Evaluation of the PAPNET System for prescreening triage of cervicovaginal smears

OBJECTIVE: To assess the PAPNET System for prescreening triage of cervical smears. STUDY DESIGN: We prospectively prescreened 5,170 consecutive cervicovaginal smears with the PAPNET System. The slides were then manually screened by cytotechnologists blinded to the PAPNET diagnoses. Cases identified as abnormal by either PAPNET or manual screening were reviewed by a cytopathologist. The PAPNET and manual diagnoses were correlated. RESULTS: Diagnostic concordance between PAPNET and manual screening was seen in 4,340 (84%) of the cases (3,167 negative, 1,038 abnormal and 135 unsatisfactory). Noncorrelation between PAPNET and manual diagnosis occurred in 794 cases (543 abnormal by PAPNET and negative manually, 228 negative by PAPNET and abnormal manually, 8 abnormal by PAPNET and unsatisfactory manually, 29 unsatisfactory by PAPNET and negative manually, 7 negative by PAPNET and unsatisfactory manually). The diagnostic sensitivity of the PAPNET System was 82%, diagnostic specificity 85%, predictive value of a positive test 66% and predictive value of a negative test 93.4%. The false negative fraction of PAPNET was 6.4% for low grade squamous intraepithelial lesion and above. CONCLUSION: PAPNET performed effectively for prescreening triage, increasing the accuracy of screening and reducing the screening time.