ABSTRACT
BACKGROUND: The AutoPap System for the initial screening and quality control of conventional cervical cytology slides previously has shown superior performance for the detection of abnormal slides of low grade squamous intraepithelial lesions and above. This report presents data regarding the important category of high grade squamous intraepithelial lesions and above (HSIL+). METHODS: All slides were run through a Current Practice (CP) arm of manual screening with 10% random quality control followed by an AutoPap (AP) arm of device initial screening with approximately 25% of slides receiving no further review; 75% received a manual rescreen with AP ranking and QC rescreening of 15% of the top ranking negative manual screen slides was performed. Detection performance for truth-determined HSIL+ cases was compared between the two study arms. Available follow-up biopsy results were correlated for cases determined to be HSIL+. RESULTS: Of 25,124 analyzed slides, 70 slides had truth-determination at the HSIL+ level (67 HSILs, 1 adenocarcinoma in situ, and 2 invasive tumors). The AP arm identified 68 of 70 cases (including both invasive tumors). Neither false-negative HSIL+ was found in the 25% "No Further Review" population. The CP arm identified 65 of 70 cases (missing both invasive tumors). The results showed statistical equivalence between the two arms (P = 0.0129). Biopsy follow-up was available in 27 of 70 HSIL+ cases identified by AP and showed abnormality at some level in 100% of cases. CONCLUSIONS: The AP arm was statistically equivalent and showed numeric superiority to the CP arm for the category of HSIL+. Follow-up data confirmed the true-positive nature of these findings. These results are significant because any overall improvement in the performance of an initial screening device must be paralleled by at least equivalence in the clinically important subset category of HSIL+. Cancer (Cancer Cytopathol)
