Page 51 - Screening for Cervical Cancer: Systematic Evidence Review
P. 51
Chapter III. Results
the same criteria outlined above, we found no studies that met the inclusion standards. One
study of 160 slides originally classified by conventional cytology as ASCUS and subsequently
55
scanned by the PapNet system is supported by histology for all samples. This study failed to
meet criteria because the reference standard was obtained up to one year after the screening test
and because the slides reviewed do not reflect a representative distribution of normal and
abnormal specimens. We summarize this work and two other studies with some use of histology
documentation as examples of the content of the literature.
The slides for this ASCUS review study, selected because the patients had subsequent
histology on record, were re-read using PapNet . Although the spectrum of disease is skewed
and reflects a high prevalence of abnormalities, performance measures can be calculated for
LSIL and HSIL given an initial ASCUS determination by conventional Pap. If histology of HPV
changes only is separated from CIN 1 and included with normals, the sensitivity, specificity,
positive predictive value, and negative predictive value of PapNet categorization of LSIL or
greater are 45.4%, 85.8%, 45.5%, and 85.8%, respectively. If HPV changes are included in CIN,
these values are, respectively, 37.7%, 92.3%, 78.8%, and 66.1%. At a PapNet threshold of
HSIL or greater for detecting CIN 2, CIN 3, or carcinoma, the sensitivity was 27.3%, specificity
was 94.2%, positive predictive value, 42.9%, and negative predictive value, 89.0%.
Jenny and colleagues hand-selected a slide set consisting of 516 abnormal slides with
accompanying histologic record of abnormality and mixed them with 684 slides from women
56
with 2 years of normal follow-up. The study was designed to evaluate the consistency of
reports between two independent PapNet evaluations and a manual screening of the same slide
set. The PapNet evaluations identified 91% and 84% of the histologically proven abnormal
specimens; conventional screening classified 78% as abnormal. The resulting true positive ratio
51
