Page 61 - Screening for Cervical Cancer: Systematic Evidence Review
P. 61
Chapter III. Results
These Hybrid Capture II estimates include an easily delineated group of women with
HIV infection who had high prevalence of HSIL 17.3% among the HIV-infected women in
63
the Womack and colleagues cohort. Excluding HIV-infected women from the aggregate test
performance estimates should have the largest influence on the predictive values, because these
are determined by prevalence. However, these women make up a small portion of the overall
total and the estimates of performance characteristics are essentially unchanged. To further
describe Hybrid Capture II performance in low-prevalence populations, we calculated aggregate
63
data for only those studies 21,62,64 or arms of studies in which the prevalence of HSIL was 10%
or lower. In these conditions, sensitivity is somewhat lower at 81.6%; specificity is similar at
78.2%; positive predictive value is reduced to 18.2% and negative predictive value is 98.6%,
essentially unchanged compared to the estimates that include all studies using Hybrid Capture II
regardless of prevalence.
HPV Screening for Detection of Low-Grade Cervical Changes
Four of the screening studies include data that allow estimation of test performance if
LSIL is included among the lesions targeted for detection by HPV testing. The sensitivity of a
positive Hybrid Capture II test for high-risk HPV types for detecting histology-proven LSIL or
greater ranged from 45.2% to 85.9%. 21,62,64 Specificities were between 47.0% and 96.7%. 21,62,64
Details of test performance for detecting low-grade and higher lesions are provided in Table 9.
Cuzick and colleagues study of women having screening at their primary care providers
offices in the United Kingdom is likely the most comparable to screening in the United States,
the relative performance of the SHARP (Digene Corp.) microtiter format PCR test and the
Hybrid Capture II is of interest. Both tests delivered similar performance: sensitivity of 64.3%
61
