Cell Blocks and the Milan System for Reporting Salivary Gland Cytopathology: A Meta‐Analysis

ABSTRACT

Background

The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) is a classification system first released in 2018. It provides a standardized categorization for salivary gland lesions. Cell blocks can be prepared from fine‐needle aspirations (FNAs) and can be utilized for the diagnostics of salivary gland lesions. This study aims to evaluate the distribution of the MSRSGC categories and the associated risks of malignancy (ROM) in salivary gland FNAs incorporating cell block preparations.

Methods

A PubMed literature search was conducted between March 3 and May 20, 2025, to identify studies with both cell block and MSRSGC applications in salivary gland FNAs. Data on sample size and the ROM for individual MSRSGC categories were extracted into Microsoft Excel and subsequently analyzed using OpenMetaAnalyst.

Results

Five studies were identified with a total of 1132 FNA samples processed to cell blocks. The pooled distributions of samples in the present meta‐analysis were 21.9% in non‐diagnostic (ND), 6.4% in non‐neoplastic (NN), 8.2% in atypia of undetermined significance (AUS), 30.0% in benign neoplasm (BN), 8.0% in salivary gland neoplasm of uncertain malignant potential (SUMP), 1.9% in suspicious for malignancy (SM), and 19.4% in malignant (MN). The pooled ROMs from this meta‐analysis was 25.3% in ND, 16.5% in NN, 26.3% in AUS, 2.0% in BN, 28.7% in SUMP, 84.8% in SM, and 97.7% in MN.

Conclusion

The ROMs of the AUS and SUMP categories were lower than the compared meta‐analyses. The cell block application utilizes architecture and ancillary studies, leading to better performance in indeterminate categories.

1. Introduction

The Milan System for Reporting Salivary Gland Cytopathology (MSRSGC) is a classification system used for fine‐needle aspiration (FNA) cytology samples from salivary glands. The MSRSGC was first published in 2018 and provides a standardized categorization for professionals working with salivary gland lesions [1]. The goal of the MSRSGC is to improve communication between pathologists and clinicians and, thus, overall patient management. The MSRSGC has six diagnostic categories: (1) non‐diagnostic (ND), (2) non‐neoplastic (NN), (3) atypia of undetermined significance (AUS), (4A) benign neoplasm (BN), (4B) salivary gland neoplasm of uncertain malignant potential (SUMP), (5) suspicious for malignancy (SM), and (6) malignant (MN).

The risk of malignancy (ROM) has been determined for each MSRSGC category based on the published studies and refined in the second edition of the MSRSGC [2]. The ROMs are calculated from cases that have undergone surgical excision and can, therefore, represent an overestimation. Some studies also report the overall ROM, which includes all cases within one category. This represents the lower range of the ROM, and the ROM from surgical excision, the upper range of the ROM [3]. ROM can differ between institutions due to patient demographics, health institution level and specialization, and differences in pathological and radiological interpretations [4].

Cell blocks can be made using various methods and are used to improve diagnostic outcomes in cytological samples. The most used methods are the plasma–thrombin method, the agar method, and commercial cell block kits [5, 6]. Ancillary studies can be performed on cell blocks. These include immunocytochemistry, special stains, and molecular techniques. Cell blocks can also be used to preserve cytological material for future studies [7].

Choosing a cell block method depends on institutional preferences, the specimens' cellularity and distribution pattern, and whether the specimen is fixed [8]. Although cell blocks play an important role in cytopathology, they do not always yield preferred results. Causes for failure include inadequate cellularity and inconsistent immunocytochemical staining from non‐formalin fixed cell blocks [9]. Other disadvantages limiting cell block use are operating costs, long processing time, and the need for excess material to produce a sufficient cell block [7, 10].

The goal of the present systematic review is to evaluate the distribution of MSRSGC categories and corresponding ROMs in salivary gland FNAs incorporating cell block preparations and to explore whether these findings differ from those reported in the broader MSRSGC literature.

2. Materials and Methods

A PubMed search was conducted between March 3 and May 20, 2025. The search terms used were “salivary gland” AND “cell block.” Additionally, one study from personal archives was inspected. The inclusion criteria were that the studies should use cell blocks in diagnosing salivary gland lesions and utilize the MSRSGC for categorizing the samples. Case studies and studies that did not address the subject were excluded. The selection process was performed according to PRISMA guidelines [11] and is illustrated as a flow chart (Figure 1).

FIGURE 1.

A flow chart displaying the study selection process according to PRISMA guidelines.

Relevant data from studies were collected using a Microsoft Excel spreadsheet (Microsoft 365 MSO, Version 2507, Build 16.0.19029.20136, 64‐bit). The following information was collected: the first author, year of publication, country of origin and continent, type of institution, period and time of study, used cell block method and cytologic stain, and whether the study was retrospective or prospective. The total number of samples and number of cell block samples from salivary glands, amount and percentage of samples per MSRSGC category, and amount and percentage of ROM per MSRSGC category were also collected.

Statistical analyses were performed using OpenMetaAnalyst (Brown University, Providence, RI, USA). A meta‐analysis of pooled proportions was conducted using a random‐effects model to account for anticipated clinical and methodological heterogeneity across the included studies. Between‐study variances were estimated using the DerSimonian–Laird method. Pooled estimates were reported with corresponding 95% confidence intervals (95% CIs). A continuity correction of 0.5 was applied in studies with zero events to allow their inclusion in the analysis. Statistical heterogeneity was assessed using Cochran's Q test and quantified by the I ² (inconsistency) index, with higher I ² values indicating greater heterogeneity. A two‐sided p‐value < 0.05 was considered statistically significant. The results are presented graphically using forest plots, showing individual study estimates with 95% CIs and the overall pooled estimate derived from the random‐effects model.

3. Results

3.1. Summary of the Studies

Five studies applying cell blocks in salivary gland diagnostics met the selection criteria and were analyzed in this meta‐analysis [12, 13, 14, 15, 16]. The studies were published between June 7, 2020 and October 20, 2023, and the combined study duration across all the studies was 33 years and 7 months. A total of 1132 FNA samples were collected from salivary glands, all of which were processed into cell blocks.

Cytological stainings varied among studies. Papanicolaou stain was used in two studies [12, 13], and May–Grunwald–Giemsa and hematoxylin and eosin stains were used in another study [15]. One study analyzed samples from multiple institutions, and multiple stains were used: Diff–Quik and Papanicolaou [13]. Two studies did not report any staining information [14, 16].

The cell block methods reported were the Cellient method [14], HistoGel method [16], and in‐house method [15]. One study did not specify the cell block method [13]. One institution used multiple methods (plasma–thrombin method, collection of visible tissue fragments, Shandon method, and in‐house method) [12].

Three of the studies were retrospective [13, 14, 16], and two, prospective [12, 15]. Only one of the studies [13] was multi‐institutional, while the others represented a single institution. Two series originated from Europe [12, 14], two from North America [13, 16], and one from Asia [15]. Four studies were performed at university hospitals [13, 14, 15, 16] and one in a publicly owned tertiary‐level private laboratory [12] (Table S1).

3.2. The MSRSGC Categories

All five articles organized the diagnosed samples into categories according to the MSRSGC. Out of the 1132 samples analyzed, 329 (29.1%) belonged to the ND, 67 (5.9%) to NN, 113 (10.0%) to AUS, 352 (31.1%) to BN, 100 (8.8%) to SUMP, 24 (2.1%) to SM, and 147 (13.0%) to MN.

The pooled distribution of specimens in each MSRSGC category was 21.9% (95% CI: 6.0%–37.9%, I ² = 98.23%) in ND, 6.4% (95% CI: 3.7%–9.1%, I ² = 70.55%) in NN, 8.2% (95% CI: 3.3%–13.1%, I ² = 90.86%) in AUS, 30.0% (95% CI: 15.5%–44.5%, I ² = 97.13%) in BN, 8.0% (95% CI: 3.5%–12.6%, I ² = 90.35%) in SUMP, 1.9% (95% CI: 1.1%–2.7%, I ² = 0%) in SM, and 19.4% (95% CI: 9.8%–29.1%, I ² = 96.71%) in MN (Figures 2 and 3).

FIGURE 2.

Prevalence and risk of malignancy (ROM) associated with the Milan System for Reporting Salivary Gland Cytopathology Categories. AUS, atypia of undetermined significance; BN, benign neoplasm; MN, malignant; ND, non‐diagnostic; NN, non‐neoplastic; SM, suspicious for malignancy; SUMP, salivary gland neoplasm of uncertain malignant potential.

FIGURE 3.

Forest plots showing the pooled prevalence of samples in each category of the Milan System for Reporting Salivary Gland Cytopathology. (A) ND, (B) NN, (C) AUS, (D) BN, (E) SUMP, (F) SM, and (G) MN. AUS, atypia of undetermined significance; BN, benign neoplasm; CI, confidence interval; I ², inconsistency index; MN, malignant; ND, non‐diagnostic; NN, non‐neoplastic; SM, suspicious for malignancy; SUMP, salivary gland neoplasm of uncertain malignant potential. [Color figure can be viewed at wileyonlinelibrary.com]

3.3. ROM

The ROM based on samples applying cell blocks for each MSRSGC category was analyzed from four studies [12, 13, 14, 16]. The ROM was not available in one study [15]. Pooled ROMs in each MSRSCG category were 25.3% (95% CI 10.2%–40.5%, I ² = 72.53%) in ND, 16.5% (95% CI −0.1% to 34.0%, I ² = 0%) in NN, 26.3% (95% CI 2.0%–50.5%, I ² = 79.4%) in AUS, 2.0% (95% CI −0.4% to 4.5%, I ² = 32.19%) in BN, 28.7% (95% CI 18.8%–38.7%, I ² = 0%) in SUMP, 84.8% (95% CI 70.0%–99.6%, I ² = 0%) in SM, and 97.7% (95% CI 94.4%–101.1%, I ² = 0%) in MN (Figures 2 and 4).

FIGURE 4.

Forest plots showing risk of malignancy (ROM) for each category of the Milan System for Reporting Salivary Gland Cytopathology. (A) ND, (B) NN, (C) AUS, (D) BN, (E) SUMP, (F) SM, and (G) MN. AUS, atypia of undetermined significance; BN, benign neoplasm; CI, confidence interval; I ², inconsistency index; MN, malignant; ND, non‐diagnostic; NN, non‐neoplastic; SM, suspicious for malignancy; SUMP, salivary gland neoplasm of uncertain malignant potential. [Color figure can be viewed at wileyonlinelibrary.com]

4. Discussion

The percentages of samples in each MSRSGC category should be approximately 10% in ND, 10% in NN, 7% in AUS, 42% in BN, 11% in SUMP, 4% in SM, and 15% in MN, according to the second edition of the MSRSGC. The pooled distributions of samples in the present meta‐analysis were 21.9%, 6.4%, 8.2%, 30.0%, 8.0%, 1.9%, and 19.4%, respectively. The results for most categories corroborate the percentages suggested by MSRSGC.

Compared with the MSRSGC recommendations and other studies, the number of samples in the ND category was higher. It varied between 3.8% and 17.5% in previous meta‐analyses of MSRSGC [17, 18, 19, 20]. The high number of ND samples is not optimal. Due to an error in acquiring or preparing a sample, a new diagnostic sample must be obtained.

The second edition of MSRSGC reports the ROM for each category as follows: ND 15%, NN 11%, AUS 30%, BN < 3%, SUMP 35%, SM 83%, and MN 98%. The pooled ROMs from the present meta‐analysis were 25.3%, 16.5%, 26.3%, 2.0%, 28.7%, 84.8%, and 97.7%, respectively. The pooled ROMs reported in this study were consistent compared with other meta‐analyses of the MSRSGC [17, 18, 19, 20, 21]. In these studies, the ROM values varied from 10.0% to 26.6% in ND, 5.0%–16.5% in NN, 28.0%–38.6% in AUS, 2.0%–4.0% in BN, 31.2%–42.0% in SUMP, 58.0%–91.0% in SM, and 91.0%–99.0% in MN.

Only the results from the AUS and SUMP categories did not fall between the results from other studies. In other studies, the AUS ranged from 28.0% to 39.6% and the SUMP from 31.2% to 42.0%. The results from this study were 26.3% and 28.7%, respectively. Lower ROMs in the AUS and SUMP categories are preferred to ensure that samples are properly classified, and the overuse of indeterminate categories is prevented. However, lower ROMs can also be a result of institutional differences in case selection and management. Diagnosing lesions as SUMP leads to surgical resection and as AUS to repeated FNA or surgical resection [1, 2]. The mean times between the FNA diagnosis and the resection were 104.8 and 88.0 days in the indeterminate AUS and SUMP categories and 32.3 and 55.7 days in the SM and MN categories, respectfully [22, 23, 24]. Diagnosing a sample in an indeterminate category leads to a longer time interval before the treatment and can influence patient outcome. Showing lower rates of cases diagnosed with AUS and SUMP in the present meta‐analysis, cell blocks could have been utilized to apply ancillary techniques to achieve more precise classifications and diagnoses for the samples that would otherwise fall into the indeterminate categories. However, use of cell blocks is not necessarily the reason for the lower ROM in AUS and SUMP categories in this meta‐analysis. The result can also be explained with ROM being lower in NN and BN and higher in SM and MN thus seemingly lowering percentage of samples in the AUS and SUMP categories.

Results of our study have a very high heterogeneity, and ROM estimates should be interpreted with caution. Several analyses demonstrated substantial statistical heterogeneity despite the use of random‐effects models. This heterogeneity likely reflects both methodological and biological variability among the included studies caused by varying staining methods, ancillary studies and cell block methods between studies as there are no standardized cell block guidelines. Differences in study design, institutional selection bias, different case mixes, and retrospective versus prospective designs could also lead to increased heterogeneity in this research. Our study also had only a limited number of studies, which may have further increased statistical heterogeneity and limited the generalizability of some pooled estimates.

ROM calculated from surgically removed cases represents upper ROM and as such it is an overestimation. Lower ROM includes all cases as denominators regardless of surgical follow‐up, assuming unresected cases as benign and as such leads to underestimation [25]. The actual ROM falls somewhere between these two values [20]. If only surgically followed cases contribute to ROM calculations, it could affect the ROMs presented in our study.

The main limitation of this study is its small sample size. Only five studies were analyzed. In addition, one study did not use cell blocks to determine ROM and was excluded from the analysis of pooled ROMs. Future research involving studies with larger sample sizes is needed to confirm these findings. There is also a large variety of different cell block methods with advantages and disadvantages. The optimal method for salivary gland diagnostics should be studied.

Ancillary testing may help to resolve diagnostic uncertainty. It is used to differentiate between benign and malignant neoplasms or low‐grade and high‐grade malignancies. Ancillary testing can also be used to improve diagnostics of samples that would be sorted out to AUS or SM categories [26]. Ancillary testing varies between samples and use of ancillary testing could have affected the results of our research.

This study shows that the use of cell blocks could possibly help in diagnostic performance of the MSRSGC. However, more research with controlled cell block methods and stainings in addition to comparing FNAs with and without cell blocks is needed to verify the present findings. If the findings of this study were proven to be true, use of cell blocks could be increased in salivary gland diagnostics leading to better results in the salivary gland diagnostics and consequently in patients' outcomes.

5. Conclusion

This meta‐analysis indicates that cell block preparations may enhance diagnostic stratification in salivary gland FNA cytology when applied within the Milan System. The lower pooled ROMs observed in the AUS and SUMP categories suggest that cell block preparations may contribute to improved diagnostic stratification through architectural assessment and ancillary studies. However, the currently available evidence does not allow direct demonstration of improved diagnostic accuracy, and further prospective studies evaluating diagnostic reclassification and performance metrics are warranted.

Author Contributions

Aada Närhi: conceptualization, methodology, data curation, formal analysis, visualization, and writing – original draft. David Kalfert: methodology, validation, formal analysis, investigation, data curation, visualization, and writing – review and editing. Henri Lagerstam: data curation and writing – review and editing. Ivana Kholová: conceptualization, investigation, data curation, writing – original draft, writing – review and editing, supervision, and project administration.

Funding

This study was funded by a VTR Research Grant X5211 from Pirkanmaa Wellbeing Service County (to I.K.) and the Cooperation Program, research area SURG (to D.K.).

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Summary of included studies and study series characteristics.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.