Atypical glandular cells of undetermined significance. Diagnostic accuracy and interobserver variability using select cytologic criteria.

Abstract

Histologic follow-up of patients with the Bethesda system cervical-vaginal diagnosis of atypical glandular cells of undetermined significance (AGUS), "favor endocervical origin", or "not otherwise specified" (NOS) shows a high percentage of clinically significant (neoplastic or preneoplastic) lesions. Using the criteria of atypical single cells, irregular nuclear membranes, and decreased cytoplasm, eight observers retrospectively reclassified 88 AGUS, "favor endocervical", or NOS smears using a probabilistic scheme. Follow-up showed 46 clinically significant and 42 benign lesions. The mean accuracy for all observers and the experienced observers was 65% and 72%, respectively. For the experienced observers, the mean specificity of a "favor clinically significant" category was 72%; the mean sensitivity of a "favor benign" category was 90%. For the less experienced observers, subclassification had poor predictive value. We conclude that experienced observers may use specific criteria to correctly subclassify AGUS lesions, and this may aid in patient management.