Page 48 - Screening for Cervical Cancer: Systematic Evidence Review

Chapter III. Results


studies (PapNet ), and 7, computer algorithms for selecting slides for rescreening or for


screening (AutoPap ).

A single study evaluating ThinPrep liquid-based cytology and other screening

modalities in a population-based cohort of 8,636 Costa Rican women met strict review criteria

51
for prospective evaluation of the test characteristics of a new screening technology:

1. Cytology specimen obtained as a screening study;

2. Performance of new technology was evaluated by colposcopy and/or histology as the

reference standard;

3. Pap and reference standard were applied within an average of three months; and

4. A validation of normal cytology results was undertaken, such that a complete 2x2

table relating the new technology to the reference standard can be completed.

The study that met these criteria used colposcopy and histology to evaluate women in

three categories: (1) atypical squamous cells of uncertain significance (ASCUS) on cytology as

assessed by ThinPrep, PapNet, or conventional smear; (2) positive cervigram; or (3) physical

examination suspicious for cancer or gynecologic emergency (rare indication). These

investigators conducted a validation substudy among a random sample of 150 women with

negative screening results, thus allowing estimation of specificity and predictive values. Their

reference standard was a composite of all screening tests and histology findings, called the final

case diagnosis. The final case diagnosis categories were negative (including ASCUS with

normal colposcopy and/or histology), low-grade squamous intraepithelial lesion (LSIL) (by

histology, or by cytologic confirmation by two or more methods), HSIL (93% histologic

confirmation), invasive carcinoma (100% histologic confirmation), and equivocal. Equivocal

included women with a single cytologic diagnosis of LSIL, isolated positive cervigram, or

48