Warthin’s Tumour Diagnostic Outcome in the Milan System Era and Cytomorphological Pitfalls

Warthin’s tumour (WT) is the second most common salivary gland neoplasm after pleomorphic adenoma (PA), comprising 5%–15% of all salivary gland tumours [1]. Recent studies have shown an increasing incidence of WT. Most people with WT have a history of smoking and are in their sixth to seventh decades of life. In the past, male individuals had a significantly higher risk of developing WT, but recently, the gap between men and women has narrowed, probably because of the increased smoking prevalence among female individuals [2‒5].

The updated second version of the Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) was published in 2023 [6]. Goals of the system were to provide a consistent classification system for salivary gland fine-needle aspiration (SG-FNA) specimens and to improve communication between professionals. SG-FNAs are classified according to the specific diagnostic criteria in six MSRSGC categories: (I) nondiagnostic, (II) non-neoplastic (NN), (III) atypia of undetermined significance (AUS), (IVa) benign neoplasm (BN), (IVb) salivary gland neoplasm of uncertain malignant potential (SUMP), (V) suspicious for malignancy, and (VI) malignant neoplasm [6‒8].

WT is usually categorized as BN according to the MSRSGC [6]. The three classic cytomorphological features of WT are the presence of oncocytic epithelial cells, lymphocytes, and granular background material [1, 6]. With the presence of these classic cytomorphological features, the diagnostic accuracy of FNAs for WT diagnosis is >95% [6]. False-negative (FN) cases are often categorized as AUS or SUMP, according to the MSRSGC. Squamous metaplasia and/or mucinous metaplasia are the leading causes of undetermined WT categorization [9, 10]. A previously published meta-analysis of WT diagnosis performance reported that the most common FN cases were cytologically diagnosed as cysts, PA, and mucoepidermoid carcinomas, whereas the most common false-positive (FP) cases were acinic cell carcinomas, mucoepidermoid carcinomas, oncocytomas, and PAs histologically [11].

Aims of this study were to evaluate the diagnostic accuracy of FNA in diagnosing WT during a 10-year period and to compare the five pre-MSRSGC years when Pap classes were used to the initial five MSRSGC years. In addition, the focus of this study was to determine which WT cytological features led to AUS or SUMP categorization in our practice.

A search of a laboratory information system was performed at the Department of Pathology, Fimlab Laboratories, Tampere, Finland, to find all salivary gland WT cytological and histological diagnoses during a 10-year period (January 1, 2013 to December 31, 2022). All histological WT diagnoses without matched cytological diagnoses were excluded. Cases with a cytological or histological diagnosis of WT were included in the study cohort, and the cytological and histological diagnoses were matched. To assess the diagnostic accuracy of FNA, we excluded cases with insufficient cytological diagnosis. For each case, the patient’s age and sex and lesion location and size were tabulated.

A radiologist performed ultrasound-guided FNAs with 22G needles. All preparations were alcohol fixed. Cytospin preparations were made and slides stained with Papanicolaou stain. Cell blocks (CBs) were prepared using four different methods: plasma-thrombin method, collection of visible tissue fragments, the Shandon/Epredia method (Thermo Fisher Scientific, USA), and the in-house method [12‒14]. Specimens were classified according to the Papanicolaou classification system in 2013–2017 and according to the MSRSGC in 2018–2022.

Histological diagnoses were considered the golden standard to place cases into three different categories: (i) true-positive (TP), when cytological and histological diagnoses were WT; (ii) FP, when cytological diagnoses were WT and histological diagnoses were other than WT; and (iii) FN, when cytological diagnoses were other than WT and histological diagnoses were WT. Cases with cytological and histological diagnoses other than WT were considered true negative. Diagnostic accuracy, sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios were calculated.

To conduct cytomorphological analysis, all FN cases that could be retrieved from the archives with a cytological classification of AUS or SUMP were reassessed and compared with randomly selected TP cases categorized as BN. (i) The background features analysed were as follows: necrosis, haemorrhage, cystic haemorrhage, inflammation, granular cystic debris, macrophages, lymphocytes, plasma cells, neutrophils, adipocytes, fibroblasts, other inflammatory cells, crystals and crystalloids, debris, keratinous debris, keratin, and mucinous like. (ii) The following architectural features were evaluated: hypocellularity, cellularity (diffuse hypercellularity), large intact cellular fragments, sheets, honeycomb sheets, formation of trabeculae, papillae, cell size variability, discohesivity, two cell populations, small groups of oncocytes, and single cells. (iii) Among the cellular features analysed were the following: monotony of the cell population, spindled cells, ciliated cells, columnar cells, cuboidal cells, oncocytic cells, oncocytoid cells, mucinous metaplasia and squamous cell metaplasia, cyst-lining cells, giant cells, acini, cytoplasmic granules, and cytoplasmic vacuoles. (iv) The nuclear characteristics analysed were enlargement, pleomorphism, size variability, hyperchromasia, crowding, elongation, grooves, irregular membrane, intranuclear inclusions, pseudoinclusions, pale powdery chromatin, presence of nucleoli, presence of small eccentric nucleoli, atypia in oncocytes, atypia in squamous cells, mitoses, and apoptotic bodies.

Data were tabulated in Microsoft^® Excel^® for Microsoft 365 MSO (version 2111 Build 16.0.14701.20254) 64 bits. Statistical analysis was performed using IBM SPSS Statistics (version 22.0; SPSS, IBM, Armonk, NY, USA). Descriptive statistics were used to analyse the data. p values were calculated using the Pearson chi-square test. p values ≤0.05 were considered statistically significant.

The Ethical Committee of the Pirkanmaa Hospital District approved the study (R17174). This study was performed in agreement with the Helsinki Declaration, and individual patient consent was not requested after approval by the Ethics Committee of Pirkanmaa Hospital District (R17174).

Overall, 1,496 salivary gland samples with a histological diagnosis were obtained over 10 years. Of these, a histological diagnosis of WT was made in 165 (11.0%) cases. Of those with a histological diagnosis of WT, 54 (37.0%) were female and 92 (63.0%) were male. The median patient age was 65.0 years (range: 43–87 years) and the average lesion size ± standard deviation was 2.7 cm ± 1.0 cm (range: 0.6–6.5 cm). Topographically, 163 (98.8%) cases were from the parotid gland and 2 (1.2%) were from the submandibular gland.

During the 10 years, there were 1,347 true-negative cases, 113 TP cases, 52 FN cases, and 3 FP cases. The overall diagnostic accuracy was 96.4%, sensitivity was 68.5%, and specificity was 99.8%. Sensitivity was slightly higher during the pre-MSRSGC years compared with the MSRSGC years (69.6% vs. 67.4%), whereas the specificities were similar (99.8% in 2013–2017 vs. 99.7% in 2018–2022). Table 1 shows the diagnostic accuracy during the overall 10-year study, and the study period was divided into pre-MSRSGC years (2013–2017) and MSRSGC years (2018–2022).

Overall, there were 55 (33.3%) falsely diagnosed cases. In our series, there were 52 (31.5%) FN cases and 3 (1.8%) FP cases. Histological diagnoses of FP cases included PA, reactive changes, and “no diagnostic abnormality.” There were no malignant cases leading to a 0.0% risk of malignancy (ROM) for cytologically diagnosed WT. During the pre-MSRSGC years, there was 1 NN case leading to a 98.2% risk of neoplasm (RON) and 1 NN case during the MSRSGC years resulting in a RON of 98.3%. For the entire study cohort, the RON was 98.3%.

FNA specimens were categorized according to the Papanicolaou classification system during the first 5 years of the study period (2013–2017) (pre-MSRSGC). Of all the FN cases, the most common cytological diagnoses were “morphologic description only” categorized as Pap class 2 in 11 cases, “neoplasm, NOS” categorized as Pap class 2 in 3 cases, and “neoplasm, NOS” categorized as Pap class 3 in 3 cases. Table 2 shows all FN cytological diagnoses of cases and their categorizations during the first 5 years (2013–2017).

During the next 5 years (2018–2022), all cases were categorized according to the MSRSGC. All TP cases were categorized as BN. FN cases were categorized as AUS (n = 22) and SUMP (n = 5), and 1 case was signed out without the MSRSGC categorization.

Separate cytomorphological features of 17 AUS cases, 5 SUMP cases, and 21 TP BN cases were evaluated. Among the features of cytological specimens studied, the following cytomorphology can distinguish between FN cases and TP cases: background of necrosis in SUMP cases (p < 0.001) and inflammatory background in BN cases (p = 0.052). Among architectural features, statistical significance was observed for cellularity (diffuse hypercellularity) (p = 0.020) and honeycombed sheets (p = 0.020) in the SUMP category and the presence of papillae (p = 0.011) and small groups of oncocytic cells (p = 0.033) in the BN category (Table 3) (Fig. 1).

In a separate analysis of CB samples, background lymphohistiocytic aggregates were most common in the BN category (42.9% of BN cases vs. 11.8% of AUS cases and 0.0% of SUMP cases, p = 0.035) and neutrophils were the dominant cell type in the SUMP category (80.0% of SUMP cases vs. 35.3% of AUS cases and 14.3% of BN cases, p = 0.014). Among the architectural cytomorphological features, papillary fragments were present in most BN cases (95.2%), but only in 5.9% of AUS cases, and there was no papillary architecture present in SUMP cases (p < 0.001). Spindled cells were seen in 1 SUMP case and in none of the AUS or BN cases (p = 0.020). Oncocytic cells were identified in every TP case, 58.8% of AUS cases, and 60.0% of SUMP cases (p = 0.004) (Fig. 2). The cytomorphological features of the CBs are summarized in Table 4.

Lymphocytes (100.0% vs. 83.7% of all cases, p = 0.012) and neutrophils (53.5% vs. 30.2% of all cases, p = 0.048) were more common in cytological specimens than in CB specimens. In addition, sheets (37.2% vs. 4.7% of all cases, p < 0.001), trabecular architecture (30.2% vs. 2.3% of all cases, p = 0.001), and small groups of oncocytic cells (46.5% vs. 18.6% of all cases, p = 0.011) were more common in cytological specimens compared with CB specimens. In contrast, in BN cases, papillary architecture was more common in CB samples than in cytological specimens (95.2% vs. 57.1%, p = 0.009) (Table 5).

This study showed that the diagnostic performance for diagnosing WT was similar in the pre-MSRSGC and MSRSGC years, with a high overall specificity (99.8%) and moderate sensitivity (68.5%). FN cases were diagnosed as AUS or SUMP during the MSRSGC years.

The overall diagnostic accuracy of this study was 96.4%, which was similar to the diagnostic accuracy of 96.1% reported by Allison et al. [10] based on WTs diagnosed according to the MSRSGC. Fisher and Ronen [11] reported a meta-analysis of 17 studies with data from pre-MSRSGC years, where WT diagnosis had a sensitivity of 93.9% and a specificity of 97.9%. The range of sensitivity in the studies included in the meta-analysis was noticeably greater than the specificity (40.0%–100.0% vs. 90.6%–100.0%) [11]. In our study, the sensitivity was also lower than the specificity, but there was no significant difference between the pre-MSRSGC and MSRSGC years.

Furthermore, in the meta-analysis by Fisher and Ronen [11], the ROM for WT diagnosis was 4.7%. In contrast, the ROM was only 1.3% in a study by Allison et al. [10]. In our study, there were no malignant cases, resulting in a ROM of 0%. Although the MSRSGC does not officially assess the RON, Fisher and Ronen [11] reported that the RON for WT diagnosis was 99.2%; for Allison et al. [10], it was 99.0%. The RON in the present study was slightly lower, with two NN FP cases leading to an RON of 98.3%.

In this analysis, the number of discordant cases was quite high during the pre-MSRSGC and MSRSGC years. The implementation of the MSRSGC system has significantly improved communication between cytopathologists and clinicians by providing standardized terminology and clear recommendations for the management of specific diagnostic categories. This standardization enables more precise communication of diagnostic findings and risk stratification, aiding clinicians in making appropriate therapeutic decisions, such as choosing between surgical intervention and conservative management. For the AUS category, the recommended management is repeat FNA or surgery, and for the SUMP category, it is surgery [6]. For the BN category, surgery is also the recommended management method, although WT can also be followed up without surgery in some cases, especially for older individuals. WT is typically slow growing and appears in older people with many comorbidities, and the risk of malignant transformations of WT is low, leading to conservative management in the elderly [15‒17]. The challenge in the pre-MSRSGC years was that FN cytological diagnoses did not indicate stratification risks to clinicians. In this study, a cytological diagnosis of “neoplasm, NOS” occurred in Pap classes 2 and 3. In particular, a diagnosis of “only morphologic description” is problematic for clinicians because of the lack of relevant information on how to manage the lesion.

WT is the most common diagnosis in the AUS category as seen in our previous study [18] and in other studies [19‒21]. In a study by Torous et al. [9], 10.5% of histologically verified WTs were categorized as AUS. In that study, the reasons for categorizing cases as AUS were squamous changes in 5 cases, and 8 cases showed hypocellularity specifically as a limiting factor for diagnostic interpretation. In the present study, 22 FN cases were categorized as AUS, and the main cytomorphological reasons for FNAs to be categorized as AUS were the lack of papillae and the presence of small groups of oncocytic cells. In addition, when cystic degeneration was not present in cytological specimens but only in cell blocks, it often led to AUS categorization. A diagnosis of WT also depends on the experience of the cytopathologist [22, 23], and in the re-evaluation, some cases that were categorized as AUS had characteristics that would allow them to be categorized as BN.

A study by Salehi et al. [24] reported a case of histologically verified WT categorized as NN. The case had lymphoid hyperplasia with mixed benign salivary glands and acellular debris, and a differential diagnosis of sialadenitis or retention cyst was made. In our study, there were no WT cases in the NN category.

WT is also diagnosed in the SUMP category. In our previous study of the SUMP category, WT was the most common benign histological diagnosis after PA (unpublished data). Allison et al. [10] reported that 54.5% of WT FN cases were categorized as SUMP. In a study by Torous et al. [9], 3.2% of WT cases verified histologically were classified as SUMP on the basis of squamous and mucinous changes. In our study, SG-FNAs were categorized as SUMP on the basis of necrosis. Furthermore, neutrophils were present in most SUMP cases, which increases concern about necrosis. Some cases present with oncocytoid cells, which leads to a diagnosis of SUMP. In addition, cellularity (diffuse hypercellularity) was one reason for the categorization of WT as SUMP because it increased the suspicion of malignancy. O'Dwyer et al. [25] showed that inflammation, fibrosis, and necrosis in WT close to the branches of the facial nerve can cause facial paralysis.

Mucinous changes in WT raise concerns about Warthin-like mucoepidermoid carcinoma [26] and lead to categorization as SUMP or an even higher diagnostic category. Squamous metaplasia in WT is a common diagnostic pitfall of squamous cell carcinoma [27]. FNA can cause infarction in WT cases, leading to squamous metaplasia, which interferes with the diagnostic process [28]. In addition, a complete infarction can lead to tumour disappearance in FNAs or in imaging studies, making diagnosis complicated [29]. Vanishing tumours are often related to the presence of oncocytes [30]. Acinic cell carcinoma with lymphoid-rich stroma is another common pitfall in WT diagnosis due to dual components [27]. Orell et al. [31] study found that acinic cell carcinoma and mucoepidermoid carcinoma were diagnosed falsely as WT due to the presence of sheets of bland epithelial cells with oncocyte-like cytoplasm. In our analysis, there were no malignant cases.

When using classic cytomorphological features, a diagnosis of WT is very accurate [6]. In our study, papillary fragments in cell blocks were the most useful cytomorphological feature that led to the correct diagnosis. In addition, small groups of oncocytes and cystic degeneration were seen in most TP cases. Klijanienko et al. [32] study found that classic cytological components were found in 32% of histologically verified WTs. In addition, two classic components (oncocytes and lymphocytes) were found in 66% of cases. In contrast, in our analysis, a lack of papillae, the presence of small groups of oncocytes, and cystic degeneration lead to WTs being diagnosed as AUS.

There was a lack of classic cytomorphological features of WT in SUMP cases, and the presence of necrosis and diffuse hypercellularity led to some cases being categorized as SUMP. Accurate diagnosis is important because the BN category is associated with a significantly lower ROM than the AUS and SUMP categories. A meta-analysis of seven prospective studies showed that ROM for the BN category was 2.4% versus 34.9% for AUS and 36.6% for SUMP. In addition, there were differences between RONs (71.3% for AUS, 98.9% for BN, and 97.8% for SUMP) [33].

In this study, 1 FP case was diagnosed as PA histologically, and a study by Allison et al. [10] reported that PA was the most common FP diagnosis for WT (n = 3). In a meta-analysis by Fisher and Ronen [11], PA was the most common FN diagnosis after cysts and the most common FP BN diagnosis after oncocytoma. Jechová et al. [17] showed that 0.5% of cytologically diagnosed WTs were PAs histologically.

In conclusion, this study showed that the diagnostic accuracy and specificity of WT diagnosis by FNAs were very high, although the sensitivity was lower. The MSRSGC improves communication among cytopathologists and clinicians, but diagnostic outcomes were similar in pre-MSRSGC and MSRSGC periods. Cytomorphological analysis identified key pitfalls, such as the absence of papillae, the presence of small groups of oncocytes, and a lack of cystic degeneration, which often resulted in WTs being falsely diagnosed as AUS. Furthermore, necrosis, oncocytoid cells, and diffuse hypercellularity increased the suspicion of malignancy, leading to SUMP categorization.

We thank J. Ludovic Croxford, PhD, from Scribendi (www.scribendi.com) for editing a draft of this manuscript.

The Ethical Committee of the Pirkanmaa Hospital District approved the study (R17174). This study was performed in agreement with the Helsinki Declaration. The need for written informed patient consent was waived by the Ethics Committee of Pirkanmaa Hospital District (decision reference number R17174).

The authors have no conflicts of interest to declare.

This study was funded by a VTR grant from Pirkanmaa Hospital District (to I.K.), the Emil Aaltonen Foundation (to H.L.), and the Charles University Cooperation Program, research area “Surgical Disciplines” (to D.K.). Funders had no role in the design, data collection, data analysis, and reporting of this study.

H.L.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing – original draft, and writing – editing; D.K.: formal analysis, methodology, and writing – editing; I.K.: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, visualization, supervision, validation, writing – original draft, and writing – editing.

All data generated or analysed during this study are included in this article. Further enquiries can be directed to the corresponding author.