Atypical glandular cells of undetermined significance (AGUS): Interobserver reproducibility in cervical smears and corresponding thin-layer preparations.

Five panelists independently reviewed 135 consecutive conventional cervical smears (CPs) originally classified as atypical glandular cells of undetermined significance (AGUS). A thin-layer slide (TP), prepared from the residual material, also was reviewed in each case. All patients underwent colposcopy that yielded at least 1 histologic specimen. Three or more of 5 reviewers retained the AGUS interpretation for 29% of CPs and 12% of the corresponding TPs. Interobserver variability infrequency of use of AGUS was marked, and interobserver agreement was poor. Agreement was improved for cases cytologically interpreted as a high-grade lesion, especially in TPs. Four of 5 reviewers retained the AGUS classification in CPs for all 7 biopsy-proven neoplastic glandular lesions. Of 95 CP interpretations made by 5 reviewers in the 19 histologically diagnosed high-grade lesions, 8 were "negative/reactive" and 6 were AGUS "favor reactive." AGUS is a poorly reproducible cytologic interpretation. Although most neoplastic glandular lesions may be distinguished by cytopathologists experienced in this area from mimics originally considered AGUS, attempts to increase the diagnostic specificity of AGUS may diminish sensitivity for an underlying high-grade precursor. Interobserver agreement was better for TPs than for the corresponding CPs. However, the split-sample TP slides may not have been fully comparable to the CPs.