Page 53 - Screening for Cervical Cancer: Systematic Evidence Review
P. 53

Chapter III.  Results



               Computerized Rescreening



                               
                       AutoPap  is a computerized quality control system that can be used for initial screening
               and for optimizing selection of high-risk slides for rescreening.  Its current routine use in the


               United States is restricted to rescreening.  We found no studies that met strict criteria for review,

               although we did identify two studies by the same group that employed a level of histologic


               verification of test performance.  The first was based on an 86-slide set of smears from

                                            58
               histology-proven HSIL cases.   This study retrospectively measured whether the system “found”

               these slides when they were reviewed as part of a larger study.  The system can be set to

               determine the proportion of slides that will be rescreened; this study evaluated 10-percent and


               20%- rescreening thresholds.  At the 10%- review threshold, AutoPap selected 77% of the HSIL

               slides for assessment; at the 20%- threshold, it selected 86% for assessment.  These figures are

               substantially higher than the expected maximum selection rate of 10% of false negatives if the


               standard random sample of 10% of negative slides were to be used to determine rescreening.

                                                                                       
                       In a subsequent study, this group focused on ability of the AutoPap  system to aid

                                            59
               identification of HSIL lesions.   This cytologic diagnosis should consistently be associated with
               clinical evaluation and histology, which improved their ability to confirm positive diagnoses


               although again they have no verification of the test-negative specimens.  Within a parent study

               with 25,124 screened slides, they identified 70 slides supported by definitive diagnoses of HSIL,


               CIS, or invasive cancer.  They compared the detection rate for these slides in the convention

                                                 
               practice screening and the AutoPap  screening portion of the parent study.  Both modalities

                                                                                                      
               detected 63 of 70 abnormal slides; conventional practice failed to detect five and AutoPap

               failed to detect two.  The performance of the two modalities was statistically equivalent.







                                                             53
   48   49   50   51   52   53   54   55   56   57   58