Efficacy of Fine-Needle Aspiration Cytology in Diagnosing Secretory Carcinoma of Salivary Gland: A Systematic Review and Meta-Analysis Free

Skalova described mammary analog secretory carcinoma (MASC) in salivary gland, in 2010, which was later designated as secretory carcinoma (SC) in 2017 and secured its place as a distinct entity in the World Health Organization classification of Head and neck tumors [1]. Since then, only a few hundred cases have been reported in the literature [2]. SC of salivary gland is a low-grade carcinoma which presents as a slow-growing painless mass in the parotid, submandibular gland, accessory parotid, peri-parotid lymph node and minor salivary gland of lip, buccal mucosa, hard palate, and soft palate [3‒7]. The median duration of symptoms is 14 months. These tumors have a slight male predilection and can affect any age-group, ranging from 9 to 78 years [4, 5, 8‒10]. The SC in salivary gland has a mean size of 2.1 cm, with a more frequent T1 or T2 stage at the time of presentation. Regional lymph node metastasis is uncommon while distant metastasis is rare, but has been reported [1, 5‒8, 11].

Earlier, these indolent neoplasms were diagnosed as acinic cell carcinoma. These neoplasms have papillary and microcystic patterns, which are unusual patterns to encounter in acinic cell carcinoma. These neoplasms harbor the characteristic translocation t(12;15) (p13;q25) resulting in an ETV6-NTRK3 fusion gene. Although histomorphology with supportive immunohistochemistry has been suggested to be sufficient for the diagnosis of SC, demonstration of the classical recurrent cytogenetic abnormality of t(12;15) (p13;q25) ETV6-NTRK3 is currently considered the gold standard for the diagnosis of this entity [3]. This cytogenetic abnormality can be identified by fluorescence in situ hybridization (FISH) or reverse transcription PCR (RT-PCR) [11]. In contrast to these costly and not so widely available molecular techniques, fine-needle aspiration cytology (FNAC) is a readily available, rapid, safe, and cost-effective outpatient procedure, which can be utilized in the diagnosis of the SC of salivary gland. However, less is known about the specific cytomorphological features of the SC. Identification of specific cytomorphological features along with triaging of specimen for cell block, which can be subjected to further immunochemistry and molecular diagnosis, will provide better diagnostic strategy.

This review aims at identifying the efficacy of FNAC in diagnosing SC of salivary gland. Secondary objective was to study the cytomorphological features of SC (salivary gland) as reported in literature. This review will help the cytopathologists to know when to suspect SC on cytology and how to accurately diagnose this entity on cytology.

For this systematic review, PRISMA guidelines were followed (Fig. 1) [12]. PubMed/MEDLINE, Embase, Science Direct, Cochrane review, and PROSPERO databases were searched for studies published between January 2010 and June 2023. The former year was chosen because SC of salivary gland was first identified in 2010. The citations of all included studies were also searched for any suitable articles. The authors were contacted if screened full texts could not be retrieved of, for any missing information.

The search strategy was to find articles describing the cytomorphological features of the SC of salivary gland. The key terms used for data search were (“secretory carcinoma of salivary gland” OR “mammary analogue secretory carcinoma of salivary gland”) AND (“Cytology”OR “Cytological features” OR “aspirate” OR “cytodiagnosis”).

We have included the studies reporting the cytodiagnosis and cytomorphological features of molecularly confirmed SC of the salivary gland in the analysis.

As diagnosis of SC of salivary gland on cytology is rare, case reports, and case series, along with original articles were included in the analysis. Only literature review, letter to editor, SC of any other organ, reports without cytomorphological features, and reports of SC without molecular confirmation were excluded. All articles, irrespective of the language of the manuscript, were included. The selection of articles was done by PSK and MG. Any disagreements were resolved by discussion with third reviewer, NT.

Two reviewers independently extracted information using a data extraction form. Information on study setting, year, country of publication, sample size, the original cytodiagnosis offered, the cytological features, any supportive technique (special stains, immunocytochemistry, cell-block preparation), histopathological diagnosis, and molecular confirmation were recorded. The molecular diagnosis was considered as the gold standard modality. A case was considered “positive” if SC was the final diagnosis or among the differential diagnosis offered on cytology. Further, we derived and combined the true positives, false positives, false negatives, and true negatives as reported in different studies, to generate a two-by-two table. The Open Meta (Analyst) software was used for obtaining summary estimates of sensitivity, likelihood ratios (LR+ and LR−), with 95% confidence interval and diagnostic odds ratio [13]. Evidence of inter-study variance due to heterogeneity was assessed by χ² test and I² test. The quality of the included studies was assessed by Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [14]. This tool has four domains: patient selection, index test, reference standard, and flow and timing. The existence of publication bias was evaluated by Funnel plot [15]. Funnel plot was generated by RevMan version 5.4.1 (Review Manager 5.4.1, The Cochrane Collaboration, Copenhagen, 2020).

Ethical approval was not required for the study as this was a systematic review of the published articles and metanalysis was performed on the secondary data. Hence, patient’s informed consent was also not applicable for the review.

A total of 413 studies were identified based on the titles and abstracts obtained via the data search. On further evaluation, 52 duplicates were excluded, 344 were excluded on the basis of the exclusion/inclusion criteria. Finally, a total of 17 studies were included in the evaluation (Fig. 1) [11, 16‒31].

A total of 17 studies/case reports included in the review yielded 45 molecularly confirmed cases of SC of salivary gland which also underwent FNAC. All studies were published in English only.

Figure 2 shows the risk of bias assessment across various domains as per QUADAS-2 tool results. It was found that almost all of the studies included in this review had a “low” to “unclear” risk of bias and concern regarding applicability across all domains. Funnel plot showed a minimal publication bias (Fig. 3).

The clinical data of all cases is shown in Tables 1 and 2. The age range of patients varied from 15 to 74 years (median = 47 years; mean = 45.07 ± 17.49 years). The male: female ratio was 1.25 (p = 0.456). The commonest site of occurrence of SC in salivary gland was parotid (n = 42/45; 93.3%), followed by submandibular gland (n = 2/45; 4.4%) and buccal mucosa (n = 1/45; 2.2%). The tumor had a predilection for parotid gland as the primary site as compared to other major and minor salivary glands (p < 0.00001). The size of the tumor varied from 0.7 to 9.8 cm; average size being 2.17 ± 1.49 cm.

In 15 out of 45 cases, staging was mentioned, the tumor was of stage I/T1 in almost 3/4th of these cases (11/15; 73.3%), followed by stage III/T3 (2/15; 13.3%) at time of diagnosis. This was statistically significant (p = 0.020). However, in majority of cases (30/45) staging was not reported.

On cytology, SC was correctly diagnosed/differentially diagnosed in only 7 cases. As the level of heterogeneity was mild (I²<25%), fixed-effect model was used for statistical analysis. It was found that FNAC had a pooled sensitivity of 27.7% (95% CI: 16.6–42.5%) in diagnosing SC of salivary gland. The LR+ was 0.654 (95% CI: 0.344–1.245), LR− was 1.023 (95% CI: 0.538–1.946) and diagnostic odds ratio (DOR) was 0.421 (95% CI: 0.129–1.374). The Forest plots for sensitivity, LR+, LR−, and DOR are shown in Figure 4. Since there were no true negatives; the specificity could not be calculated. This was the shortcoming of the current metanalysis.

A range of both benign and malignant diagnosis/differential diagnosis were given on initial cytological evaluation in different studies viz. Acinic cell carcinoma, mucoepidermoid carcinoma, adenocarcinoma, atypia, low-grade salivary gland neoplasm, oncocytic neoplasm, pleomorphic adenoma, and monomorphic adenoma, due to variable and overlapping cytomorphological features. A confident diagnosis of SC on cytology was made in three cases (6.6%), while in four cases it was in differentials (Table 2).

Papillary structures followed by clusters, tubular arrangement, and singles were common architectural features (Fig. 5; Table 3). Characteristic finding was finely vacuolated cytoplasm observed in 91% cases (41/45), in two cases cytoplasmic vacuolation was not seen and in two cases it was not mentioned. Background mucin/metachromatic stroma was observed in 60% (27/45) cases leading to diagnostic dilemma with mucoepidermoid carcinoma and pleomorphic adenoma.

Cell block from the aspirate was prepared in 13 cases with immunohistochemistry in four cases and molecular FISH in one case (Table 4). Immunohistochemistry was performed on cell block from aspirates and resection specimens. S100 and mammaglobin were positive in 100% (41/41) cases, in other four cases it was not performed. Other common immunomarkers performed were p63, SMA, EMA, MUC1/MUC4, and keratins in different cases; however, expression of latter was not significant.

Molecular profiling for ETV6-NTRK3 fusion/ETV6 rearrangement is the gold standard for diagnosis and criteria for selection of cases in the review. FISH was done in majority of cases (75.6%) followed by PCR (22.2%) and next-generation sequencing (NGS) (2.2%).

Reasons behind the correct cytodiagnosis of SC were cytomorphology alone (2/7), cytomorphology with immunocytochemistry (1/7), supportive immunohistochemistry on cell block (1/7), histochemistry and immunohistochemistry on cell block (1/7) and diagnostic FISH on cell block (1/7). Second opinion from another specialized center helped in one of the cases. The most commonly used immunological (IHC/ICC) markers used on cell-block and FNAC smears were S100 and mammaglobin.

Before this entity was introduced in 2010 by Skalova, these tumours were misclassified as acinic cell carcinoma, mucoepidermoid carcinoma, adenocarcinoma (NOS), or pleomorphic adenoma [4, 32]. In a study by Terada et al. [8], it was found that of 24 salivary gland neoplasm diagnosed as acinic cell carcinoma, when tested for ETV6-NTRK3 fusion, revealed to include 14 misdiagnosed cases of SC. SC needs to be differentiated from other salivary gland neoplasms preoperatively, so that a specific biological behavior can be predicted and a targeted therapy can be offered in selected cases [27].

Overall, the diagnostic accuracy of FNAC in salivary gland is around 95%; however, its performance in the diagnosis of SC is poor [33]. The sensitivity and specificity in diagnosis of malignancy in major salivary gland varies from 38% to 97% and 81–100%, respectively [34‒37]. The sensitivity and specificity of FNAC in minor salivary gland tumor diagnosis are 66.7% and 91%, respectively [38]. Considering acinic cell carcinoma and SC as low to intermediate grade malignancy, 50% of SC were correctly graded as low-grade malignancy in a retrospective single-center study [8]. However, the accuracy of cytology in diagnosing SC “specifically” has not been reported till date. In the present review, the sensitivity of FNAC in diagnosing SC of salivary gland was found to be 27.7%, which is quite low. This is because the cytological diagnosis of SC is not straightforward. In majority of cases, the cytological features were retrospectively looked upon after a final histological diagnosis with or without supportive molecular diagnosis was given. In a survey on 109 cytologists, it was found that a case of SC was correctly diagnosed by only 2 cytologists. The tumor was classified as Papanicolaou category I/II, III, and IV/V by 14%, 53%, and 33% of participating cytologists [24]. The Milan System for reporting cytopathology is a six-tiered reporting system with implication at predicting risk of malignancy. Wiles et al. [39] observed MRSGC diagnostic category for SC as malignant in majority of FNA specimens (50%), suspicious of malignancy (26%), salivary gland neoplasm of uncertain malignant potential (18%), and atypia of undetermined significance in 6% cases.

High-grade transformation has been reported in SC [40]. In addition to above mentioned typical cytological features of SC, presence of necrosis, mitosis, small cells with hyperchromatic nuclei with coarsely granular chromatin and distinct nucleoli can suggest high-grade transformation. Ki67 index assessment on cell block preparations can be helpful in such cases [23].

SC can be misdiagnosed as pleomorphic adenoma, low-grade adenocarcinoma, low-grade mucoepidermoid carcinoma, acinic cell carcinoma, low-grade salivary duct carcinoma, and oncocytic neoplasm (Table 5). An increased number of passes can give sufficient material to hint on the correct diagnosis [39].

Mucoepidermoid carcinoma (MEC), especially low-grade and SC share features of vacuolated cytoplasm of tumor cells, presence of mucinophages and mucinous extracellular material (Table 5). However, MEC smears are hypocellular and have a variable admixture of 3 cell types-mucinous, intermediate, and squamoid cells. On the other hand, in SC, the smears are cellular and the tumor cells are uniform and have eosinophilic granular to vacuolated cytoplasm. The mucinous cells of MEC are often univacuolated, while the tumor cells of SC often have multivacuolated cytoplasm [42]. Also, nuclear membrane irregularity/nuclear grooves may be observed in SC, this feature is not seen in MEC [26]. Mucicarmine stains the mucin in the background and the cytoplasm of mucinous cells in MEC. The extracellular and intracellular material seen in SC is positive for mucicarmine [43]. However, the intracellular/extracellular staining is focal and weak in case of SC [25].

SC is often misdiagnosed as acinic cell carcinoma, as observed in most of the case reports/studies in this review (37.7%). Acinic cell carcinoma and SC share common features of high cellularity, tumor cells with granular or vacuolated cytoplasm and low-grade nuclei, arranged in loose clusters, and presence of arborizing blood vessels. Besides, the cellular smears of SC also show papillary arrangement of cells along with the presence of arborizing blood vessels (Fig. 5). Hobnail cells covering the papillae have also been described in cytology of SC [27]. The cells of acinic cell carcinoma have coarser chromatin, prominent nucleoli, and intracytoplasmic zymogen granules, which are basophilic and fine in appearance on Giemsa-stained smears [25]. Stripped nuclei are a common finding in acinic cell carcinoma. For long period, the SC has been considered as zymogen-poor ACC. Chiosea et al. [32] reclassified 11 of 26 zymogen-poor acinic cell carcinoma as SC. In SC, the intracellular PAS-positive diastase resistant substance is often seen as larger, globoid, or more irregular granules or inclusions (Fig. 5). This is in contrast to the more uniform and finer PAS-diastase positive zymogen granules seen in acinic cell carcinoma [43].

In addition, intracytoplasmic and extracellular secretions may be demonstrated by mucicarmine. The intracytoplasmic vacuoles can also show positivity with adipophilin immunostain. These findings indicate that these secretions are either glycogen, mucin or lipid and may not be only degenerative changes [22, 41, 44].

The Warthin tumor has a classical cytomorphology comprising sheets of oncocytes, presence of lymphocytes in background and the granular background. The two tumors can share common features of cells with granular eosinophilic cytoplasm and presence of mucinous material in the background. However, the tumor cells in SC also have vacuolated cytoplasm, along with presence of histiocytes and lack of lymphocytes in background [33, 41, 44].

Various ancillary studies can be performed on the cytology specimen to increase the diagnostic yield. FNAC material can be useful in assessing cytomorphology, special cytochemical stain, immunocytochemistry, and FISH on destained smears. Oza et al. [27] successfully diagnosed SC preoperatively on cytology, by additionally using immunocytochemistry on destained wet-fixed cytology smears. A panel of markers S100, mammaglobin, and p63 can increase the sensitivity of FNA. Liquid-based cytology may further improve the diagnostic utility of FNAC and immunocytochemistry as in other salivary gland neoplasms [45].

In a study by Mahendru et al. [46] FISH was performed on the destained cytology smears of histopathologically diagnosed SC. They found FISH interpretable in 93.3% cases (14 of 15 cases) and confirmed ETV6 gene rearrangement in 50% cases (7 of 14 cases). In the remaining 7 cases, NR4A3 rearrangement was tested and 21.4% (3 of 14 cases) cases were found positive, suggesting the diagnosis of Acinic cell carcinoma. In total, a confirmed diagnosis was offered in 66.7% cases (10 of 15 cases). The results correlated completely with those performed on histopathological specimens [46]. From this study, it is also worth inferring that histopathology may not be confirmatory of SC and a molecular confirmation is essential. FISH has been successfully performed on the cytology smears (unstained or destained) in salivary gland neoplasms, with a sensitivity of 66.7% and specificity of 100%. Combined sensitivity of cytology and FISH in salivary gland neoplasms have been reported to be 93.3% [47].

Further, the aspirated material can be utilized to prepare cell-block, get hints about histomorphology, and proceed with histochemistry, immunohistochemistry, and molecular investigations (Table 4). However, insufficient material can be a limitation of the cell-block. Histologically, these tumors can have tubules, microcysts, papillae, or macrocysts. Eosinophilic colloid-like secretions are present in the macrocysts/microcysts. The tumor cells have granular eosinophilic cytoplasm or apocrine-appearance. These can have vacuolated cytoplasm, giving a clear cell or signet-ring cell appearance. The nuclei are round to oval with minimal pleomorphism, open chromatin, single indistinct to prominent nucleolus. Mitotic count is usually low. Necrosis is absent [4, 48]. Fine-needle aspiration-related changes can be found.

Immunohistochemistry is an invaluable tool, useful in arriving at a diagnosis due to characteristic expression of various immunomarkers (Table 5). SC is positive for cytokeratin 7, cytokeratin 18, S-100, mammaglobin, GATA-3, AE1/AE3, Vimentin, MUC1, MUC4, EMA, and pSTAT5 are positive immunohistochemical markers in SC. S100 immunostaining is generally intense but heterogenous or even negative staining can also be encountered [16]. Negative markers include SMA, calponin, p63, CD10, DOG-1, GCDFP-15, Her2Neu, estrogen, receptors, progesterone, and androgen receptor. P63 may be focally positive in SC, but calponin and SMA are almost always negative in SC [4, 49]. Ki 67 proliferation index is usually low in most of the cases but can be 30–40% in an occasional case [4, 5, 39]. Phosphorylated signal transducer and activator of transcription-ion 5 (pSTAT5a) has been successfully demonstrated in immunocytochemistry specimens of SC of salivary gland and is not expressed in any other salivary gland cancers [44]. A new immunohistochemical marker, Pan-TRK has been reported to be a cost-effective substitute for the FISH/RT-PCR molecular investigations [29].

Acinic cell carcinoma is mammaglobin negative, 1/3rd cases may express S100, express DOG-1(diffusely, in a canalicular pattern), HER2 expression may be present in an occasional case [50]. Co-expression of mammaglobin and S100 can also be seen in polymorphous adenocarcinoma and adenoid cystic carcinoma [51]. High-grade salivary duct carcinoma frequently express mammaglobin, HER2 overexpression, lack expression of DOG-1. Low-grade salivary duct carcinoma expresses S100 and mammaglobin. None of these tumors, except SC, show ETV6 rearrangement [50].

Molecular studies for ETV6-NTRK3 fusion due to underlying t(12;15) (p13;q25) mutation are considered diagnostic of the tumor. ETS variant transcription factor 6 (ETV6) is located on chromosome 12 and Neurotrophic tyrosine kinase receptor type 3 (NTRK3) is located on chromosome 15 [7, 9]. The fusion gene encodes for a chimeric tyrosine kinase, also activates RAS-MAP kinase and PI3K-AKT pathways [28]. Hence, it plays a role in oncogenesis and carries a risk of high-grade transformation [2]. This fusion gene can be detected by FISH, RT-PCR, or NGS. Other neoplasms with this characteristic translocation include secretory breast carcinoma, congenital fibrosarcoma, congenital mesoblastic nephroma, ALK negative inflammatory myofibroblastic tumor, a subset of gastrointestinal tumor, and acute myeloid leukemia [11, 52]. Recently, ETV6-RET and ETV6-MET fusions have also been described [52, 53]. These mutations are not observed in other salivary gland neoplasms [54]. On the other hand, Acinic cell carcinoma can have genetic alterations or deletions at chromosomes 4p, 5q, 6p, 6q, and 17p [2]. MECT1/MAML2 gene fusion is found in 35–65% cases of mucoepidermoid carcinoma, PLAG1 rearrangement is found in pleomorphic adenoma and MYB rearrangement or MYB-NFIB fusion is found in adenoid cystic carcinoma.

Thus, if a preoperative diagnosis of SC is warranted, a cell block prepared from the fine needle aspiration material can be utilized for immunohistochemistry and molecular diagnosis. In appropriate context, a short panel of three IHC markers S-100, mammaglobin and p63/SMA can help exclude most of the close differentials of SC. FISH analysis has been successfully reported in previous studies. The utility of cell block in molecular diagnosis is limited in some cases by inadequate specimen or degenerative changes in the specimen [16, 23, 55].

Surgical excision with free resection margins with/without neck dissection is the primary treatment for SC. The rate of lymph node metastasis has been reported to be 17.6% in SC as compared to 7.9% in acinic cell carcinoma [32]. Local recurrence rate is low (∼4%) and can occur after 50 months of primary treatment [5]. Some studies have mentioned the role of post-operative radiotherapy in patients of salivary gland carcinoma with incomplete resection, close margin, T3/T4 stage tumors, and regional lymph node metastasis [5, 56]. The 5-year and 10-year overall survival rates were 95% [5]. Even the recurrent cases and those with residual tumor have an indolent long-term clinical course [57]. Rarely, death due to distant metastasis, high-grade transformation, or multiple recurrences in T3/T4 disease can occur [58]. Acinic cell carcinoma has 5-year and 10-year survival rates of 91% and 88%, respectively [2]. Tropomyosin receptor kinase (TRK) inhibitors, e.g., entrectinib and crizotinib, have been tested in the SC patients, but the efficacy is still questionable. Acquired resistance has been reported after the initial response [17, 59, 60]. Nevertheless, TRK inhibitors or c-Met inhibitors are potential chemotherapeutic agents that can be considered in high-grade transformed SC, locoregional recurrence, and distant metastasis [61].

Currently, the efficacy of cytology in diagnosing SC of salivary gland is low. But, if FNAC is used diligently, a pre-operative diagnosis of SC is possible and can be confirmed by special stains and molecular investigations on the destained cytosmears and/or cell block preparation.

Ethical approval was not required for the study as this was a systematic review of the published articles and metanalysis was performed on the secondary data. Hence, patient’s informed consent was also not applicable for the review.

The authors declare that there are no conflicts of interest involved.

None.

P.S.K. and M.G. did the data search and compilation. Quality of studies was assessed by P.S.K. and M.G. Discrepancies were resolved and data was verified by N.T. The data were analyzed by P.S.K. with the help of statistician. The final version of the manuscript was approved by all three authors.

All data generated or analyzed during this study has been included in this article. Further inquiries can be directed to Dr Pooja Sharma Kala ([email protected]).