Subcategorization of atypical glandular cells is useful to identify lesion site

Background In the subcategorization of atypical glandular cells (AGCs), origin of cells should be mentioned to estimate lesion sites for diagnosis. However, cases without subcategorization are often encountered due to limited reproducibility. We evaluated whether the subcategorization of AGC based on the Bethesda terminology can estimate lesion sites. Methods We retrospectively investigated cases whose cervical smears were interpreted as AGC and underwent pathological assessment at our institution between June 2009 and September 2017. AGC was subcategorized based on the Bethesda System. Not-otherwise-specified (NOS) was subcategorized into endocervical cells (NOS-EC), endometrial cells (NOS-EM), or glandular cells (NOS-G). Favor neoplastic (FN) was subcategorized into endocervical cells (FN-EC) or glandular cells (FN-G). FN-G was further subcategorized into endometrial cells (FN-EM) or unknown origin (FN-UO). Clinicopathological data were retrieved from the medical records. Results Of 88 AGC cases, there were 30 NOS-EC (34.1%), 2 NOS-EM (2.3%), 25 FN-EC (28.4%), 22 FN-EM (25.0%), and 9 FN-UO (10.2%). A significantly higher proportion of neoplastic lesions occurred in FN than in NOS (P <.001). The concordance of AGC subclass and lesion site was 88.0%, 70.7%, and 77.3% in FN-EC, FN-G, and FN-EM, respectively. The concordance of FN-EM and lesion site increased to 88.9% in patients aged >50 years. Of nine cases of FN-UO, six experienced nonendometrioid endometrial cancer and extrauterine malignancy. Conclusion Subcategorization of NOS and FN would be useful in estimating neoplastic lesions. Further subcategorization into FN-EC, FN-EM, and FN-UO would similarly be beneficial in estimating the lesion site, especially for small endometrial and extrauterine lesions.