3% to 5% of all head and neck cancers.[1] The fine needle aspiration cytology (FNAC) of salivary gland lesions has high sensitivity and predictive potential in the evaluation of salivary gland lesions. It is a rapid, simple, and cost-effective test that can improve the communication between clinicians and pathologists, leading to a consistent approach for the preoperative management of the patients.

Although FNAC for salivary gland cytology is a worldwide established method, there was no classification system available until 2017. The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) was introduced in 2017 for salivary gland aspirates and the properties for each of the six categories were listed in detail. These categories are listed as non-diagnostic, non-neoplastic, atypia of undetermined significance (AUS), benign neoplasm, salivary gland neoplasm of undetermined malignant potential (SUMP), and suspicious for malignancy (SM) and malignant.[2]

The MSRSGC provides a standardized reporting scheme of salivary gland lesions and standardized clinical management of patients. In our study, we aimed to evaluate the diagnostic accuracy and the risk of malignancy values for each category.

MATERIALS AND METHODS

The FNAC materials of the cases who presented to our pathology department between 2013 and 2019 were reviewed. The data of 578 FNAC procedures were obtained with repeated aspirates of 514 different patients. Among these, 85 patients underwent surgery in our Head and Neck Surgery Department. In our study, we included cases that underwent surgery and excluded cases that had missing clinical data or radiological data. FNACs usually were performed under ultrasound guidance by a radiologist along with a pathologist. An average of five passes were performed using a 22–23 gauge needle. The aspirated material was ideally spread over two slides; one slide was dried in air and stained with May Grunwald Giemsa stain; the other slide was fixed in alcohol and was stained with Papanicolaou stain. The remaining portion was processed for cell-block preparation that was submerged in a 9:1 mixture of 96% ethyl alcohol and 4% buffered formalin. The cytological specimens that had surgical follow-up were classified based on MSRSGC by two pathologists blinded to the original cytological and histopathological diagnosis.

The risk of malignancy (ROM) and the frequency of each category were determined. The ROM value was determined by dividing the number of cases that were histopathologically diagnosed as malignant in a category to the total number of cases in the same category. To calculate the sensitivity, specificity, positive predictive rate, and negative predictive rate, we divided the cases into benign and malignant groups.

Ethical approval for this study was obtained from the Ethics Committee of Marmara University (reference number-09.2019.858).

RESULTS

A total of 578 FNAC procedures of 514 patients were performed in our department between 2013 and June 2019. Of these, 85 cases had surgical follow-up. The age of these patients in the surgical follow-up category ranged from 6 to 91 years, with a mean age of 49.09 years. There were 41 females and 44 males. The recategorization of cases according to the Milan System was non-diagnostic in 7 (8.2%), non-neoplastic in 3 (3.5%), AUS in 9 (10.5%), benign neoplasm in 43 (50.5%), SUMP in 7 (8.2%), suspicious for malignancy in 10 (11.7%), and malignant in the remaining 6 (7%) patients [Table 1]. The distribution of cases in these categories with localization and follow-up histopathological diagnosis are presented in Table 2.

Table 1

The recategorization of cases according to MSRSGC

	n	(%)
Nondiagnostic	7	(8.2)
Nonneoplastic	3	(3.5)
AUS	9	(10.5)
Benign neoplasm	43	(50.5)
SUMP	7	(8.2)
Suspicious for malignancy	10	(11.2)
Malignant	6	(7)
Total	85 (100)	(100)

Table 2

Comparison of MSRSGC diagnoses with histopathological diagnoses and tumour sites

MSRSGC	Histopathological diagnosis	Tumor site	Number of cases
ND (n = 7)	Chronic sialadenitis	Parotid	1
	Chronic sialadenitis	Parotid	1
	Normal salivary gland tissue	SMG	1
	Granulocytic sarcoma	Parotid	1
	AdCC	Parotid	1
	Hemangioma	Parotid	1
	WT	Parotid	1
NN (n = 3)	Normal salivary gland tissue	SMG	1
	Normal salivary gland tissue	Parotid	1
	Chronic sialadenitis	Parotid	1
AUS (n = 9)	MALT lymphoma	Parotid	1
	Cellular PA	Parotid	1
	PA	Parotid	1
	Sclerosing polycystic adenosis	Parotid	1
	MEC	Parotid	1
	Chronic granulomatous sialadenitis	SMG	1
	Chronic sialadenitis	SMG	1
	PA	SMG	1
	Hodgkin’s lymphoma	SMG	1
BN	PA	Parotid	30
BN (n = 43)	WT	Parotid	11
	PA	SMG	1
	Lipoma	Parotid	1
SUMP (n = 7)	BCA	Parotid	1
	BCA	Parotid	1
	BCA	Parotid	1
	BCAC	Parotid	1
	Cellular PA	Parotid	1
	Salivary duct carcinoma	Parotid	1
	Oncocytoma	Parotid	1
SM (n = 10)	AdCC	Parotid
	WT	Parotid	1
	ACC	Parotid	1
	Myoepithelial carcinoma	Parotid	1
	Secretory carcinoma	Parotid	1
	AdCC	SMG	1
	MEC	SMG	1
	SCC	SMG	1
	Lymphoepithelial carcinoma	SMG	1
M (n = 6)	Nasopharyngeal carcinoma	Parotid	1
	AdCC	Parotid	1
	ACC	Parotid	1
	MEC	Parotid	1
	SCC	Parotid	1
	DBCL	Parotid	1

MSRSGC, Milan System for Reporting Salivary Gland Cytopathology; ND, non-diagnostic; NN, non-neoplastic; AUS, atypia of undetermined significance; BN, benign neoplasm SUMP, salivary gland neoplasm of uncertain malignant potential; SM, suspicious for malignancy; M, malignant SMG, submandibular gland; AdCC, adenoid cystic carcinoma; DBCLL, diffuse B cell lymphoma; ACC, acinic cell carcinoma MEC, mucoepidermoid carcinoma: SCC, squamous cell carcinoma; WT, Warthin tumor; PA, pleomorphic adenoma; BCA, basal cell adenoma BCAC: basal cell adenocarcinoma

The calculated ROM for each category were 28.5%, 0%, 33%, 0%, 28.5%, 90%, and 100%, respectively, and these date are presented in Table 3. We have also determined the diagnostic accuracy for benign and malignant cases.

Table 3

The risk of malignancy (ROM) according to MSRSGC classification

	ROM (%)
Nondiagnostic	28.5
Nonneoplastic	0
AUS	33
Benign neoplasm	0
SUMP	28.5
Suspicious for malignancy	90
Malignant	100

DISCUSSION

In our study, salivary gland lesions were most frequently found in the parotid gland. The age of the cases ranged from 6 to 91 years, consisting of 41 females and 44 males. These data are similar to those reported by previous studies.[3456] Viswanathan et al.[4] reported a male:female ratio of 1:1.2 with an age range of 13 to 100 and Rohilla et al.[6] reported a male:female ratio of 1.7:1 with an age range of 1 to 95 years.

Diagnostic sensitivity and specificity in our study were 85.71% and 89.66%, respectively, and these values were generally compatible with previously published values in the literature.[3789] Song et al. reported diagnostic sensivity and specifity in malignant cases to be 84.8% and 94.1%, respectively.[7] Hughes et al.[8] had diagnostic sensivity and specifity of 73% and 91%, respectively.

The Milan system suggests non-diagnostic category to represent less than 10% of all cases. In our study, the rate was 8.2% and most of the cases were diagnosed as chronic sialadenitis (28.5%). Kala et al.[9] also reported several cases in this category with chronic sialadenitis as the leading diagnosis. The authors of these recent publications emphasized that there was wide variation among the cases in this category.[5910] Wei et al.[5] had 100 cases in non-diagnostic category, and review of histopathological diagnosis revealed that 31% of the cases were non-neoplastic lesions, 44% were benign neoplasms, and 25% were malignant neoplasms. Singh et al.[10] reported three cases in this category with diagnoses of a benign cyst, chronic sialoadenitis, and Whartin tumor.

Our current study had three cases in non-neoplastic category, which represented 3.5% of the total cases. In the literature, again, there was a wide variation among these rates as well.[69101112]

Rohilla et al.,[6] Kala et al.,[9] Sadullahoglu et al.,[11] and Gaikwad et al.[12] reviewed cases with histopathological correlation and found that non-neoplastic category represented 22.3%, 11.6%, 24.4%, and 31.7% of all their cases, respectively. The rate of non-neoplastic category was higher in studies that had only cytological cases such as Rohilla et al.'s[6] study, which reported 55.8% of non-neoplastic rate, and the lower rate of 3.5% in our study might be due to study design because non-neoplastic cases are not generally excisionally biopsied and our study only included cases that had surgical follow-up.

In our study, two out of three (66.6%) non-neoplastic cases were diagnosed as normal salivary gland tissue and the other as chronic sialoadenitis in histopathological diagnosis. These cases were initially diagnosed as a mass radiologically. Excisions were total submandibular gland exicision at one case and two superficial parotidectomies containing the lesions according to the radiologic and surgical data. Excisions were performed because of radiological–cytological discrepancy. Follow-up of these patients revealed no complaint or any sign of disease.

Histopathological diagnoses of three out of nine total cases (33.3%) in the AUS category were pleomorphic adenoma. In similar studies in the literature, lymphomas are generally the most malignant final diagnosis in this category.[1314]

Of the 43 cases in the benign neoplasm category, 31 were diagnosed as pleomorphic adenoma and 11 were diagnosed as Warthin tumor. In previous studies, it was reported that most of the cases in benign neoplasm category were pleomorphic adenoma.[3561314] This is expected and compatible with our study because pleomorphic adenoma is the most common benign neoplasm of salivary glands.

Moreover, in our study, 8.2% of cases were under the SUMP category. This value was similar to 8.2% and 7.6% that were reported in previous works.[1314] There were seven cases in the SUMP category in our study, including three cases of basal cell adenomas. The MRSGCC recommends surgical management in most cases of SUMP.[2] In the following years, if this reporting system is embraced in many other institutes and more studies comparing histopathological diagnosis with cytological data are done, more detailed databases of this category will be available, and thus more reliable results can be achieved.

Furthermore, in our current study, we found SM cases to be 11.2% of all the cases. This value was higher compared to that in other published studies.[7915] In their studies that did not have a cytohistopathological correlation, Song, et al.[7] and Kala et al.[9] reported that SM cases accounted for 2.7% and 2.4% of all their total cases, respectively.A rate of 8.5% was reported by Hafez et al.[15], and the study explained this situation by the overestimated rates of surgery due to the cytology reports with atypical and undefined features. Finally, six cases evaluated in the malignant category were the most heterogeneous category, as shown in Table 2, with variation in histopathological diagnoses. This category represented 7% of all cases in our study and this rate varies in the literature, reported between 19.5% and 9.2%. In the studies of Hafez[15] and Song et al.,[7] ROM values were 100% and 98.5%, respectively, similar to our study.

The ROM value for each category was calculated as 28.5%, 0%, 33%, 0%, 28.5%, 90%, 100%, respectively. The ROM values of similar studies are summarized in Table 2. In the study of the Milan group, these rates were found as 25%, 10%, 20%, <5%, 35%, 60%, and 90%, respectively.[2] When compared with other studies, it is seen that although ROM values are 0% in the benign category, this value reaches 100% in the malignant category. It is possible to explain this finding with the fact that the diagnostic criteria are very clear in the “determinate”categories unlike ambiguities of indeterminate categories that cause diagnosic differencies and interobserver variability. Another striking point is the high ROM value in the non-diagnostic category. Non-diagnostic category in the MSRGCC guide is determined as hypocellular or poorly prepared specimens with dense artifacts, specimens containing nonneoplastic salivary gland tissue obtained from a radiologically/clinically mass-forming lesion, and non-mucinous cyst fluids without epithelial component. We used those criteria specified in the guide while evaluating our cases. It was suggested that these cases should comprise less than 10% of all cases.[2] In our study, the non-diagnostic category represented 8.2% of all cases and the ROM value was 28.5%. In management of these cases, the Milan guideline recommends aspiration repetition or a clinico-radiological correlation.[2] Because we only included patients who underwent surgery in our study, the final histopathological diagnoses were as follows: two chronic sialadenitis, one Warthin tumor, one hemangioma, one adenoid cystic carcinoma, one normal salivary gland tissue, and finally one granulocytic sarcoma. When we re-evaluated the aspirate slides of these cases, we found that the slides of both adenoid cystic carcinoma and chronic sialadenitis were insufficient preparations for a diagnosis. The cellular elements in the smears of granulocytic sarcoma and chronic sialadenitis cases were representing normal salivary gland tissue. The aspiration material of the Warthin tumor only had cyst content [Figure 1a], and re-aspiration was suggested in the pathology report. However, surgical excision was done due to the incompatibility of the pathologic and radiologic data [Figure 1b]. Based on these data and results, we can draw the following conclusion: clinican–pathologist–radiologist communication has to be particularly strong in cases of this category, and in times of insufficient aspiration material or clinical–radiological incompatibility, one should not assume that the lesion was not biopsied as it could very well be representative of the lesion just like of a Warthin tumor cyst content. Or it was and the result is negative for neoplasm like our cases of chronic sialoadenitis and granulocytic sarcoma, which had normal salivary gland tissue on smears. Hence, evaluating the facts and communicating in this situation can be very beneficial.

Figure 1

(a) Case of Warthin non-diagnostic on cytology representing cyst content, PAP (200×), (b). A final histopathological diagnosis of Warthin tumor, (H and E, 100×).

In the Milan System, “indeterminate” category is divided into two sub-groups: AUS and SUMP. Variable ROM values of “indeterminate categories” in previous studies were observed [Table 4]. Alruwaii et al.[13] emphasized the necessity of clarifying the diagnostic criteria of these categories, especially in the guide, pointing out the variability in their studies.

Table 4

The risk of malignancy (ROM) for each diagnostic category in published other studies

Number of cases	Faquin et al.[2]MSRSGCROM (%)	Rossi et al.[3]709/1729ROM (%)	Viswanathan et al.[4]323/627ROM (%)	Rohilla et al.[6]94/631ROM (%)	Kala et al.[9]172/293ROM (%)	Sadullahoglu et al.[11]129/459ROM (%)	Hollyfield et al.[14]77/134ROM (%)	Alruwaii et al.[13]694/1560ROM (%)	Bhutani et al.[18]172/899ROM (%)	Current study85/514ROM (%)
Nondiagnostic	25	17	4	0	25	22.2	38	17.1	0	28.5
Nonneoplastic	10	16	3.9	17.4	5	22.2	17	10	20	0
AUS	20	53	18.4	100	20	60	33	37.5	-	33
Neoplasm
a) Benign neoplasm	<5	6	3	7.3	4.4	0	4	2.9	2.6	0
b) SUMP	35	-	21	50	33.3	14.3	33	40.7	33.3	29
Suspicious of malignancy	60	79	76.5	-	85.7	79.2	67	100	66.6	90
Malignant	>90	100	61	96	97.5	89.9	100	98.3	100	100

In the Milan System, AUS is defined as an aspirate that lacks either qualitative or quantitative cytomorphologic features to be diagnosed with confidence as “nonneoplastic or neoplastic.”[2] In our current study, the ROM value for AUS category was 33%. One of the nine cases was diagnosed as mucoepidermoid carcinoma in the final histopathology report [Figure 2a]. A review of the slides [Figure 2b] revealed that the diagnosis of AUS was done due to the presence of atypical cells, but with insufficient cellularity. Two of the nine cases were diagnosed as lymphomas—Hodgkin's lymphoma and MALT lymphoma—in the follow-up. In these two cases, the radiologic appearance of the lesions was suggestive of malignancy with infiltrative borders and increased vascularity. After the AUS diagnosis at cytopathology report with notes of lymphoid rich population, indefinite for lymphoproliferative disorder, physicians performed excision in both cases, revealing that the lesions were intraglandular lymph nodes, affected by MALT and Hodgkin's lymphomas. In the Milan system, it is suggested that slides that are indefinite for neoplasm and slides with reactive atypia, metaplastic changes or mucinous materials with an absent epithelial component should be classified under AUS. Salivary gland lymph nodes or indeterminate lymphoproliferative lesions are also recommended to be included in the AUS category in the Milan system.[2] Previous studies found that one of the most common malignancies in the AUS category was lymphomas.[1314] Alruwaii et al.[13] found 7 of 15 malignant cases that were diagnosed as AUS were diagnosed as lymphomas at follow-up. Hollyfield et al.[14] also found similar results in their studies and obtained different data by removing lymphomas from this evaluation.

Figure 2

(a) A final histopathological diagnosis of mucoepidermoid carcinoma, H and E (100×, (b). Case of mucoepidermoid carcinoma on cytology diagnosed as AUS representing just a few atypical cells, PAP stain (200×).

In light of this knowledge, we recommend that while evaluating monotonous or atypical lymphoid proliferations, if a definitive diagnosis of high-grade lymphoma cannot be made with help of ancillary techniques because it is difficult for cytopathologists, even experienced ones, to rule out low-grade lymphomas on cytomorphological basis. AUS diagnosis should be considered because lymphomas are the major malignant final diagnosis of this category and majority of the false-negative cases as well.

Alruwaii et al.[13] suggested that the rapid on side evaluation (ROSE) technique could be beneficial in the diagnoses of lymphoproliferative lesions. With the ROSE technique, focusing on obtaining cell block material that allows the use of ancillary techniques such as immunohistochemistry, which are of great importance in diagnosis of lymphoproliferative lesions, can also be done. As is known, another technique that is particularly important in lymphoproliferative lesions is the “flow cytometric immunophenotyping.” The MSRSGC guide recommends the use of flow cytometry because it improves the diagnostic value of FNAC in lymphoproliferative lesions.[2] In this study, we used the ROSE technique to collect material for cell blocks, but did not have the opportunity to use flow cytometry on lymphoproliferative lesions.

The other “indeterminate” category is SUMP. It is recommended that this category should be used for cases when a malignant neoplasm cannot be excluded, such as cases with cytologic features implying cellular basaloid neoplasms, cellular oncocytic/oncocytoid neoplasm, and cellular neoplasm with clear cell features. In our study, we found the ROM value for the SUMP category to be 28.5%. The FNAC of a case from a basaloid adenoma [Figure 3a] that was diagnosed as SUMP had rare to a few cohesive cell groups of bland, basaloid-appearing cells with mild nuclear pleomorphism [Figure 3b]. FNAC of another SUMP case from a cellular pleomorphic adenoma had high cellularity, absence of chondromyxoid stroma, accompanied by mild cytologic atypia.

Figure 3

(a) Case of a basaloid adenoma called SUMP on cytology showing uniform basaloid cells, PAP (200×), (b). Basal cell adenoma was composed of bland basaloid cells separated by many abundant amorphous hyaline stroma, H and E (100×).

In the literature, basaloid neoplasms are reported to be the most common diagnostic problem in salivary gland cytology.[151617181920] In our study, the histopathological diagnoses of four cases, which constituted 57% of the SUMP category, were basal cell lesions. In the final histopathology report, three of them were diagnosed as basal cell adenoma and one was diagnosed as basal cell adenocarcinoma. This finding differs from Liu et al.'s[16] study, which reported a lower rate (7.4%). Viswanathan et al.[4] reported 24% (9/38) of their SUMP cases were basaloid neoplasms. Chen et al.[21] found that 19% (6/31) of their SUMP cases were basaloid neoplasms. Sadullahoglu et al.,[11] reported that only 3 out of 129 (2.3%) and Gargano et al.[22] reported 19 out of their 50 cases (38%) were basaloid neoplasms. Lower number of SUMP cases in our study might have contributed to our higher ratio (57%) of basaloid neoplasms compared to other studies.

In our study, one of the SUMP cases was diagnosed as basal cell adenocarcinoma, and two cases with histopathological diagnosis of basal cell adenoma had initial cytological diagnoses of non-diagnostic and AUS. Cantley et al.[19] and Seethala et al.[23] emphasized that distinction of benign and malignant at cases where a definite diagnosis can not be made should be avoided and descriptive reporting should be done. It is also stated in the Milan system that descriptive reporting can be an ideal approach in cases where specific diagnostic techniques such as immunohistochemistry or molecular studies could not be applied.[2]

In the Milan system, it is recommended that the SM category is to be used in cases where specific criteria for malignancy could not be met.[2] Although the diagnostic criteria for this category are very clear, the expectation to minimize the interobserver variability with the use of ancillary techniques is emphasized. The ROM value for the SM category on previous studies ranges from 60% to 100% [Table 4]. In this regard, while re-evaluating our cases of SM category, lack of cell block material, which we could have been used for ancillary techniques, was noted. We think that cytological diagnosis can be made more clearly with the help of such materials that can be used for ancillary techniques. There was one false-positive case of Warthin tumor, where the initial cytological diagnosis was suspicious for malignancy. Cytological preparations of that case revealed highly cellular slides with pleomorphic cells and the presence of an area suggesting tumor necrosis. Warthin tumors heterogeneity is very well known, because of the overlapping morphological features, and may create challenges at diagnosis.[6]

Finally, very specific definitions have been made in the MRSGCC for the malignant category: “The smears that clearly exhibit malignant morphology with their cytomorphological features” It is also suggested that if possible, exact tumor diagnosis and histological grading should also be reported.[2] The malignant category was the most heterogenous group with six cases, all malignancies had a different kind. Due to the limited number of cases, we did not evaluate the cases by the grades. In the future, molecular techniques might be helpful and more accesible diagnostic tools in each case malignant case.

In conclusion, we think that MSRSGC is a highly functional and practical reporting system in salivary gland cytology. It is also confirmed by the data obtained from previous studies that it can provide standardization in indeterminate categories, where communication with the clinician is of great importance. However, to reach more clear data in terms of standardization of clinical management multicenter, meta-analysis studies with the help of ancillary techniques are required.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.