Should cervical cytologic testing be augmented by cervicography or human papillomavirus deoxyribonucleic acid detection?

Criticism of the Papanicolaou smear in the lay press and recent federal legislation regulating cytology laboratories indicate a need to reappraise cervical cancer screening programs. This study directly compares three potential screening tests, used alone or in combination. A total of 1012 women aged 18 to 35 years were screened by cytologic testing, cervicography, and hybridization for human papillomavirus deoxyribonucleic acid, with discrepancies being referred by the last two authors. After findings from the entire lower genital tract were combined, 116 women (11.5%) showed definite clinical abnormalities (either exophytic vulvovaginal condylomas or cervical squamous intraepithelial lesions). Another 72 (7.2%) had positive Southern blot hybridizations without accompanying viral expression, yielding a cumulative frequency for established disease or latent infection of 18.6%. When associated vulvovaginal condylomas are disregarded, final groupings with regard to cervical pathologic classification were: 23 high-grade and 71 low-grade squamous intraepithelial lesions, 164 cases of equivocal atypia (34 of which had detectable human papillomavirus deoxyribonucleic acid), and 754 cases with negative results (38 of which had detectable human papillomavirus deoxyribonucleic acid). Cervical screening tests were compared principally by plotting increasingly liberal recall criteria onto receiver operating characteristic curves (i.e., graphs of true-positive results on the Y axis versus false-positive results on the X axis). Used individually, each screening test was valid, but none was substantially better than the others. No matter how liberal the recall criteria, no single test was able to detect all of the 23 definite precursors in this sample. Applying conventional recall criteria (i.e., high- or low-grade lesion suspected), cytologic testing alone detected 12 high-grade squamous intraepithelial lesions (52.2%), at a cost of having to perform colposcopy in 8.7% of the sample. Combining all three tests and setting the end point as just a high-positive result by at least one test, 19 high-grade squamous intraepithelial lesions (83%) were detected, with a recall of 7%. Optimal test performance (96% sensitivity, 4% recall) would have been attained by recalling all patients with high-grade cytologic results or positive cervicography results, plus any patients with low-grade morphologic atypia in which hybridization detected an oncogenic human papillomavirus type. Our conclusions are as follows: (1) Cytologic detection rates are markedly improved by a second or third test; (2) increased screening costs could be offset by not recalling patients with minor lesions with no apparent potential for progression.