Accuracy of thin-layer cytology in patients undergoing cervical cone biopsy

OBJECTIVE: To compare the accuracy of thin-layer cytology with Autocyte PXEP (R) (TriPath Imaging Inc., Burlington, North Carolina, U.S.A.) with conventional smears in 500 women under-going cervical cone biopsy. STUDY DESIGN: The study was performed among 500 consecutive women presenting for cone biopsy for high grade cervical intraepithelial neoplasia (CIN) on biopsy in 350 (70%) and discrepant cytology/colpohistology in 150 (30%). Before performing a cone biopsy, two cervical samples were collected for conventional smears and thin-layer cytologic slides, with randomization of the order. Conventional smears were stained and diagnosed at Pasteur Cerba, while thin-layer cytologic slides were processed at a local TriPath office (Meylan, France) and sent in a masked fashion for screening at Pasteur Cerba. Any slides initially read as normal were reviewed again and reported without knowledge of the other cytologic oil cone biopsy data. The final cytologic diagnoses for the two methods were compared with histopathology of the cone biopsy. RESULTS: The conventional smear was unsatisfactory in 58 (11.6%) of cases, while there were 4 (0.8%) unsatisfactory thin-layer cytologic slides (P < .001). Endocervical cells were missing from 31 (6.2%) of conventional smears and 34 (6.8%) of thin-layer cytologic slides. For the pooled data, sensitivities of conventional smear and thin layer for detecting high grade CIN (0.82% and 0.86%, respectively) were similar as were specificities (0.40% and 0.43%, respectively). When first samples were compared, the sensitivities of the conventional smear and thin layer for high grade CIN were 0.79% and 0.89%, respectively (P = .02), with corresponding specificities of 0.41% and 0.36% (P < .01). CONCLUSION: When controlled for sample order, the sensitivity of thin-layer cytology for detecting high grade CIN teas significantly higher than that of conventional smears in patients with previous abnormal cytology, but at the expense of specificity.