Glandular abnormalities on cervical smear: a study to compare the accuracy of cytological diagnosis with underlying pathology.

The interpretation of glandular abnormalities detected by cervical smear provides a diagnostic dilemma. This study aims to compare the accuracy of cytological diagnosis with underlying pathology so that guidelines for the investigation and management of abnormal glandular smears may be formulated. A retrospective review of 150 women with glandular abnormalities reported on cervical smear collected over 12 months from 1996 in a University hospital was performed. Smears were graded by the initial report into 3 groups, dependent on the severity of abnormality. Investigation, treatment and subsequent 3-year follow-up were recorded. The accuracy of prediction for a significant neoplastic or preneoplastic glandular pathology only was 0% with mild, 9% (3/35) with moderate, and 24% (9/38) with severe abnormalities. When squamous lesions were included, the chance of finding any dysplastic squamous or glandular abnormality was 16% (12/77), 51% (18/35) and 82% (31/38), respectively, following a smear showing a suspected glandular abnormality only. Our results highlight the poor specificity of predicting glandular neoplasia or preneoplasia from cervical smears, with a final diagnosis of high grade CIN in 35% (17/49) of patients with dyskaryotic glandular cytological changes only and 83% (20/24) where concomitant squamous dyskaryosis was reported. The reporting of reactive or minor changes in endocervical cells was of no diagnostic value. Management protocols for moderate and severe glandular abnormalities should include visualisation and biopsy of the uterine cavity to exclude endometrial neoplasia.