Predictive value of diagnoses of endocervical glandular abnormalities in cervical smears.

To determine the positive predictive value (PPV) of cervical smear diagnoses of 'definite' and 'possible' endocervical adenocarcinoma in situ or invasive adenocarcinoma, and whether diagnostic accuracy can be improved.The study examined cervical smears reported as definite or possible high-grade glandular abnormality between 1992 and 1998. PPV was calculated by comparing smear diagnoses with the subsequent histopathology report. All available smears were reviewed without knowledge of follow-up results, and were reclassified by consensus.Thirty-two smears were diagnosed as high-grade glandular lesions, with adequate biopsy follow-up in 31 cases (96.9%). A high-grade epithelial abnormality (HGEA) was detected in 29 cases (PPV, 93.5%), with a high-grade glandular lesion in 24 (PPV, 77.4%). Very few smears were reclassified on review. Seventy-three smears were initially diagnosed in the 'inconclusive' glandular or indeterminate cell-type category. There was adequate biopsy follow up for 54 cases (74.0%). On follow-up, 31 cases had a HGEA (PPV, 57.4%), with 14 cases having a high-grade glandular abnormality (PPV, 25.9%). In the review of 'inconclusive' smears, 12 were reclassified as squamous abnormalities and none of these had a glandular lesion on biopsy. Eight were reclassified as negative; seven contained endometrial stroma and the glandular cells in question were considered to be of lower uterine segment (LUS) origin. No significant lesion was present on follow-up of these cases.For clinicians using our laboratory, large loop excision of the transformation zone (LLETZ) or cone biopsy should follow a 'definite' cytological diagnosis of a high-grade endocervical glandular lesion. However, cone biopsy may not be the appropriate initial management in the 'possible' high-grade glandular group because of a significantly lower predictive value of the diagnosis. The slide review highlighted the importance of (1) caution in classifying sheets of abnormal cells as glandular, and (2) endometrial stroma as a marker of LUS material.