Evaluation of PAPNET testing as an ancillary tool to clarify the status of the "atypical" cervical smear

To assess the utility of the PAPNET system (Neuromedical Systems Inc., Suffern, NY) in clarifying the status of cervical smears showing borderline abnormalities, we analyzed the results of five cytotechnologists who reclassified 200 "atypical" smears by evaluating PAPNET images only. The interobserver agreement (reliability) of the PAPNET reviewers was computed, and their readings were compared with three standards: the consensus diagnosis of five pathologists who used light microscopy, the detection of cancer-associated human papillomavirus DNA by Southern analysis, and the correlation with diagnoses of biopsy specimens obtained during passive follow-up. The PAPNET reviewers classified 18 to 65% of cases as normal, 25 to 42% as equivocal, and 10 to 55% as abnormal. Unanimous interobserver agreement was achieved in only 24 (13%) cases. Four of the five PAPNET reviewers agreed moderately well with the results of the pathology reference panel. In four of the five PAPNET reviews, classification of cases as abnormal was strongly correlated with the detection of cancer-associated types of human papillomavirus. Consensus PAPNET results of abnormal were predictive of abnormal histologic findings at follow-up. Theoretically, if colposcopy had been performed on all of the women with equivocal or abnormal PAPNET results (based on the consensus of the panel), as much as 95% of biopsy-confirmed lesions could have been detected, but 79% of women would have been referred. Restriction of colposcopy referral to women with definitely abnormal PAPNET readings would have reduced referrals to 31%, but the sensitivity of the triage would have dropped to 51% of biopsy-confirmed lesions.