Abstract
Background:
The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) describes several salivary gland fine needle aspiration cytology (SGFNAC) morphologies developed by Griffith. Basaloid neoplasms are pleomorphic (PB) or monomorphic with fibrillary (MBFIB), hyaline (MBHy), or other (MBOther) matrix. Oncocytoid neoplasms can be pleomorphic (PO), show granular/vacuolated cytoplasm (OGV), or be monomorphic with mucinous (MOMuc), cystic (MOCyst), or other (MOOther) background. We explore interobserver agreement (IOA) and risk of malignancy (ROM) for these subcategories.
Methods:
The study includes 169 SGFNAC cases with surgical follow-up. Four reviewers categorized these using the Griffith criteria with consensus determined by majority. For all morphologic categories, IOA- using Fleiss’ kappa- and ROM were calculated.
Results:
ROMs for basaloid categories were- PB: 100% (1/1), MBHy: 71.4% (5/7), MBFib: 50.0% (3/6) and MBOther: 47.4% (9/19). ROMs for oncocytoid neoplasms were- OGV: 100% (10/10), MOMuc: 92.3% (12/13), PO: 88.9% (8/9), MOOther: 33.3% (5/15), MOCyst: 0 (0/1). The system showed substantial agreement overall (κ=0.69). For basaloid neoplasms, the IOA results were: MBHy (κ=0.59), MBFib (κ=0.41), MBOther (κ=0.41), PB (κ=0.11). For oncocytoid neoplasms, the IOA results were: MOMuc (κ=0.88), OGV (κ=0.67), PO (κ=0.63), MOOther (κ=0.57), MOCyst (κ=0.18).
Conclusions:
The SGFNAC scheme proposed by Griffith and incorporated into MSRSGC shows substantial agreement overall, with particularly high agreement for MOMuc, OGV, PO, and MBHy categories. PB and MOCyst categories show slight agreement and may be improved by revised criteria. PB, PO, MOMuc, and OGV categories show high ROM, and the latter two might best be classified as suspicious for malignancy.
Keywords: Salivary gland cytopathology, fine needle aspiration, Milan System, interobserver agreement, reproducibility, risk of malignancy, basaloid, oncocytoid
Precis:
In this study, four cytopathologists independently categorize 169 salivary gland cytology cases with surgical follow-up using a scheme devised by Griffith and adopted by the Milan System, finding substantial interobserver agreement overall and particularly strong agreement for categories such as monomorphic oncocytoid neoplasm with mucinous background, oncocytoid neoplasm with granular/vacuolated cytoplasm, pleomorphic oncocytoid neoplasm, and monomorphic basaloid neoplasm with hyaline matrix. The study also confirms a high risk of malignancy for pleomorphic basaloid neoplasms, pleomorphic oncocytoid neoplasms, monomorphic oncocytoid neoplasms with mucinous background, and oncocytoid neoplasms with granular/vacuolated cytoplasm, suggesting the latter two entities might best be classified under Milan category V: Suspicious for malignancy.
Background
Fine needle aspiration cytology (FNAC) has long represented an important test in the workup of salivary gland masses worldwide(1). Salivary gland FNAC is extremely reliable at distinguishing non-neoplastic from neoplastic disease, which can help determine the need for surgery(2). Among neoplastic lesions, FNAC generally shows good sensitivity (~70–80%) and high specificity (>90%) for diagnosing malignancy, however these test characteristics have shown wide heterogeneity across institutions(2,3). Salivary gland FNAC may also distinguish low- from high-grade malignancy(4), which can alter the surgical approach from a limited excision to a radical resection- potentially with facial nerve sacrifice- and ipsilateral neck dissection(4). In the last several years, however, the interinstitutional variability in salivary gland FNAC performance combined with the less than perfect sensitivity for malignancy led to increasing interest in a universal risk-stratification system for salivary gland FNAC classification(5,6). The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) was introduced to address these issues, creating a standardized system to improve the practicing cytologist’s diagnostic approach and better guide subsequent therapeutic intervention(7).
In addition to providing a tiered framework for salivary gland FNAC, the MSRSGC recommends several morphologic categories with specific differential diagnoses, allowing the cytologist to narrow his or her differential diagnosis and perform appropriate ancillary testing when necessary. Several of these categories had been introduced by Griffith et al in their original risk stratification scheme for salivary gland FNAC (Fig 1)(5). Per their system, the neoplastic categories included benign neoplasms with characteristic morphology (pleomorphic adenoma and Warthin tumor) as well as several morphologic classes divided among two umbrella categories based on the amount of tumor cell cytoplasm: basaloid neoplasms have minimal cytoplasm, imparting a blue appearance on Romanowsky-type stains, while oncocytoid neoplasms have more abundant cytoplasm. Both basaloid and oncocytoid categories have pleomorphic subcategories which are strongly suggestive of malignancy. For those neoplasms with more monomorphic cells, the Griffith criteria subcategorize the basaloid neoplasms by background stroma (fibrillary, hyaline, or mixed/other). The oncocytoid category is first subdivided by cytoplasmic characteristics; cases showing coarsely granular/vacuolated cytoplasm are separated from the other monomorphic oncocytoid subcategory, which is further divided by background contents (mucinous, cystic, other). The MSRSGC adopted basaloid neoplasm and oncocytoid neoplasm and added an additional category of cellular neoplasm with clear cell features as subcategories of its Category IVB: Salivary gland neoplasm of undetermined malignant potential (SUMP)(7). For the first two, MSRSGC provides morphologic scenarios that echo those of Griffith, with basaloid neoplasms further divided by stroma and oncocytoid neoplasms divided by background.
Figure 1.
Original Griffith categorization scheme
Ideally, the wide acceptance and utilization of the MSRSGC would lead to a standardized and reproducible system for categorizing salivary gland FNAC. With the MSRSGC, the cytology community now has a guide for how to distribute various morphologies into different diagnostic categories. We will only be able to achieve reproducible results, however, if we can agree on how to classify individual cases into a specific morphologic category to begin with. In this study, we attempt to determine the extent to which we can do just that. We had multiple pathologists review select salivary gland FNAC cases in order to determine the interobserver agreement (IOA) for the various categories proposed by Griffith et al and adopted by the MSRSGC. A second goal was to determine the risk of malignancy (ROM) for these categories at our institution after consensus classification of cases.
Methods
Following approval by the Institutional Review Board at Memorial Sloan Kettering Cancer Center (MSKCC), we performed a retrospective search of the electronic database in order to retrieve all salivary gland FNA cases performed at MSKCC between January 2007 and December 2018, yielding 976 cases, of which 373 had surgical follow-up. Cases were excluded from analysis if surgical follow-up demonstrated a hematolymphoid malignancy or metastasis or if the cytology specimen was originally classified as non-diagnostic, leaving a total of 169 cases for analysis.
The cases were then categorized by four reviewers (DJL, OL, DJB, XW) according to the Griffith criteria as either non-neoplastic, benign neoplasm (pleomorphic adenoma or Warthin tumor), or as one of the various basaloid or oncocytoid subcategories. Reviewers based their interpretations on an analysis of smeared slides and liquid-based preparations.
A consensus cytology designation was assigned to each case by simple majority of reviewer designations. ROM was then calculated for all categories. Cases without a consensus cytology designation were excluded from ROM analysis.
The level of interobserver agreement (IOA) for each of the morphologic subcategories was calculated using Fleiss’ Kappa with corresponding 95% confidence intervals (CIs) for 12 evaluation categories (Warthin tumor, pleomorphic adenoma, the 9 morphologic subcategories and an “other” category consisting of alternative diagnoses including non-diagnostic and non-neoplastic). Data analysis was generated in Excel version 16.16 using the Real Statistics Resource Pack software (Release 6.8), Copyright (2013 – 2020) Charles Zaiontz. For categories with only slight agreement, all cases which were designated as such by even a single observer were reviewed to help determine the source of the poor IOA.
Results
Initial search revealed 169 cases which met inclusion criteria and had slides available for review. The distribution of FNAC cytology consensus designations by ultimate surgical pathology diagnosis is demonstrated in Table 1. Using the Griffith criteria, there were 6 cases classified as non-neoplastic, 70 cases classified as benign neoplasms, 33 classified as basaloid neoplasms, and 48 cases classified as oncocytoid neoplasms. There were 12 cases for which consensus was not reached. The overall ROM for each morphologic category in our study in comparison to prior studies is depicted in Table 2.
Table 1.
Distribution of consensus cytology designation results.
| Surgical pathology diagnosis | Consensus Cytology designation | No consensus | Total | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Benign | Basaloid categories | Oncocytoid categories | ||||||||||||
| Non-neo | WT | PA | MBFib | MBHy | MBOther | PB | MOCyst | MOOther | MOMuc | OGV | PO | |||
| Chronic sialadenitis | 5 | - | - | - | - | - | - | - | - | - | - | - | - | 5 |
| Granulomatous inflammation | 1 | - | - | - | - | - | - | - | - | - | - | - | - | 1 |
| Pleomorphic adenoma | - | - | 58 | 3 | 2 | 5 | - | - | - | - | - | - | 4 | 72 |
| Oncocytoma | - | - | - | - | - | - | - | - | 10 | - | - | 1 | 1 | 12 |
| Basal cell adenoma | - | - | - | - | - | 2 | - | - | - | - | - | - | - | 2 |
| Oncocytic cystadenoma | - | - | - | - | - | - | - | 1 | - | - | - | - | 1 | 2 |
| Lymphoepithelial cyst | - | - | - | - | - | 1 | - | - | - | - | - | - | - | 1 |
| Myoepithelioma | - | - | - | - | - | 1 | - | - | - | - | - | - | - | 1 |
| Intercalated duct adenoma | - | - | - | - | - | 1 | - | - | - | - | - | - | - | 1 |
| Warthin tumor | - | 12 | - | - | - | - | - | - | - | 1 | - | - | 13 | |
| Mucoepidermoid carcinoma | - | - | - | - | - | - | - | - | 2 | 12 | - | - | 2 | 16 |
| Acinic cell carcinoma | - | - | - | - | - | - | - | - | - | - | 9 | - | 9 | |
| Myoepithelial carcinoma | - | - | - | 1 | 1 | 2 | 1 | - | 1 | - | - | 1 | 1 | 8 |
| Epithelial-myoepithelial carcinoma | - | - | - | 1 | 1 | 2 | - | - | 1 | - | - | - | 1 | 6 |
| Salivary duct carcinoma (de novo) | - | - | - | - | - | - | - | - | - | - | - | 3 | 1 | 4 |
| Salivary duct carcinoma (ex PA) | - | - | - | - | - | - | - | - | - | - | - | 4 | - | 4 |
| Adenoid cystic carcinoma | - | - | - | - | 3 | 1 | - | - | - | - | - | - | - | 4 |
| Basal cell adenocarcinoma | - | - | - | - | - | 3 | - | - | - | - | - | - | - | 3 |
| Adenocarcinoma, NOS | - | - | - | 1 | - | 1 | - | - | - | - | - | - | - | 2 |
| Carcinoma ex PA, NOS | - | - | - | - | - | - | - | - | - | - | - | - | 1 | 1 |
| Oncocytic carcinoma | - | - | - | - | - | - | - | - | 1 | - | - | - | - | 1 |
| Secretory carcinoma | - | - | - | - | - | - | - | - | - | - | 1 | - | - | 1 |
| Total | 6 | 12 | 58 | 6 | 7 | 19 | 1 | 1 | 15 | 13 | 10 | 9 | 12 | 169 |
MBFib=monomorphic basaloid neoplasm with fibrillary matrix, MBHy=monomorphic basaloid neoplasm with hyaline matrix, MBOther=monomorphic basaloid neoplasm with other matrix, MOCyst=monomorphic oncocytoid neoplasm with cystic background, MOMuc=monomorphic oncocytoid neoplasm with mucinous background, MOOther=monomorphic oncocytoid neoplasm with other background, non-neo=non-neoplastic, OGV=oncocytoid neoplasm with granular/vacuolated cytoplasm, PA=pleomorphic adenoma, PB=pleomorphic basaloid neoplasm, PO=pleomorphic oncocytoid neoplasm, WT=Warthin tumor
Table 2:
Risk of malignancy for various morphologic categories.
| Study | Morphologic category | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Warthin tumor | PA | Monomorphic basaloid | Pleomorphic basaloid | Monomorphic oncocytoid | Oncoytoid Gran/Vac | Pleomorphic oncocytoid | |||||
| Fibrillary | Hyaline | Mixed/Other | Cystic | Mucinous | Other | ||||||
| Griffith et al. 2015 | 0 (0/29) | 4.1% (2/49) | 15.4% (2/13) | 42.9% (3/7) | 60.0% (6/10) | 100.0% (4/4) | 0 (0/11) | 80.0% (8/10) | 21.1% (4/19) | 84.6% (11/13) | 100.0% (21/21) |
| Griffith et al. 2017 | 0 (0/16) | 14.3% (3/21) | 14.3% (2/14) | 83.3% (5/6) | 60.0% (6/10) | 75.0% (3/4) | 25% (1/4) | 80.0% (24/30) | 40.0% (2/5) | 100.0% (17/17) | 100.0% (17/17) |
| Current study | 0 (0/12) | 0 (0/58) | 50% (3/6) | 71.4% (5/7) | 47.4% (9/19) | 100.0% (1/1) | 0 (0/1) | 92.3% (12/13) | 33.3% (5/15) | 100.0% (10/10) | 88.9% (8/9) |
| Total | 0 (0/57) | 3.9% (5/128) | 21.2% (7/33) | 65.0% (13/20) | 53.8% (21/39) | 88.9% (8/9) | 6.3% (1/16) | 83.0% (44/53) | 28.2% (11/39) | 95.0% (38/40) | 97.8% (46/47) |
PA=pleomorphic adenoma
For the 76 cases classified as benign, all were confirmed as such upon surgical resection. This included 6 non-neoplastic cases, 58 cases classified as pleomorphic adenoma, and 12 cases classified as Warthin tumor.
There were 33 cases classified by consensus opinion as basaloid neoplasms, including 32 monomorphic cases and a single pleomorphic case (Fig 2). Among monomorphic basaloid neoplasms, there were 6 cases with fibrillary matrix (MBFib) (Fig 3), 7 with hyaline matrix (MBHy) (Fig 4), and 19 with mixed/other matrix (MBOther) (Fig 5). For the monomorphic basaloid neoplasms, ROM was highest for MBHy cases (71.4%, 5/7) followed by MBFib (50.0%, 3/6) and MBOther cases (47.4%, 9/19). The single case demonstrating a pleomorphic basaloid (PB) neoplasm proved to be a myoepithelial carcinoma on resection.
Figure 2:
Pleomorphic basaloid neoplasm. This was the single case with a consensus designation of pleomorphic basaloid neoplasm. Note the significant anisonucleosis of the smaller group of cells. Follow-up revealed myoepithelial carcinoma.
Figure 3:
Monomorphic basaloid neoplasm with fibrillary matrix. This case revealed cellular smears with only minimal amounts (<25%) of pink fibrillary matrix. Follow-up revealed pleomorphic adenoma.
Figure 4:
Monomorphic basaloid neoplasm with hyaline matrix. Note the abundant dense pink globules of matrix material. Follow-up revealed adenoid cystic carcinoma.
Figure 5:
Monomorphic basaloid neoplasm with mixed/other matrix. This example demonstrates bland basaloid cells without any appreciable matrix material. Follow-up revealed basal cell adenocarcinoma.
There were 48 cases classified by consensus opinion as oncocytoid neoplasms, including 29 monomorphic cases, 10 oncocytoid neoplasms with granular/vacuolated cytoplasm (OGV) (Figs 6a and 6b), and 9 pleomorphic oncocytoid (PO) neoplasms (Fig 7). Among the monomorphic oncocytoid neoplasms, ROM was highest for those with a mucinous background (MOMuc) (92.3%, 12/13) (Fig 8) followed by those without a significant background component (MOOther) (33.3%, 5/15) (Fig 9). A single monomorphic oncocytoid case with cystic background (MOCyst) demonstrated an oncocytic cystadenoma on resection (Fig 10). All 10 OGV cases were malignant (9 acinic cell carcinomas and 1 secretory carcinoma). The ROM for PO cases was 88.9% (8/9), with a case of oncocytoma representing the lone benign case with pleomorphic cytologic features.
Figures 6a and 6b:
Two cases categorized as oncocytoid neoplasm with granular/vacuolated cytoplasm. The first case (6a) demonstrates sheets of cells with fragile granular to microvacuolated cytoplasm. Note the numerous naked nuclei. Follow-up revealed an acinic cell carcinoma. The second case (6b) shows cells with abundant cytoplasm with frequent microvacuoles and a single cell distended with a massive pink cytoplasmic vacuole. Follow-up revealed a secretory carcinoma.
Figure 7:
Pleomorphic oncocytoid neoplasm. Cells with moderate amounts of pink cytoplasm demonstrating marked nuclear pleomorphism. Follow-up revealed salivary duct carcinoma.
Figure 8:
Monomorphic oncocytoid neoplasm with mucinous background. Note the abundant dense pink mucin in the background. Follow-up revealed mucoepidermoid carcinoma.
Figure 9:
Monomorphic oncocytoid neoplasm with other background. A pure population of oncocytes without a cystic or mucinous background. Follow-up revealed an oncocytoma.
Figure 10:
Monomorphic oncocytoid neoplasm with cyst contents background. Follow-up revealed an oncocytic cystadenoma.
There were 12 cases for which consensus could not be achieved (Table 3), 6 of which proved to be benign and 6 malignant. For all of these cases, there was a universal consensus to the umbrella category (basaloid vs oncocytoid), but there was a discrepancy between the categorical subsets. The discord resulted for all cases from an even split of the four reviewers between two diagnoses. The most common scenario was between diagnoses of MBFib and MBOther, which occurred in 4 cases: 3 pleomorphic adenomas and 1 epithelial-myoepithelial carcinoma. There were 2 cases of mucoepidermoid carcinoma without a consensus designation, each of which was placed in the OGV category by 2 reviewers. For one of these cases, the other 2 reviewers labeled the case as MOOther, while for the other case the other 2 reviewers labeled the case as MOMuc.
Table 3:
Cases without consensus.
| Case | Reviewer number | Diagnosis | Cause of discrepancy | |||
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | |||
| 1 | MBOther | MBFib | MBFib | MBOther | Pleomorphic adenoma | MBOther vs MBFib |
| 2 | MBFib | MBFib | MBOther | MBOther | Pleomorphic adenoma | MBOther vs MBFib |
| 3 | MBFib | MBOther | MBOther | MBFib | Pleomorphic adenoma | MBOther vs MBFib |
| 4 | OGV | PO | PO | OGV | Pleomorphic adenoma | OGV vs PO |
| 5 | OGV | OGV | PO | PO | Oncocytoma | OGV vs PO |
| 6 | WT | MOCyst | WT | MOCyst | Oncocytic cystadenoma | WT vs MOCyst |
| 7 | PB | MBOther | PB | MBOther | Carcinoma ex pleomorphic adenoma | MBOther vs PB |
| 8 | MBOther | MBFib | MBFib | MBOther | Epithelial-myoeithelial carcinoma | MBOther vs MBFib |
| 9 | MOOther | OGV | OGV | MOOther | Myoepithelial carcinoma | OGV vs MOOther |
| 10 | OGV | MOOther | OGV | MOOther | Mucoepidermoid carcinoma | OGV vs MOOther |
| 11 | MOMuc | OGV | OGV | MOMuc | Mucoepidermoid carcinoma | OGV vs MOMuc |
| 12 | PO | MOOther | MOOther | PO | Salivary duct carcinoma | PO vs MOOther |
MBFib=monomorphic basaloid neoplasm with fibrillary matrix, MBOther=monomorphic basaloid neoplasm with other matrix, MOCyst=monomorphic oncocytoid neoplasm with cystic background, MOMuc=monomorphic oncocytoid neoplasm with mucinous background, MOOther=monomorphic oncocytoid neoplasm with other background, OGV=oncocytoid neoplasm with granular/vacuolated cytoplasm, PB=pleomorphic basaloid neoplasm, PO=pleomorphic oncocytoid neoplasm, WT=Warthin tumor
Overall, the IOA analysis found substantial agreement between the four reviewers using the Griffith criteria (κ=0.69; 95% CI 0.67–0.71). There was almost perfect agreement for both pleomorphic adenoma (κ=0.92; 95% CI 0.86–0.98) and Warthin tumor (κ=0.83; 95% CI 0.77–0.89). Regarding basaloid neoplasms, there was moderate agreement for the categories of MBHy (κ=0.59; 95% CI 0.53–0.65), MBFib (κ=0.41; 95% CI 0.35–0.47), and MBOther (κ=0.41; 95% CI 0.35–0.47), and only slight agreement for PB (κ=0.11; 95% CI 0.05–0.17). There were 9 cases which were classified by even a single observer as PB, of which 7 were malignant (Table 4). For oncocytoid neoplasms, there was almost perfect agreement for the MOMuc category (κ=0.88; 95% CI 0.81–0.94), substantial agreement for OGV (κ=0.67; 95% CI 0.61–0.73) and PO (κ=0.63; 95% CI 0.57–0.69), moderate agreement for MOOther (κ=0.57; 95% CI 0.51–0.64), and only slight agreement for MOCyst (κ=0.18; 95% CI 0.12–0.24). There were 5 cases classified by even a single observer as MOCyst, of which only 1 was malignant (Table 5).
Table 4:
Cases classified by even a single reviewer as pleomorphic basaloid neoplasm.
| Case | Surgical pathology diagnosis | Reviewer interpretations | Consensus | |||
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | |||
| 1 | Pleomorphic adenoma | PB | MBOther | MBFib | MBFib | MBFib |
| 2 | Basal cell adenoma | MBOther | MBOther | PB | MBOther | MBOther |
| 3 | Carcinoma ex PA, NOS | PB | MBOther | PB | MBOther | None |
| 4 | Myoepithelial carcinoma | MBOther | PB | MOOther | PB | PB |
| 5 | Myoepithelial carcinoma | MOOther | PO | PB | PO | PO |
| 6 | Epithelial-myoepithelial carcinoma | MBFib | MBOther | PB | MBFib | MBFib |
| 7 | Epithelial-myoepithelial carcinoma | MBOther | PB | NDx | MBOther | MBOther |
| 8 | Basal cell adenocarcinoma | MBOther | MOOther | PB | MBOther | MBOther |
| 9 | Basal cell adenocarcinoma | MBOther | MBOther | PB | MBOther | MBOther |
MBFib=monomorphic basaloid neoplasm with fibrillary matrix, MBOther=monomorphic basaloid neoplasm with other matrix, MOOther= monomorphic oncocytoid neoplasm with other background, NDx=non-diagnostic, PB=pleomorphic basaloid neoplasm, PO=pleomorphic oncocytoid neoplasm
Table 5:
Cases classified by even a single reviewer as monomorphic oncocytoid neoplasm with cystic background.
| Case | Surgical pathology diagnosis | Reviewer interpretations | Consensus | |||
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | |||
| 1 | Oncocytic cystadenoma | WT | MOCyst | WT | MOCyst | None |
| 2 | Oncocytic cystadenoma | MOCyst | NDx | WT | MOCyst | MOCyst |
| 3 | Lymphoepithelial cyst | MBOther | Non-neo | MOCyst | MBOther | MBOther |
| 4 | Pleomorphic adenoma | PA | MBOther | MOCyst | MBOther | MBOther |
| 5 | Acinic cell carcinoma | OGV | MOCyst | OGV | OGV | OGV |
MBOther=monomorphic basaloid neoplasm with other matrix, MOCyst=monomorphic oncocytoid neoplasm with cystic background, NDx=non-diagnostic, non-neo=non-neoplastic, OGV=oncocytoid neoplasm with granular/vacuolated cytoplasm, PA=pleomorphic adenoma, WT=Warthin tumor
Discussion
Salivary gland FNAC can be challenging due to the wide array of diagnostic considerations and considerable overlap in cytomorphology among entities with vastly different clinical behavior. For years it has been the equivocal cytologic categories that have provided the most discordance among cytologists. Before MSRSGC publication, a survey of 468 pathologists, cytotechnologists, and pathology trainees was performed to determine the level of agreement between respondents for its various proposed categories(8). The survey utilized 75 salivary gland FNAC images which were felt by expert consensus to be reflective of the categories and which would subsequently be included as examples in the published MSRSGC. Compared to the expert consensus opinion, the respondents demonstrated the highest level of agreement for the Neoplasm-Benign category (58.9%) and the lowest agreement for Atypia of Uncertain Significance (17.7%), Salivary Gland Lesion of Uncertain Malignant Potential (23.6%), and Suspicious For Malignancy (22.7%)(8).
The MSRSGC system was established in part to address this kind of disparity and improve interobserver variability. Initial reports suggest mixed success on this front. With two reviewers applying MSRSGC criteria to 135 adequate salivary gland FNACs, Garg et al found disagreement in only 10 cases (unweighted Cohen’s Kappa score 0.812)(9). On the other hand, the equivocal categories do still appear to create challenges. In a study of seven reviewers and 134 cases, Hollyfield et al found fair agreement for the AUS category (Fleiss’ kappa 0.127, P<0.05) and only slight agreement for the SUMP category (Fleiss’ kappa 0.024 P=0.74)(10). The question then becomes which morphologic categories are particularly problematic.
In a study of the applicability of their classification scheme, Griffith et al found good overall IOA among four reviewers, with 24 of 42 cases (57.1%) demonstrating uniform or majority agreement. There was still minimal to no agreement on almost a third of cases (13/42, 30.9%), however. In their analysis, the authors reveal the surgical resection diagnoses associated with these challenging FNACs, but they did not seek to quantify the level of disagreement by the various morphologic categories. For the individual categories, then, the level of IOA remained largely unknown. Our study begins to address this issue by studying the IOA of challenging salivary gland FNAC cases in our institution to determine the extent of agreement for each category.
Risk of malignancy
The monomorphic basaloid neoplasm categories have demonstrated a relatively inconsistent ROM from one study to the next, possibly due to a limited number of cases in each group. We found a relatively high ROM for both the MBFib and MBHy categories compared to prior studies. This may be partly due to our having a particularly high-risk patient population, as the malignancy rate of our salivary gland FNAC cases is twice to three times that seen in other large academic centers(11). Regardless, overall the monomorphic basaloid neoplasm categories all appear to carry a significant enough ROM to warrant a SUMP categorization.
The ROMs for various oncocytoid morphologies in our study mirror those observed in prior studies using the same categories. We found a high ROM for the MOMuc category (92.3%, 12/13), with all malignant cases representing mucoepidermoid carcinomas. Our lone benign case echoes the prior impression that Warthin tumor with mucinous metaplasia is the primary alternate diagnosis to consider for this category(12). That a significant proportion of Warthin tumors can demonstrate such metaplastic changes has long been known(13), however our results confirm that even in the absence of clinical context, salivary gland FNAC specimens demonstrating oncocytoid cells in a background of abundant mucin might be considered as highly suspicious for malignancy, a finding suggested by prior studies utilizing these criteria(5,12).
Similarly, the high ROM for the OGV category in combination with the substantial agreement in its use raise questions regarding how this category should be properly classified under the tiered MSRSGC system. One issue with the application of MSRSGC is that there appears to be some overlap between its SUMP subcategories of “Oncocytoid neoplasm with granular/vacuolated cytoplasm” and “Cellular neoplasm with clear cell features.” The latter is described as having “neoplastic cells with clear cell features: clear, foamy, granular, or vacuolated cytoplasm, or any combination thereof; features are not characteristic of true oncocytes,” and the figures provided as examples included three examples of acinic cell carcinoma. This morphology description could easily apply to any of the cases in our study categorized as OGV which indeed included all 9 cases of acinic cell carcinoma and the lone case of secretory carcinoma. Additionally, the differential diagnosis provided by MSRSGC for the OGV category includes acinic cell carcinoma, secretory carcinoma, and metastatic renal cell carcinoma. It would seem, then, that these two SUMP subcategories could be condensed into a single one. With its high ROM, one might even argue that the OGV category may be better classified along with the MOMuc category under the Suspicious for Malignancy category or higher. That these categories showed high IOA only lends credence to this idea.
Interobserver agreement
Overall, there was substantial agreement for the select morphologic categories in the IOA analysis. Categories with particularly strong agreement included MOMuc, OGV, PO, and MBHy. The IOA analysis did identify two problematic categories, however: MOCyst and PB.
On further analysis, the MOCyst category appeared to overlap with several other morphologic subcategories. Of the cases interpreted by any reviewers as MOCyst, there was only a single malignant case (acinic cell carcinoma), and this was interpreted as MOCyst by just a single reviewer, with the remaining three reviewers interpreting it as OGV. While this interpretation would have contributed to a lower presumed risk of malignancy for this patient, it likely would not have significantly impacted care, as the proper management for this low grade carcinoma is usually a limited resection. Regardless of the reproducibility of this category, in combination with those of Griffith et al, our results suggest that it is associated with a low risk of malignancy. In theory, the main salivary gland malignancy to exclude in this setting would be a low-grade mucoepidermoid carcinoma, as these can present as purely cystic contents, making them a notorious cause of false negative salivary gland FNAC(5,11,14). In practice, however, this did not bear fruit among our cases; none of the mucoepidermoid carcinomas in our study were interpreted as fitting the MOCyst category by any of the reviewers.
The primary issue regarding the use of the PB category appeared to be defining pleomorphism among the basaloid neoplasms, as by far the most common competing categories among cases with at least one PB interpretation were monomorphic basaloid categories. Perhaps better agreement could be achieved with a different description of the category, which was initially defined as containing “pleomorphic nuclei suggestive of a high-grade malignancy;” more specific criteria may improve IOA. In a study restricted to basaloid neoplasms, the presence of prominent nucleoli or even a single mitotic figure were both found exclusively in malignant cases, and the presence of necrosis/apoptotic debris or anisonucleosis ≥2:1 were also strongly predictive of malignancy(15). Future studies would be required to determine whether these more specific criteria could improve IOA. In the meantime, individual institutions may benefit from pursuing a consensus agreement regarding the level of atypia required to consider a basaloid neoplasm pleomorphic. Cytologists should in any event take solace in the knowledge that in our cohort, there were 9 cases interpreted by at least one reviewer as PB, and 7 of these proved to be malignant, so even a single reviewer identifying a case as such suggested a relatively high ROM (77.8%). Still, had the two benign cases that were each interpreted by a single reviewer as PB been reported as consistent with a high-grade carcinoma, the patients may have been subject to unnecessary radical resection and/or neck dissection.
Limitations
There were a few limitations to our study. One limitation is that in utilizing the Griffith criteria, we included a category that combined monomorphic basaloid neoplasms with mixed stroma with those that had no stroma (MBOther). The Milan system has recommended separating these categories, and had we done so, we would have been able to assess whether there was a difference in ROM between the two. The study by Gargano et al suggested that basaloid neoplasms without stroma may have a particularly high ROM. Should further study confirm this, it would justify its inclusion as a separate morphologic subcategory. Our analysis also excluded metastatic and hematolymphoid malignancies, so we are unable to determine the extent to which their inclusion in the various morphologic categories might alter our results.
Conclusion
In summary, we utilized the salivary gland FNAC categorization scheme proposed by Griffith et al in order to retrospectively classify cases with known surgical follow-up at our institution. We found substantial reproducibility using this system overall, with particularly high agreement for diagnoses such as MOMuc, OGV, PO, and MBHy. The two categories with only slight agreement included PB and MOCyst. A majority of the cases interpreted by even a single reviewer as PB were malignant, while a majority those interpreted by even a single observer as MOCyst were benign. Our study finds a particularly high ROM for the PB, PO, MOMuc and OGV categories. While the former two categories are recognized to be at least suspicious under Milan criteria, the finding of high ROM for the latter two categories in conjunction with their high level of IOA suggest that these morphologies might be best classified by the Milan system under category V: Suspicious for malignancy.
Acknowledgments
Funding Statement: Research reported in this publication was supported in part by the Cancer Center Support Grant of the National Institutes of Health/National Cancer Institute under award number P30CA008748. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
This work is original and has not been published elsewhere nor is it currently under consideration for publication elsewhere.
References
- 1.Rossi ED, Faquin WC, Baloch Z, Barkan GA, Foschini MP, Pusztaszeri M, et al. The Milan System for Reporting Salivary Gland Cytopathology: Analysis and suggestions of initial survey. Cancer Cytopathol. 2017;125(10):757–66. [DOI] [PubMed] [Google Scholar]
- 2.Schmidt RL, Hunt JP, Hall BJ, Wilson AR, Layfield LJ. A systematic review and meta-analysis of the diagnostic accuracy of frozen section for parotid gland lesions. Am J Clin Pathol. 2011;136(5):729–38. [DOI] [PubMed] [Google Scholar]
- 3.Liu CC, Jethwa AR, Khariwala SS, Johnson J, Shin JJ. Sensitivity, Specificity, and Posttest Probability of Parotid Fine-Needle Aspiration: A Systematic Review and Meta-analysis. Otolaryngol - Head Neck Surg (United States). 2016;154(1):9–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Johnson DN, Onenerk M, Krane JF, Rossi ED, Baloch Z, Barkan G, et al. Cytologic grading of primary malignant salivary gland tumors: A blinded review by an international panel. Cancer Cytopathol. 2020;1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Griffith CC, Pai RK, Schneider F, Duvvuri U, Ferris RL, Johnson JT, et al. Salivary Gland Tumor Fine-Needle Aspiration Cytology: A Proposal for a Risk Stratification Classification. Am J Clin Pathol. 2015;143(6):839–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rossi ED, Wong LQ, Bizzarro T, Petrone G, Mule A, Fadda G, et al. The impact of FNAC in the management of salivary gland lesions: Institutional experiences leading to a risk-based classification scheme. Cancer Cytopathol. 2016;124(6):388–96. [DOI] [PubMed] [Google Scholar]
- 7.Faquin WC, Rossi ED, Baloch Z, Barkan GA, Foschini MP, Kurtycz DFI, et al. The Milan System for Reporting Salivary Gland Cytopathology. First. Cham, Switzerland: Springer International Publishing AG; 2018. 1–182 p. [Google Scholar]
- 8.Kurtycz DFI, Rossi ED, Baloch Z, Pavelec D, Madrigal E, Vielh P, et al. Milan Interobserver Reproducibility Study (MIRST): Milan System 2018. J Am Soc Cytopathol [Internet]. 2020;In Press. Available from: 10.1016/j.jasc.2019.12.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Garg K, Karnezis AN, Rabban JT. Uncommon hereditary gynaecological tumour syndromes: pathological features in tumours that may predict risk for a germline mutation. Pathology. 2018;50(2):238–56. [DOI] [PubMed] [Google Scholar]
- 10.Hollyfield JM, O’Connor SM, Maygarden SJ, Greene KG, Scanga LR, Tang S, et al. Northern Italy in the American South: Assessing interobserver reliability within the Milan System for Reporting Salivary Gland Cytopathology. Cancer Cytopathol. 2018;126(6):390–6. [DOI] [PubMed] [Google Scholar]
- 11.Lubin D, Buonocore D, Wei XJ, Cohen JM, Lin O. The Milan System at Memorial Sloan Kettering: Utility of the categorization system for in-house salivary gland fine-needle aspiration cytology at a comprehensive cancer center. Diagn Cytopathol. 2020;48(3):183–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Griffith CC, Schmitt AC, Pantanowitz L, Monaco SE. A pattern-based risk-stratification scheme for salivary gland cytology: A multi-institutional, interobserver variability study to determine applicability. Cancer Cytopathol. 2017;125(10):776–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Goonewardene SA, Nasuti JF. Value of mucin detection in distinguishing mucoepidermoid carcinoma from Warthin’s tumor on fine needle aspiration. Acta Cytol. 2002;46(4):704–8. [PubMed] [Google Scholar]
- 14.Klijanienko J, Vielh P. Fine-needle sampling of salivary gland lesions IV.: Review of 50 cases of mucoepidermoid carcinoma with histologic correlation. Diagn Cytopathol. 1997;17(2):92–8. [DOI] [PubMed] [Google Scholar]
- 15.Gargano SM, Sebastiano C, Solomides CC, Griffith CC, HooKim K. Cytohistologic correlation of basaloid salivary gland neoplasms: Can cytomorphologic classification be used to diagnose and grade these tumors? Cancer Cytopathol. 2020;128(2):92–9. [DOI] [PubMed] [Google Scholar]