Hologic ThinPrep, Руководство оператора

Страница: 9 / 230

MAN-08723-001 Rev. 001 page 5 of 28
G.2 DESCRIPTIVE DIAGNOSIS SENSITIVITY AND SPECIFICITY ESTIMATES
A panel of three independent Cytopathologists adjudicated slides from all discordant (one-grade or
higher cytologic difference) descriptive diagnosis cases (639), all concordant positive cases (355)
and a random 5% subset of the 8550 negative concordant cases (428). The Cytopathologists on the
adjudication panel were board-certified, all of whom had a subspecialty certification in
Cytopathology. Their experience levels in Cytopathology ranged from 6 to 12 years. Two of the
adjudicators were from university practices and one adjudicator was from a private medical center.
The volumes for the adjudicator’s institutions ranged from 12,000 to 30,000 ThinPrep
® Pap Tests
annually.
A consensus diagnosis was defined as agreement by at least 2 of 3 Cytopathologists. All slides sent
to the panel of Cytopathologists were not identified by site nor ordered in any fashion. When a
consensus diagnosis could not be obtained by at least 2 of 3 Cytopathologists, the full panel of
Cytopathologists reviewed each case simultaneously using a multi-headed microscope to determine
a consensus diagnosis.
The adjudicated results were used as a “gold standard” to define the following major “true”
descriptive diagnosis classifications of the Bethesda System: Negative, ASCUS, AGUS, LSIL,
HSIL, Squamous Cell Carcinoma (SQ CA) and Glandular Cell Carcinoma (GL CA). Estimates of
sensitivity and specificity together with 95% confidence intervals were calculated for the
Manual
Review and Imager Review arms of the study. The differences in sensitivity and specificity between
the two arms, together with their 95% confidence intervals were also calculated. Among the random
5% subset of 8,550 cases (428 slides) that were found to be negative by both arms and adjudicated,
there were 425 “true” negative and 3 “true” ASCUS slides. A multiple imputation technique was
used to adjust the numbers of true positives and true negatives for the 8,550 negative concordant
cases based on the 5% of cases that were adjudicated
3 .
Tables 2-4 below summarize the descriptive diagnosis sensitivity and specificity estimates with 95%
confidence intervals for each of the four sites and all sites combined for “true” ASCUS+, LSIL+
and HSIL+.
Table 2: Adjudicated Review Versus Imager And Manual Reviews ASCUS+
Descriptive Diagnosis Summary.
Sensitivity is a percent of “true” ASCUS+ (combined ASCUS, AGUS, LSIL, HSIL, SQ CA and GL CA) slides
classified in either study arm as ASCUS+ and specificity is a percent of “true” Negative slides classified in either
study arm as Negative.
Sensitivity Specificity
Site/
Number
Cases
Manual Imager Difference
Site/
Number
Cases
Manual Imager Difference
Site 1
77.2% 78.3% +1.1%
Site 1
98.7% 99.2% +0.4%
180 (70.4, 83.1) (71.6, 84.1) (-5.8, 8.0) 2132 (98.1, 99.1) (98.7, 99.5) (-0.1, 1.0)
Site 2
63.1% 77.5% +14.4%
Site 2
95.8% 96.1% +0.3%
230
(56.5, 69.3) (71.4, 82.6) (8.2, 20.5)
2210
(94.9, 96.6) (95.2, 96.9) (-0.7, 1.3)
Site 3
80.6% 94.2% +13.6%
Site 3
98.5% 98.8% +0.4%
103 (71.6, 87.7) (87.8, 97.8) (4.3, 22.9) 2196 (97.9, 99.0) (98.3, 99.2) (-0.3, 1.0)
Site 4
87.2% 84.4% -2.8%
Site 4
97.3% 97.0% -0.3%
179 (81.4, 91.7) (78.2, 89.4) (-10.6, 5.0) 2313 (96.6, 97.9) (96.2, 97.7) (-1.1, 0.5)
All
75.6% 82.0% +6.4%
All
97.6% 97.8% +0.2%
692 (72.2, 78.8) (78.8, 84.8) (2.6, 10.0) 8851 (97.2, 97.9) (97.4, 98.1) (-0.2, 0.6)
Numbers in parentheses represent 95% confidence intervals.
The results presented in Table 2 show that for ASCUS+, the increase in sensitivity of the Imager
Review over the Manual Review was statistically significant with the lower limit of the 95%
confidence interval being 2.6% for all sites combined. The observed difference between sensitivities
for ASCUS+ varied among the sites from –2.8% with a 95% confidence interval of (– 10.6%; 5.0%)
to +14.4% with a 95% confidence interval of (8.2%; 20.5%). The difference in specificity results
between the
Imager Review and the Manual Review was not statistically significant with a 95%
confidence interval of -0.2% to +0.6%. The observed differences between specificities varied among
the sites from –0.3% to +0.4%.