Page 49 - Screening for Cervical Cancer: Systematic Evidence Review
P. 49
Chapter III. Results
conflicting results on the basis of all the data. Because the threshold for colposcopy was ASCUS
or higher, the majority of equivocal final case diagnoses are presumed to be equivocal low-grade
findings although the precise make-up of this group is not clearly specified in the article.
Bearing this classification system in mind, we have calculated performance measures for
ThinPrep using two different assumptions about equivocal final diagnosis; one assumption
includes them in normal, and the other in LSIL. Based on the findings of no abnormal histology
among the random sample of 150 women with normal final diagnoses (i.e., all screening test
negative) who were evaluated with colposcopy and biopsy, estimates of sensitivity, specificity
and predictive value are based on the presumption of no false negatives, which is plausible based
on the number of tests applied and the consensus process for assigning the final diagnoses.
Table 6 summarizes test characteristics at different thresholds; assignment of the equivocal
cases is indicated.
Two studies of ThinPrep , known for the large size of the populations screened (>35,000
each), did not meet strict criteria. These studies provide histology results for a subset of subjects
52,53
obtained within undisclosed periods of time from the screening ThinPrep . Neither used a
colposcopy/histology reference standard to verify test negatives, and both studies appear to rely
on histology specimens associated with clinical care for calculating sensitivity and specificity of
the ThinPrep test compared to conventional cytology.
Among 10,694 US patients in the latter study screened by ThinPrep , 630 women had
cytology diagnoses of ASCUS or more severe; 1,395 of 39,408 women screened by conventional
53
cytology had cytology of ASCUS or more severe. However, the estimates of sensitivity and
specificity reported in the publication are based on 54 biopsy reports in the ThinPrep group and
89 in the conventional cytology group. The authors do not specify how these available
49
