Application of the Milan System for Reporting Salivary Gland Cytopathology: A Single Institutional Experience of 354 Cases with Cytologic-Histologic Correlation

Fine-needle aspiration cytology (FNAC) of salivary gland tumors is challenging due to the morphologic heterogeneity of primary salivary gland tumors, with a high degree of overlap between distinct entities, including benign and malignant neoplasms [1, 2]. However, FNAC is considered the standard of care in distinguishing neoplastic versus non-neoplastic lesions of the salivary gland, and its accuracy is comparable to that of frozen section in the diagnosis of salivary gland lesions [3]. It can reliably differentiate benign from malignant salivary gland lesions and low-grade versus high-grade tumors [4, 5]. This distinction impacts patient care in terms of surgical management, providing information to select the type and extent of the surgical procedure and avoiding unnecessary surgery in some patients. There are still diagnostic challenges and pitfalls encountered in the interpretation of these samples because of overlapping features between different lesions [2, 6‒8]. The application of the Milan system for reporting salivary gland cytopathology (MSRSGC) classification system, with diagnostic categories I–VI, was introduced in April 2018 and has been recommended to standardize the diagnostic reporting of salivary gland lesions by FNAC [9]. The current study aims to reclassify archival salivary gland FNAC samples using MSRSGC, correlate with surgical resections, and calculate the institutional risk of malignancy (ROM) for each category. Furthermore, we summarize our institutional experience with the MSRSGC to collate information to inform patient management and help refine the classification system as it inevitably evolves moving forward.

A review of salivary gland FNAC specimens from the cytology laboratory files at the University of Miami from 2013 to 2018 was performed. Cases with corresponding surgical follow-ups were selected for the study. Three cytopathologists (J.V.T., Y.Z., and C.G.F.) independently reviewed all FNAC reports. A review of slides was reserved for cases where the cytology diagnoses were inconclusive or discordant with the histologic diagnosis. All cytology cases were reclassified using MSRSGC categories: nondiagnostic (ND)-I, non-neoplastic (NN)-II, atypia of undetermined significance (AUS)-III, benign neoplasm (BN)-IVA, salivary gland neoplasm of uncertain malignant potential (SUMP)-IVB, suspicious for malignancy (SM)-V, and malignant (M)-VI. Histological diagnoses were reviewed and correlated with the cytological findings. Histology results were used to calculate the overall ROM for each cytologic category. The ROM was calculated as the ratio of malignant tumors divided by the total number of cases in each category. The sensitivity, specificity, positive predictive value, and negative predictive value were calculated. The diagnostic accuracy of separating benign and malignant neoplasms was also calculated.

A total of 354 FNAC specimens from 354 patients with salivary gland lesions from our cytopathology archives during the study period were reclassified using the MSRSGC categories. All cases were reclassified into one of the 7 MSRSGC categories as follows: ND 17.0% (60), NN 1.4% (5), AUS 11.0% (39), BN 49.4% (175), SUMP 10.7% (38), SM 3.4% (12), and M 7.1% (25). The overall ROM for each category was ND 22%, NN 20%, AUS 15%, BN 2%, SUMP 53%, SM 75%, and M 96%. The findings are summarized in Table 1.

Among our 354 patients, the mean patient age was 56 years (range, 15–94 years), and the female-to-male ratio was 1.2:1 (194 female and 160 male). The aspirate sites included parotid (322), parapharyngeal space (19), submandibular gland (9), palate (3), and buccal mucosa (1).

For the ND category (60), histologic follow-up revealed 35 benign neoplasms that included Warthin tumor (WT) (16), pleomorphic adenoma (PA) (13), basal cell adenoma (2), and cystadenoma (4). The malignant neoplasms (13) included mucoepidermoid carcinoma (MEC) (6), acinic cell carcinoma (ACC) (4), lymphoma (2), and adenocarcinoma, NOS (1). Finally, non-neoplastic lesions (12) included chronic sialadenitis (2), lymphoepithelial cysts (5), salivary duct cysts (4), and cystic hygroma (1).

In the NN (5) and BN (175) categories, histologic follow-up revealed 5 false-negative results (Table 2). One was in the NN category and represented a mucosal-associated lymphoid tissue lymphoma, and 4 were in the BN category and included carcinoma ex pleomorphic adenoma (CAEXPA) (2) and MEC (2). The benign neoplasms (171) in the BN category included PA (130), followed by WT (31), basal cell adenoma (4), monomorphic adenoma (2), atypical PA (2), oncocytoma (1), and schwannoma (1).

Among the AUS category (39), histologic follow-up showed 26 benign neoplasms, 6 malignant neoplasms, and 7 non-neoplastic lesions. The benign neoplasms included WT (15), PA (7), oncocytoma (2), myxoma (1), and sebaceous lymphadenoma (1). The malignant neoplasms included metastatic squamous cell carcinoma (SCC) (3), MEC (2), and lymphoma (1).

For the SUMP category (38), histologic follow-up showed 18 benign neoplasms and 20 malignant neoplasms. The most common benign neoplasm was PA (12), followed by basal cell adenoma (2), WT (2), myoepithelioma (1), and oncocytoma (1). Various types of malignant neoplasms were classified in the SUMP category, including MEC (6), CAEXPA (4), epithelial-myoepithelial carcinoma (2), myoepithelial carcinoma (2), cribriform cystadenocarcinoma (1), adenoid cystic carcinoma (AdCC) (1), salivary duct carcinoma (1), ACC (1), polymorphous adenocarcinoma (1), and dermatofibrosarcoma protuberans (1).

In the SM category (12), histologic follow-up showed 9 malignant neoplasms and 3 benign neoplasms. The malignant neoplasms included metastatic SCC (2), MEC (2), adenocarcinoma NOS (2), ACC (1), poorly differentiated carcinoma (1), and lymphoma (1). The false-positive cases included PA (2) and WT (1) (Table 3).

Finally, for the M category (25), histologic follow-up showed 24 malignant neoplasms and 1 Langerhans cell histiocytosis (LCH) (Table 3). The malignant neoplasms included ACC (6), MEC (2), adenocarcinoma, NOS (3), AdCC (2), CAEXPA (2), salivary duct carcinoma (1), metastatic SCC (2), metastatic melanoma (2), metastatic Merkel cell carcinoma (2), lymphoma (1), and metastatic fibrosarcoma (1).

The results of malignant and benign neoplasms per MSRSGC categories with histologic follow-up are summarized in online supplementary Tables 1 and 2 (see www.karger.com/doi/10.1159/000525263for all online suppl. material).

Upon excluding the specimens classified as AUS and SUMP, defining NN and BN categories as “benign” and defining suspicious for malignancy and malignant categories as “malignant,” the false-negative rate was 2.7%, the false-positive rate was 10.5%, the positive predictive value was 89%, and the negative predictive value was 97%. The sensitivity and specificity in our institutional cohort were 87% and 98%, respectively. FNAC was 96% accurate for the diagnosis of benign and malignant neoplasms.

FNAC is a highly reliable and effective initial diagnostic procedure for the preoperative diagnosis of salivary gland lesions in the hands of experienced cytopathologists. The technique is minimally invasive and cost-effective [3, 10]. It serves to determine the nature of the lesions (neoplastic or non-neoplastic and – if neoplastic – benign or malignant). The preoperative diagnosis of benign or malignant neoplasms assists the surgeon in planning the extent of surgery [4, 8, 11]. The goal of the MSRSGC is to create a uniform reporting system for salivary gland cytopathology and help effectively communicate the findings, in particular the implied ROM, between the pathologist and the treating physician. This 7-tiered classification system uses criteria for categorizing lesions with proposed clinical and surgical management for each category [12, 13]. To further understand the performance of the MSRSGC in practice, we reclassified a large cohort of salivary gland FNAs over 5 years based on the criteria proposed by the MSRSGC.

MSRSGC defines an ND aspirate as less than 60 lesional cells, non-neoplastic salivary gland elements in the setting of a clinically or radiologically defined mass, non-mucinous cyst contents, or poorly preserved cellular material. Excluded from this category are mucinous cyst contents, significant cytologic atypia, or abundant extracellular matrix [14]. The ND category constituted 17.0% of our cohort, which is higher than the recommended rate of ≤10%; however, it falls within the reported range, which varies from 1.1% to 28.2%, depending on the series [15, 16]. The ROM in this category was 22%, which is similar to the target established by the MSRSGC of 25%. Our cohort included non-mucinous cyst contents, few oncocytes, benign salivary gland elements, few epithelial cells, few lymphoid cells, inflammatory cells, and blood content only. On histologic follow-up, most cases were benign neoplasms (35), followed by malignant neoplasms (13) and, lastly, non-neoplastic lesions (12). Most neoplastic lesions that comprised this category were poorly sampled cystic neoplasms, including WT (18) and low-grade MEC (6). Given that a variety of non-neoplastic and neoplastic lesions can be cystic, sparsely cellular aspirates with nonspecific findings (histiocytes, scant mucin, and acellular debris) should be classified as ND with a note explaining the cytologic findings. It has been reported that performing rapid onsite evaluation (ROSE) in salivary gland FNAC decreases the number of ND specimens. Thus, performing ROSE should be considered when aspirating cystic lesions to assess the adequacy of the sample [17]. This recommended management is a clinical and radiological correlation, with repeat fine-needle aspiration if there is a physically or radiologically defined mass [18].

As per MSRSGC, the NN category includes benign entities such as reactive, metaplastic, and inflammatory changes [19]. The NN category constituted 1.4% of our cohort and was composed of reactive lymph nodes (2) and chronic sialadenitis (2). There was 1 false-negative case in this category (Table 2), which was interpreted as a reactive lymph node on cytology, and surgical follow-up revealed a mucosal-associated lymphoid tissue lymphoma. The ROM in this category was 20%, which is higher than the 10% suggested by MSRSGC, but it is still within the reported range of 0–20% [9, 20]. Our study design is likely a contributing factor as we only included cases with histologic follow-up, leading to selection bias. Cases, where the cytologic and clinico-radiologic correlations are concordant and non-worrisome, are typically managed nonoperatively and thus excluded from our study. Indeed, a goal of the NN category is to spare patients from unnecessary surgery. Studies have shown that lymphoma is a leading cause of false-negative results in the NN category, given that a subset of low-grade lymphomas can yield an aspirate with a heterogeneous appearance, mimicking a reactive population of lymphoid cells [21‒23]. Therefore, it is essential that in cases where lymphoma is a clinical consideration, material should be collected for flow cytometry analysis. The recommended management for this category is follow-up with clinical and radiologic correlation.

The AUS category constituted 11.0% of the cohort, which is slightly higher than the recommended MSRSGC criteria of keeping the use of AUS to ≤10% of all salivary gland FNAs [9, 12]. The use of this category should be limited, and every attempt should be made to classify specimens in a more specific category. This category should only be used in indeterminate samples for a neoplasm with limited cellular or architectural atypia (e.g., reactive/reparative atypia, squamous, oncocytic, or mucinous metaplastic changes, mucinous cystic lesions, and atypical lymphoid lesions) [14]. The ROM in this category was 15%, which is somewhat lower than the 20% suggested by the MSRSGC [20]. Our cohort included atypical squamous lesions, oncocytic cell proliferation, oncocytes admixed with lymphocytes and macrophages, myxoid lesions, few myoepithelial/spindle cells, and atypical epithelial cells admixed with mucin. On histologic follow-up, most cases were benign neoplasms (26), followed by non-neoplastic lesions (7), and, lastly, malignant neoplasms (6). Given that most cases in this category were poorly sampled cystic neoplasms (WT and MEC), the use of ROSE should be considered to improve diagnostic yield. The suggested management is clinical and radiological correlation with appropriate follow-up or repeat fine-needle aspiration [14].

The BN category was the most frequent category and comprised 49.4% of our cohort. The ROM in this category was 2%, which is in accordance with the one suggested by MSRSGC of <5% [24]. PA was the most common tumor with 74% (130) of cases, followed by WT with 18% (31) of cases. There were 4 false-negative cases in this category (Table 2). The management for this category is conservative surgery, or in selected cases, clinical follow-up.

The SUMP category comprised 10.7% of our cohort. The ROM in this category was 53%, which is higher than the 35% suggested by MSRSGC. This category includes neoplasms, where a definitive diagnosis of a specific entity cannot be made, and the possibility of malignancy cannot be excluded with certainty [14]. The majority of cases were malignant neoplasms (53%), followed by benign neoplasms (47%) [15]. This management of this category is conservative surgical resection with intraoperative consultation to help determine the extent of surgery [9].

The SM category comprised 3.4% of our cohort. The ROM was 75%, which is higher than the 60% suggested by MSRSGC [9]. This category includes aspirates that are highly suggestive of a malignant neoplasm, but cytomorphologic features are not definitive. There were 3 false-positive cases (Table 3). The management of this category is conservative surgical resection with intraoperative consultation to help determine the extent of surgery [9].

The M category comprised 7.1% of our cohort. The ROM was 96%, which is slightly higher than the 90% suggested by MSRSGC but within the reported range (57–100%) [11]. This category includes aspirates that are diagnostic of malignancy (primary salivary gland neoplasms, lymphomas, sarcomas, and metastatic carcinomas). The most common malignancy seen in this category was ACC (6), followed by adenocarcinoma, NOS (3). Only 1 false-positive case (Table 3) was identified, which on histologic follow-up was classified as LCH. The aspirate for this case was moderately cellular and contained discohesive cells with enlarged, irregular nuclei with occasional folds, prominent nucleoli, and cytoplasmic granular pigment, raising the possibility of melanoma (Fig. 1a). Immunohistochemistry performed on resection specimen revealed that the atypical cells represented CD1a- and langerin-positive histiocytes (Fig. 1b, c). LCH is cytologically similar to and can mimic melanocytic tumors [25]. Both these lesions are positive for S100 protein, a diagnostic pitfall when only limited material is available, which may further complicate the evaluation. For the M category, it is recommended that a specific histologic classification be provided, when possible, given that this information can help with clinical decisions. If a definitive classification is not possible, the tumor should be subclassified into low grade or high grade. The management of this category is surgery, and the extent of surgery will depend upon histologic type and grade [9]. The sensitivity and specificity in our institutional cohort were 87% and 98%, respectively, which match reported range of 54–98% and 88–98%, respectively [21].

In keeping with other reports, MEC represented the most frequent malignant neoplasm in our cohort. This tumor was most commonly placed in ND or AUS categories (8), followed by SUMP (6), SM or M (4), and lastly BN (2). The MECs classified as ND and AUS can be explained by sampling error [15]. MEC also represented the most common malignant tumor associated with false-negative results, similar to other studies [21, 23, 26, 27]. Two cases classified as BN included one case in which the mucinous background was misinterpreted as the chondromyxoid matrix of PA and another case in which the intermediate cells were interpreted as oncocytic cells of WT.

PA was most the common benign tumor, comprising 74% of the BN category, followed by 32% of SUMP, 18% of AUS, and 16% of SM. The cases classified as SUMP were cellular basaloid neoplasms with scant matrix, whereas PAs in the AUS category were composed of few spindled myoepithelial cells with scant stroma. Two cases classified as SM included one PA with adenoid cystic-like areas (Fig. 2) and one PA with cytologic atypia. Adenoid cystic-like areas are encountered in approximately 5% of PAs and are known to pose diagnostic pitfall. Up to 20% of PAs can demonstrate cytological atypia, raising the possibility of malignancy. PAs with atypia without overt features of malignancy can be classified as PA with atypia under the SUMP category. Most PAs harbor recurrent translocations affecting PLAG1 on 8q12 (>50%) and HMGA2 on 12q14–15 (10–15%) [28‒33]. In challenging cases, immunohistochemistry for PLAG1 (positive in 65–100% of PAs) and MYB (positive in up to 89% of AdCCs) can be performed on cell block preparation to help resolve this issue [32‒37].

CAEXPA is a challenging diagnosis for salivary gland FNAC and represents another significant cause of false-negative results, given that the diagnosis can only be made if both benign and malignant components are present in the aspirate. Therefore, FNAC interpretation is difficult, and definitive diagnosis is best rendered by histologic evaluation. Our experience is similar to that of several studies, confirming CAEXPA as a consistent cause of false-negative diagnoses in salivary gland FNAC [2, 17, 21]. CAEXPA was the second most common malignant neoplasm in SUMP category comprising 20% of cases.

WT was commonly classified as ND (16) or AUS (15), and most neoplastic lesions that comprised these categories were poorly sampled cystic neoplasms. It was the second most common tumor in the BN category (31). The single WT interpreted as suspicious for SCC on FNAC showed an infarcted WT with squamous metaplasia and reactive epithelial atypia on histologic follow-up. Infarcted WT can show a range of cytological findings, including necrotic debris, squamous metaplasia with atypia, inflammatory background, and mucin; features that may mimic malignancy, especially SCC or MEC [11, 38‒40].

As previously stated, several of our categories showed higher ROM (NN, SUMP, SM, and M) than the targeted ROM suggested by MSRSGC, but still fell within the reported range. Potential reasons for this higher ROM are multifactorial and include differences in patient populations (high-risk patients treated at a large cancer referral center), quality of FNAC specimens, and cytopathologist threshold for each diagnostic category. Also, it is important to note that calculating ROM based solely on cases with surgical follow-up is expected to overestimate true ROM since clinically non-worrisome cases are typically managed nonsurgically.

A possible limitation of our study is that it was performed at a single institution and therefore reflects our practice and may not necessarily reflect the experiences of other facilities. To that effect, as the MSRSGC becomes more widely used in practice, individual cytopathology laboratories might choose to estimate their own ROM, which would provide more tailored information to the treating physicians of each institution and their patients.

The results reported in this study represent our institutional experience using the MSRSGC system to classify a cohort of salivary gland FNAC samples retrospectively. A review of our archived salivary gland aspirates reported before 2018 revealed a broad range of diagnostic descriptions, particularly for indeterminate cases. As with other recently developed classification systems for cytologic reporting, we found that the MSRSGC is a useful classification system that allows for the standardization of salivary gland FNAC diagnostic reports. This system helps provide useful information for appropriate clinical management and helps ensure the best patient care. It also offers guidelines for the recommended rates in each of the categories. Specifically, the ND and AUS categories, likely representing poorly sampled lesions, should each be limited to ≤10% of the cases. Several factors that can help improve the diagnostic accuracy of FNAC of salivary gland lesions include ROSE and ancillary studies such as flow cytometry, immunohistochemistry, and molecular testing. Institutions would do well to review the quality of their own salivary gland cytology diagnostic reporting by applying the MSRSGC classification system. Collecting and evaluating data across multiple institutions enriches our knowledge about salivary gland cytology. More importantly, it is necessary to refine this system which helps with risk stratification and patient management. Since 2018, we have implemented the MSRSGC system for our salivary gland FNAC diagnostic reporting. Our experience with the prospective use of this system is the subject of an ongoing study.

All procedures performed in this retrospective data analysis involving human participants were in accordance with the ethical standards of the Institutional Review Board at the University of Miami (approval number 20180047), which did not require informed consent.

The authors have no conflicts of interest to declare.

No specific funding was disclosed.

Monica Garcia-Buitrago, Carmen Gomez-Fernandez, Merce Jorda, Darcy A. Kerr, Youley Tjendra, Jaylou M. Velez Torres, and Yiqin Zuo contributed to the study’s conception and design, read, and approved the final manuscript. Material preparation, data collection, and analysis were performed by Carmen Gomez-Fernandez, Youley Tjendra, Jaylou M. Velez Torres, and Yiqin Zuo. Jaylou M. Velez Torres wrote the first draft of the manuscript and Carmen Gomez-Fernandez and Darcy A. Kerr reviewed and revised it. Carmen Gomez-Fernandez and Jaylou M. Velez Torres provided micrographs.

All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.

Dr. Kerr’s current affiliation is Department of Pathology and Laboratory Medicine, Dartmouth-Hitchcock Medical Center and Geisel School of Medicine at Dartmouth, Lebanon, NH, USA.