Multicenter Comparison of Manual and Automated Screening of AutoCyte Gynecologic Preparations

Abstract

To compare AutoCyte SCREEN-assisted evaluation of AutoCyte PREP liquid-based preparations with manual microscopic screening of the same preparations in a masked, multisite trial. AutoCyte PREPs were made using the CytoRich automated, liquidbased method from the residual cellular material on the collection device after a conventional cervical smear had been made. The study involved 1,676 samples collected sequentially from high-risk patients at two medical centers. The AutoCyte PREPs were then screened manually by cytotechnologists at one of two laboratory sites. All abnormal slides were reviewed by the site pathologists for final diagnosis. The PREPs were then remasked and screened using the AutoCyte SCREEN automated, interactive screening system, designed to select potentially abnormal slides for manual review while allowing the direct sign-out of negative slides. The AutoCyte SCREEN-assisted practice result was determined by combining the interactive SCREEN result with manual evaluation for those cases selected by SCREEN for manual review. All slides deemed abnormal were manually reviewed by an independent reference pathologist. The original manual review results were then compared to the AutoCyte SCREEN-assisted practice results stratified by the Bethesda categories of abnormal diagnoses as determined by the reference pathologist. Of the 1,676 cases, 494 were determined to be abnormal (ASCUS+) by one or both of the study methods and also by the independent reference pathologist. Of these 494 abnormal cases, 312 had a reference diagnosis of LSIL+, and 139 had a reference diagnosis of HSIL or cancer. The remainder of these cases were ASCUS or AGUS. Sensitivities and false negative proportions were stratified by the reference pathologist based on Bethesda categories as "truth" and compared. For LSIL+ cases, manual screening alone had a sensitivity of 89% as compared to 98% for the AutoCyte SCREEN-assisted practice. Manual screening demonstrated 90% sensitivity to HSIL or greater abnormality as compared to 99% sensitivity by the AutoCyte SCREEN-assisted practice. There was a concurrent significant reduction in the false negative fraction using the AutoCyte SCREEN as part of screening practice. Specificity for both screening practices was equivalent.