Page 121 - Screening for Cervical Cancer: Systematic Evidence Review

Appendix C. Evidence Tables

Evidence Table 2. New Methods for Preparing or Evaluating Cervical Cytology (cont'd)

Quality
Patients & Methods Outcomes Measured Study Results & Limitations* Considerations†
AutoPap
86 cases of known Sensitivity of AutoPap In 86 biopsy–positive cases Quality Score=4
HSIL+ from archives of to detect known with cytological diagnosis of Ref. Std: 2
two labs HSIL+ HSIL+, AutoPap selected 66 Blind: 0
(77%) at 10% review fraction Verification: 1
GS: histology and 74 (86%) at 20% review Consecutive: 0
fraction Spectrum: 0
Publication: 1 Industry: 0
Limitation: not rescreening use

25,124 Pap smear Estimated sensitivity Est Se (ASCUS+) = Quality Score=6
slides, excluding “high- Relative TPR 1199/1397 (85.8%) Ref. Std: 0
risk” slides Relative FPR Est Se (LSIL+) = 321/348 Blind: 2
(92.2%) Verification: 1
GS: cytology Relative TPR (AutoPap/conv) Consecutive: 1
(independent panel = 1199/1106 = 1.08 Spectrum: 1
review of discrepant Relative FPR (AutoPap/conv) Publication: 1 Industry: 0
cases). = 1123/1322 = 0.85
No histological
verification. Limitation: results not reported
for performance of AutoPap
System on slides with manual
screen witin normal limits.
AutoPap system uses a
different algorithm from
AutoPap 300 QC

25,124 slides, excluding Pap results: Bethesda Prevalence: Quality Score=
“high risk” slides method HSIL+ = 70/25,124=0.28% 5
Ref. Std:0
Reference standard: AutoPap: Blind:2
HGSIL and above Se (ASCUS+, HSIL+)=86% Verification:0
Se (LSIL+, HSIL+)=92% Consecutive:0
Se (HSIL+, HSIL+)=97% Spectrum:1
Publication:1
Manual reading: Industry:1
Se (ASCUS+, HSIL+)=79%
Se (LSIL+, HSIL+)=86%
Se (HSIL+, HSIL+)=93%

Specificity cannot be
calculated

*Format for display of results of conventional Pap test: Prevalence (GS threshold)=disease/total=%; Sensitivity (Pap
threshold/GS threshold)=true positive/disease=%; Specificity (Pap threshold/GS threshold)=true negatives/non-
diseased=%

C-27