Milan System for Reporting Salivary Gland Cytopathology- An Experience from Western Indian Population.

INTRODUCTION: Fine-needle aspiration cytology (FNAC) can be challenging to provide a precise diagnosis in salivary gland cytopathology due to diversity of lesions and cytomorphological convergence between the tumors and within the same tumor of salivary gland. The recently proposed Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) provides a risk stratification-based classification system with an intrinsic risk of malignancy (ROM) for each diagnostic category, which aims to furnish useful information to the clinicians. This study was undertaken to evaluate the diagnostic utility and validity of MSRSGC. METHODS AND MATERIAL: In this retrospective study, FNAC done for all salivary gland lesions over a period of two years were retrieved. All cases were categorized according to MSRSGC and correlated with histopathological follow-up, wherever available. ROM was calculated for each category.
RESULTS: The cases belong to following categories: non-diagnostic (1.27%), non-neoplastic (30.38%), atypia of undetermined significance (5.06%), benign neoplasm (46.84%), salivary gland neoplasm of uncertain malignant potential (1.27%), suspicious for malignancy (1.27%), and malignant (13.92%). Out of 79 cases, 50.63% had follow-up. The ROM were 0% for category II and IVa, 50% for category III, and 100% for category IVb, V, and VI. The sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy were recorded as 77.78%, 100%, 100%, 91.3%, and 93.33%, respectively.
CONCLUSIONS: Application of MSRSGC has immense value for standardization of reporting of salivary gland FNAC. Our data corresponds to the studies done worldwide and recommends the use of MSRSGC for future diagnostic purposes. Copyright:

INTRODUCTION

Fine-needle aspiration cytology (FNAC) is a vital investigative tool for salivary gland lesions. The technique is minimally invasive, well-tolerated, and cost-effective. Salivary gland FNA can help ascertain if a lesion is non-neoplastic or neoplastic, benign or malignant, and low-grade or high-grade. Hence, it helps to guide clinical management and surgical planning.[123] The accuracy of FNA to differentiate benign lesions from malignant lesions has been observed to be as high as 81% to 100%. However, the accuracy of FNA to provide a specific diagnosis has a wider range of 48% to 94%.[456] The wide spectrum of cytological features within a particular tumor type, as well as the morphological overlap between various entities, pose challenges for the pathologist to accurately diagnose salivary gland lesions by FNA.[78] In addition, a variety of new salivary gland tumors have been recognized by the World Health Organization (WHO), making an accurate subtyping of lesions all the more arduous.[9]

Due to these challenges, a need for a tiered classification system of salivary gland FNA was felt, to ensure uniformity in reporting and to effectively provide relevant information to clinicians.[1011] In 2015, the American Society of Cytopathology and International Academy of Cytology proposed a risk stratification-based classification for salivary gland FNA called 'The Milan System for Reporting Salivary Gland Cytopathology' (MSRSGC). This comprised of six categories including non-diagnostic (category I), non-neoplastic (category II), atypia of undetermined significance (category III), benign neoplasm (category IVa), salivary gland neoplasm of uncertain malignant potential (SUMP) (category IVb), suspicious for malignancy (category V), and malignant (category VI).[12] A significant feature of this classification is that these categories are well-defined and provide implied risk of malignancy (ROM), thus guide patients for further definitive management.

In this study, we stratified all salivary gland lesions according to MSRSGC and thus tried to evaluate its efficacy by cytohistological correlation and calculate ROM for each category.

MATERIALS AND METHODS

This was a retrospective study done over a period of two years from January 1, 2017 to December 31, 2018 in the Department of Pathology at the tertiary care institute attached to Medical College. Local ethics committee sanction was obtained. The study included cytology smears of patients who visited the cytology section of the Department of Pathology for FNA of salivary gland lesions as a diagnostic workup during the above period. Thus, all FNAs of lesions involving major salivary glands, i.e. parotid, submandibular, and submental glands as well as minor salivary glands were included. The cases with lost or damaged cytological materials were excluded from the study. The detailed clinical and radiological findings were noted from the records.

As a routine practice in our department, in this study, all FNAs were already performed after taking informed consent from the patients. The lesions were aspirated using a 22-23 gauge needle via a direct percutaneous or transoral route by trained cytopathologists. If large swelling, aspirate was taken from multiple sites to avoid the diagnostic error. In the case of fluid aspiration, the fluid was centrifuged and smears were prepared from the sediment. Also, repeat aspirate was performed from any solid lesion remained after evacuating cystic contents. Aspirations were guided radiologically wherever solid cystic lesion was found. Of all smears prepared in each case, 50% were air-dried for Giemsa stain and 50% were alcohol-fixed for Papanicolaou (Pap) stain, respectively.

All cytology smears were retrieved and reviewed by two cytopathologists and assigned to one of the six categories after application of strict criteria given by MSRSGC.[12] The review was done blindly, independent of the histopathological diagnosis. All cases were evaluated in reference to the location of salivary gland involvement, age and sex of patient, and type of lesion. The cytological diagnoses were then correlated with clinical and histopathological follow-up, wherever available. Considering histopathology as a gold standard, the sensitivity, specificity, positive and negative predictive values, and diagnostic accuracy of FNA to detect malignant lesions were calculated. ROM was determined by dividing the number of malignant cases by a total number of histopathological follow-up available in the particular category.

RESULTS

A total of 79 patients visited for FNAC of salivary gland lesion at cytology section of the Department of Pathology of our institute over a period of two years. The male (44.30%) to female (55.70%) ratio was 1: 1.26. The patients' age ranged from 7to 85 years with the mean age being 46 years. Parotid gland (68.35%; N = 54) was involved in the majority of cases while the submandibular gland was affected in 26.58% (N = 21) of cases. The minor salivary gland was affected only in 3.8% (N = 3) of cases.

All the cytology cases were reviewed and categorized according to the MSRSGC into six categories [Table 1]. The majority of the cases (46.84%; N = 37) belong to category IVa, i. e., benign neoplasm followed by category II, i.e. non-neoplastic lesions (30.38%; N = 24). There were 13.92% cases in category VI (malignancy) while 5.06% were labelled under atypia of undetermined significance (AUS) category. One of each case was included in category I, IVb, and V, respectively.

Table 1

Categorization of FNAC cases according to Milan system and correlation with histopathological findings

Milan category	No of Cases (79)	Primary cytology diagnosis	No of cases	HP Avail-able	HP diagnosis	ROM
I (non-diagnostic)	1 (1.27%)	Blood only	1	0	-	-
II (non- neoplastic)	24 (30.38%)	Acute sialadenitis	5	1+3*	Sialadenitis	0
		Chronic sialadenitis	10	4*
		Sialadenosis	2	0
		Lymphoepithelial cyst	7	1*
III (AUS)	4 (5.06%)	Few atypical cells seen	2	1	Intermediate grade MEC$	50%
		Obstructive sialadenitis but low gradeMEC cannot be ruled out	1	1	Sialadenitis
		Cystic lesionD/D Mucocele vs low grade MEC	1	0
IVA (benign neoplastic)	37 (46.84%)	Pleomorphic adenoma (PA)	30	18	PA (17)Canalicular adenoma (1)	0
		PA vs myoepithelioma	4	1	PA
		Basal cell adenoma (BCA)	1	1	BCA
		Monomorphic adenoma	1	0	-
		Warthin’s tumor	1	0	-
IVB (SUMP)#	1 (1.27%)	BCA vs Adenoid cystic carcinoma (AdCC)	1	1	AdCC	100%
V (Suspicious for malignancy)	1 (1.27%)	Suspicious of AdCC	1	1	AdCC	100%
VI (Malignant)	11 (13.92%)	MEC	3	3	MEC	100%
		Duct cell carcinoma	3	1	Metastasis of lung carcinoma
		High grade carcinoma	3	0	-
		Acinic cell carcinoma (ACC) vs MASC^	1	1	MASC
		Acinic cell carcinoma	1	1	ACC
		Malignant round cell tumorD/D Rhabdomyosarcoma vsNon Hodgkin’s lymphoma	1	1	Embryonal RMS

*Clinical follow up as swelling decreased with medication alone with no further intervention required. #Salivary gland neoplasm of uncertain malignant potential, $Mucoepidermoid carcinoma, ^ Mammary Analogue of Secretary Carcinoma

The MSRSGC Category I consisted of one case, which had only blood on cytology and the patient did not follow up. The MSRSGC Category II consists of chronic sialadenitis (10 cases) as the most common lesion. Histopathology follow-up was available in one case and the diagnosis was sialadenitis. In addition, nine cases had clinical follow-up with a history of disappearance of the swelling with medication alone requiring no further intervention. This was consistent with their diagnosis of non-neoplastic inflammatory diseases.

No discordant case was found in this study after the application of strict MSRSGC criteria while one of each case from the indeterminate categories, i.e. AUS and SUMP turned out to be malignant on histopathology. There was one case in Category V reported as suspicious for adenoid cystic carcinoma (AdCC), which turned out to be AdCC on histopathology [Figure 1]. Out of 11 cases of Category VI, 63.64% cases had follow-up and all were malignant on histopathology. No follow-up was available in the non-diagnostic category (one case only) while no malignant case was reported on histopathology in non-neoplastic and benign neoplasm categories.

Figure 1

Moderately cellular smears showing clusters of larger basaloid cells around acelluar matrix; cribriform pattern (a: Pap stain, 40x). Abundant acellular, non-fibrillary hyaline matrix (b, Pap stain, 10x). Although features are highly suggestive of adenoid cystic carcinoma, the cellular details were compromised due to drying artefact (c, Pap stain, 40x) hence this case was labelled as suspicious for adenoid carcinoma (AdCC). Follow up of the case confirmed AdCC on histopathology (d, H and E, 40x)

Accordingly, ROM was calculated for all categories [Table 2]. Statistical analysis revealed the sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy as 77.78%, 100%, 100%, 91.3%, and 93.33%, respectively.

Table 2

Cytohistopathological correlation of salivary gland cytopathology with risk of malignancy (ROM)

Milan category	No of cases	HP available (%)	Non-neoplastic	Benign neoplastic	Malignant	ROM (%)
I	1	0	-	-	-	-
II	24	1+9*(41.66%)	1+9*	0	0	0
III	4	2 (50%)	1	0	1	50%
IVA	37	20 (54.05%)	0	20	0	0
IV B	1	1	-	-	1	100%
V	1	1 (100%)	0	0	1	100%
VI	11	7 (63.63%)	0	0	7	100%
TOTAL	79	31+9*(50.63%)	2+9*	20	9	22.5%

*Clinical follow up as swelling decreased with medication alone with no further intervention required

DISCUSSION

MSRSGC aims at providing a definite diagnostic category for salivary gland lesion cytology and furnishes useful information to the clinician including ROM.[12] We applied the same criteria to define our cases and check the utility in our study population and compared the results with studies done worldwide.

In our study, there was a slight female preponderance similar to findings by other studies.[131415] The mean age group was from 3rd to 4th decade, similar to various studies from India, e.g. by Mishra et al., Karuna et al., Kala et al.[131617] This finding was in contrast to studies done worldwide, where the mean age group was on the higher side, i.e. in the 6th decade.[1415181920] This may be due to the referral practices in their institutions.

The majority of the cases in our study were from the parotid (69.2%) followed by the submandibular gland (25.6%). This distribution is identical to most of the studies in literature.[14151617181920212223] Although the incidence of neoplasm is lesser in the submandibular gland, ROM in tumors arising from the submandibular gland has been shown to be almost double (41-54%) that of tumors in the parotid gland.[24] Maleki et al. studied FNAs from submandibular gland lesions. They found that the ROM of each MSRSGC category for submandibular gland cytology was similar to that reported for the parotid gland and thus concluded that MSRSGC can be reliably applied to submandibular gland lesion cytology.[24]

MSRSGC recommends a minimum of 60 lesional cells be treated as a criterion for adequacy and the rate of non-diagnostic aspirates to be less than 10%.[12] In their review of 29 studies, Wei et al. found the share of non-diagnostic cases to be ranging from 1.1% to 7.8%.[1] This was similar to the majority of studies done in India (range 2.08% to 6.1%).[1316172123] However, Chen et al. Thiryayi et al. Malleki et al. and Vallenthaiel et al. found a higher proportion of non-diagnostic cases (18.4%, 21.3%, 21.4% and 23%, respectively).[19202425] Chen et al. observed that the adherence to the application of strict adequacy criteria as per MSRSGC increases the number of non-diagnostic cases but at the same time decreases the number of false-negative cases of malignancy.[19] Maleki et al. studied submandibular gland lesions with 21.4% of cases in the non-diagnostic category. The reasons partly attributed to chronic sclerosing sialadenitis (Küttner tumor), which is more common in submandibular as compared to the parotid gland and aspirates from this lesion are often paucicellular and difficult to obtain. Also, the authors suggested that higher rates of non-diagnostic aspirates from the submandibular gland pointed out the fact that aspirates from submandibular gland lesions are more difficult to obtain than that from parotid gland lesions and the use of ultrasound guidance and rapid on-site evaluation (ROSE) will definitely improve the results.[24]

MSRSGC has ascertained ROM for Category I to be 25% while it ranged from 0-67% in various studies.[12] Vallenthaiel et al. observed ROM for non-diagnostic category to be 44%, which is higher compared to other studies [Table 3].[131516171819202122232426] This was attributed to the re-categorization of negative samples into non-diagnostic category. They reinforced the value of a mandatory repeat guided aspiration or biopsy for such cases in the presence of a mass lesion.[1925] Non-diagnostic FNAs are affected by various preanalytical factors including operator experience, availability of radiological assistance, intralesional heterogenicity, and presence of cystic component etc.[18] Thus, availability of adequate clinical information and radiological findings is imperative to limit the number of non-diagnostic aspirates.

Table 3

Comparison of category-wise ROM with other studies

Authors	I (%)	II (%)	III (%)	Iva (%)	IVb (%)	V (%)	VI (%)
Milan system[12]	25	10	20	<5	35	60	90
Savant et al.[14]	0	0	33.3	0.8	40.9	100	100
Vishawanathan et al.[15]	6.7	7.1	38.9	5	34.2	92.9	92.3
Karuna et al.[16]	0	0	50	2.44	33.33	100	93.33
Kala et al.[17]	25	5	20	4.4	33.3	85.7	97.5
Katta et al.[21]	33.33	11.1	100	6.9	50	66.6	87.5
Rohilla et al.[19]	0	17.4	100	7.3	50	-	96
Pujani et al.[22]	0	10	50	2.5	50	100	100
Chen et al.[19]	8.6	15.4	36.8	2.6	32.3	71.4	100
Thiyayi et al.[20]	8.5	1.6	0	1.9	26.7	100	100
Vallenthaiel et al.[25]	44	8	0	8	44	81	100
W Park et al.[27]	19.5	6.9	0	2.4	26.2	83.3	100
Present study	-	0	50	0	100	100	100

The ROM for the non-neoplastic category (II) in our study was 0%, which is consistent with many other studies in literature.[12141623] MSRSGC has estimated ROM for non-neoplastic cases to be 10%. However, a wide range of ROM from 0-20% has been observed in many published studies. MSRSGC believes that this overestimation of ROM may be due to selection bias for surgery in non-neoplastic cases.[12] Rohilla et al., Chen et al., and Pujani et al. observed that the most common cause for false-negative diagnosis in non-neoplastic category was low-grade mucoepidermoid carcinoma (MEC) while lymphoma was the most common cause in their study by Vishwanathan et al.[15192223] The low-grade MEC always poses a diagnostic problem due to low cellularity, lack of cytologic atypia, and abundant mucinous background along with inflammatory cells, thus causing significant morphological overlap with other cystic lesions leading to false-negative diagnosis.[22] The inclusion of such cases in AUS category will definitely decrease false-negative diagnoses.

We had four (5.06%) cases under category AUS. This was similar to the recommendation by MSRSGC (less than 10%) and various studies in literature (0.2%-6%).[121314151617192021222324252627] Interestingly, Hollyfield et al. found 11% of their cases belonged to AUS and also noted fair agreement in interobserver reliability in these cases. They mentioned that the observers favored AUS or benign neoplasm category for SUMP cases, probably because of less familiarity with the SUMP as a diagnostic category.[18] The ROM for AUS in our study was 50%, which is higher than that recommended by the MSRSGC (20%) but in concordance with other studies in literature (0-100%).[20212227] The extremes of findings are likely because of the low number of cases and inadequate histopathologic follow-up.

Majority of the cases (46.84%) in this study were benign neoplasms similar to most of the studies in literature.[1416181920212223252627] In the contrary, Maleki et al. found only 18.3% of their cases to be in the benign neoplasm category. This low number could be because of the inclusion of only submandibular gland lesions and having non-neoplastic category as the most common category, the reason already discussed.

In our study, ROM for benign neoplasm was calculated to be 0%, which was in agreement with ROM of <5% as per MSRSGC and various studies in literature.[12] Although no false-negative diagnosis was observed in our study, Rohilla et al. found two cases of MEC and one case of oncocytic carcinoma as false negative in category IVa. Chen et al. in their study observed six cases of carcinoma ex pleomorphic adenoma (PA) reported as PA on FNA cytology and concluded that it was likely due to sampling error.[19]

There was only one (1.27%) case under Category IVb (neoplasm of uncertain malignant potential, SUMP). Also in various other studies, SUMP is not a common diagnosis (1.7% to 12%).[1315161718192021222324252627] Our case was a 52-year-old lady with a 2 × 2 cm preauricular mass for one year. On cytology, the aspirate was very cellular with sheets and clusters of basaloid cells and scant to moderate amount of nonfibrillar matrix. The cells were slightly larger in size but with low nucleocytoplasmic (N:C) ratio and round, normochromic nuclei. With careful examination, few clusters showed a high N:C ratio and hyperchromatic nuclei. But necrosis or mitosis was absent. The aspirate was very cellular with focal atypical features. Thus, our case was placed in the SUMP category under cellular basaloid neoplasm. The differential diagnoses for cellular basaloid neoplasm with hyaline matrix include basal cell adenoma/carcinoma, adenoid cystic carcinoma, epithelial-myoepithelial carcinoma, or polymorphous low-grade adenocarcinoma.[12] Hence, histological evaluation is essential to differentiate benign from malignant entities. Histopathological examination of our case showed adenoid cystic carcinoma, thus ROM was 100% for this category in our study. Although MSRSGC recommended ROM for SUMP to be 35%, the overall finding in different studies varies from 0-100%.[12] The data from Indian studies show the ROM for Category IVb is in the range of 33.33% to 50%.[131617212223] This is potentially due to the low number of cases in the SUMP category.

We had one more case with basaloid cells and hyaline matrix. The cells were large having high N: C ratio with hyperchromatic nuclei. The hyaline matrix was abundant, globular at places with focal fibrillary appearance as well. Although the smears were cellular, there was drying artifact at many places. Hence, it was labeled as suspicious for adenoid cystic carcinoma (Category V). The histopathology of this case was consistent with adenoid cystic carcinoma with ROM for Category V being 100% in our study. The overall rate of Category V in various studies is less than 5%, which is similar to our study (1.27%).[1315161718192021222324252627] Exceptionally, Vallonthaiel A G et al., who included only parotid gland lesions in their study, found 14% of their cases to be in Category V.[25] This rate is higher than the other studies already mentioned. The ROM for Category V ranges from 0-100% in various studies but MSRSGC recommended that it should be around 60%.[12]

Malignancy (Category VI) was unequivocal in 13.92% of the aspirates (11 cases). Histopathology was available in 63.63% of cases and all were positive for malignancy [Figure 2]. The rate of malignancy ranged from 2.5% to 26.6% and ROM ranged from 87-100% for the same in various studies in literature.[1315161718192021222324252627] The ROM for Category VI in our study came out to be 100%, which although higher than ROM of 90% as per MSRSGC, was similar to findings in various studies.[14181920232527]

Figure 2

Cellular smears comprising of monomorphic cells with round nuclei and abundant cytoplasm with delicate vacuolation. This was reported as acinic cell carcinoma on cytology (Category VI) (a, Giemsa stain, 40x) and confirmed on histopathology. (b, H and E stain, 40x]. Another case showing monomorphic cell population with cells having low nucleo-cytoplasmic ratio and vacuolated cytoplasm having cystic background. This was reported as Category VI with features are of Acinic cell carcinoma-papillary cystic variant vs Mammary Analogue of Secretory Carcinoma (MASC) (Category VI) (c, Pap stain, 40x and d, Giemsa stain, 40x). Histopathology confirmed the diagnosis of MASC (not shown in figure)

Sensitivity and specificity of salivary gland FNA to differentiate benign and malignant lesions have been reported to be high ranging from 54% to 98% and 88% to 98%, respectively.[14] This was comparable to our study [Table 4].

Table 4

Comparison of statistical parameters

	Sensitivity	Specificity	PPV	NPV	Diagnostic accuracy
Mishra et al.[13]	96.84%	80.95%	95.83%	85%	-
Karuna et al.[16]	85%	98.14%	94.44%	94.64%	94.54%
Kala et al.[17]	83.3%	98.3%	95.7%	92.8%	-
Katta et al.[21]	73.34%	95.56%	84.62%	91.49%	90%
Rohilla et al.[22]	79.4%	98.3%	96.4%	89.2%	91.4%
Pujani et al.[23]	81.8%	100%	100%	96.4%	96.9%
Chen et al.[19]	70.4%	99.2%	90.5%	96.7%	80.8%
Present study	75%	100%	100%	92.8%	94.1%

Although various studies based on MSRSGC are coming up from India, we have not come across any study from western India. Thus, this represents the first study from western India.

There were certain limitations in this study, the most critical being the comparatively smaller sample size. We understand that more salivary gland FNAs should have been included in the study group to evaluate the validity of MSRSGC.

In conclusion, our data show consistent results compared to studies done worldwide and thus recommend the MSRSGC for the standardization of salivary gland FNA reporting. The provision of AUS and SUMP will definitely help to decrease the false-negative diagnosis and will appropriately communicate the degree of suspicion of malignancy to the clinician. Thus, the implied ROM for each category will help to triage the patients for effective clinical management.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.