In this retrospective study, the records of Seoul National University Hospital were searched within the period from January 1995 to December 2020 to identify all reported cervical cytology cases with glandular abnormalities. These cases were prepared either as LBC slides or conventional smears. Among the 683,197 Pap tests over this 25-year period, 985 cases from 923 patients were reported with a glandular abnormality, while the 657 of these with available surgical pathology follow-up were included in the study. Of the 985 Pap tests, 322 (32.7%) were prepared as conventional smears, whereas 663 (67.3%) as LBC. The cytologic interpretations regarding the glandular abnormalities were made by board-certified cytopathologists of our laboratory, based on the following categories of the Bethesda system [5,26]: atypical glandular cells not otherwise specified (AGC-NOS), atypical endocervical cells (AGC-EC), atypical endometrial cells (AGC-EM), atypical glandular cells favor neoplastic (AGC-FN), and ADC. Several Pap tests that were diagnosed as atypical glandular cells of undetermined significance, favor reactive based on the Bethesda system 1991, were considered AGC-NOS lesions in 2014 Bethesda system. Surgical pathology follow-up included any of the following samples: cervical biopsies, endocervical curettages, loop electrosurgical excision procedure or conizations, endometrial curettages, and hysterectomies.

A few cytomorphologic features of selected Pap test cases exhibiting glandular abnormalities and cytohistologic discrepant findings were also compared between the cases subsequently found to be histologically benign and malignant, to identify cytologic criteria associated with malignancy. These features included architectural characteristics (overlapping; presence of single cells; and presence of three-dimensional clusters), nuclear enlargement, irregular nuclear membranes, increased nuclear-to-cytoplasm (N/C) ratio, nuclear hyperchromasia, prominent nucleoli, and mitotic activity. The Fisher exact test was used to identify the importance of cytomorphologic features in the diagnosis of malignancies. A p-value of <.05 was considered statistically significant.

Although the efficacy of cervical cytology diagnosing squamous cell carcinoma and its precursors is well-established, detecting glandular abnormalities is a challenge due to their rarity, pathologists’ lack of experience with their cytomorphologic criteria, and poor interobserver reproducibility [10-12]. The aim of the present single-center study was to investigate the diagnostic distribution and assess the cytohistologic concordance of glandular abnormalities in cervical cytology, also to identify selected cytomorphologic findings associated with malignancy in follow-up surgical pathology.

According to our findings, glandular abnormalities accounted for 0.144% of all cervical cytology cases in our center from 1995 to 2020, showing their rarity and the diagnostic challenge interpreting them. This finding is consistent with what is reported in the literature [15,28,29]. In respect of the distribution of glandular lesions, none was diagnosed as AIS in cytology. Of the 16 cases that were histologically-confirmed to be cervical AIS, 13 were interpreted as AGC-NOS and three as ADC in cytology. This indicates that identifying AIS in cytology is a rather challenging task. Previous studies have identified the pitfalls of the AIS cytologic diagnosis, also its suboptimal accuracy and low levels of interobserver agreement [30-32]. Some authors have also reported that many AIS lesions could be interpreted as AGC, AGC-FN, ADC, or squamous intraepithelial lesions in cervical cytology [32,33].

Assessing the cytohistologic correlation of the glandular abnormalities detected in our cervical cytology cohort, we found that most Pap test cases reported to have glandular abnormalities cytologically actually had some significant lesions in their subsequent histopathology. One hundred eighty-eight out of 657 cases (28.6%) were turned out to be completely concordant, all of which were diagnosed as malignant tumors in follow-up histology (including cervical 92, endometrial, and ovarian 73). Also, there were 121 cases reported as AGC in cervical cytology, which turned out to be malignant tumors of various origins. On the other hand, 213 AGC Pap tests were histologically benign. Zhao et al. [25] analyzed 662 patients with an AGC cytologic interpretation and available tissue biopsy material and found that AGC cytology revealed cancer in 15.3% of the cases during histologic follow-up, most likely located in the cervix (8.3%), the endometrium (6.3%), and the ovaries (0.6%), respectively. In another study by Pradhan et al. [34], the histologic diagnoses of 3,709 AGC cases consisted of: negative (70.5%), cervical intraepithelial neoplasia/low grade squamous intraepithelial lesion (LSIL) and high grade squamous intraepithelial lesion (HSIL) (20.7%), endometrial ADC (5.5%), endocervical AIS and ADC (1.9%), and metastatic carcinomas (0.5%).

We also reviewed 74 AGC Pap tests to identify any cytomorphologic feature differences between the cases confirmed as benign and malignant during follow-up surgical pathology. In this examination, irregular nuclear membranes, three-dimensional clusters, single-cell pattern, and presence of mitoses were found to be significantly associated with malignancy in subsequent histology. On the contrary, no significant association was identified regarding increased N/C ratio, nuclear hyperchromasia, prominent nucleoli, or overlapping. These particular findings are largely concordant with what has so far been reported the literature. Raab et al. [35] showed that the presence of single dysplastic cells, nuclear membrane irregularities, and reduced amount of cytoplasm indicated cancer or a cancer precursor in histology. In another study by Yucel Polat et al. [29], the presence of feathering, papillary pattern, polarity loss, 3D clusters, irregular nuclear membranes, and prominent nucleoli were found to be significant; in contrast, the formation of rosettes, overlapping, increased N/C ratio, and nuclear hyperchromasia were not significantly associated with cancer in follow-up histology. Torres et al. [36] showed that finding cells exhibiting high N/C ratio and dyskeratosis indicated intraepithelial neoplasia or cancer, whereas glandular lesions were composed of cells with reduced amount of cytoplasm, nuclear membrane irregularity, and macronuleoli. Lastly, Reynolds et al. [37] reported that the presence of single dysplastic cells, 3D clusters, intracytoplasmic neutrophils, increased N/C ratio and larger nuclei, nuclear border irregularity, reniform-shaped nuclei, polarity loss, overlapping, and macronucleoli were significantly associated with a clinically significant lesion (HSIL or cancer) in surgical pathology follow-up.

Whenever possible, it is necessary to identify the cytomorphologic differences between squamous and glandular abnormalities in Pap tests, in addition to the cytologic features observed in benign glandular lesions that could mimic malignancy, to reduce potential misinterpretations. AGC is often found to be a benign lesion or a squamous intraepithelial lesion (e.g., a HSIL with endocervical gland involvement), rather than a glandular abnormality, in follow-up histology [34,38]. HSIL with endocervical gland involvement could indeed be misinterpreted as a glandular abnormality in cervical cytology. However, the loss of polarity within the hyperchromatic crowded groups, the flattening of the cells at the periphery of such groups, and the identification of single squamous dyskaryotic cells could help identify such lesions as squamous rather than glandular [26,38].

Some cases in our cohort included histologically benign cases interpreted as ADCs in cytology. Interestingly, among the cases interpreted as AGCs or ADCs in cytology, there were a few cases that were neither ADC nor squamous cell carcinoma in histology, including a smooth muscle tumor of uncertain malignant potential, an ovarian diffuse large B-cell lymphoma, an extramammary Paget disease of vulva, a small cell carcinoma, and a poorly differentiated carcinoma of cervix. The pathologists’ lack of experience and/or the presence of ambiguous cytomorphological criteria may be responsible for their misinterpretation.

A limitation of this study is that a few of our cases cytologically containing both glandular and squamous abnormalities (e.g., ACG-NOS and atypical squamous cells-cannot exclude HSIL (ASC-H) were solely classified based on their glandular component in this study, before assessing our cases’ distribution and cytology-histology correlation. According to the literature, glandular and squamous lesions in the cervix often coexist [39,40]. In our study, 50 of the 657 cases with histologic follow-up diagnosis were found to have both glandular and squamous abnormalities in their cytology reports. Forty-four of 409 AGC-NOS cases exhibited such coexistence, whereas 19 of them were finally diagnosed as cervical squamous neoplasias (LSIL, HSIL, or invasive squamous cell carcinoma) during surgical pathology follow-up, without any glandular lesions present. As previously mentioned, we evaluated only the glandular component for the cytohistologic correlation assessment of such cases. For example, a case interpreted as AGC-NOS and ASC-H in cytology followed by an LSIL diagnosis in histology, was regarded as a lesion with AGC-NOS only, ignoring its cytologic interpretation about squamous component. This way of classification, however, can be controversial, as it may not carry a major clinical impact when AGC patients end up having a squamous lesion.

In conclusion, this study showed the current status of glandular abnormalities detected by cervical cytology in a single center over a 25-year-period. They revealed a few cytomorphologic features, such as nuclear membrane irregularity, three-dimensional clusters, single-cell pattern and presence of mitoses, associated with malignancy in follow-up histology. As detecting glandular lesions is rare in Pap tests and pathologists are often unfamiliar with them, we hope this study will add some value to the relevant literature.