Context: Human papillomaviruses (HPVs) are known to cause most cervical cancer worldwide, but the utility of HPV DNA testing in cervical cancer prevention has not been determined.
Objective: To provide comprehensive data on the screening performance of HPV testing for the most common carcinogenic types, at different levels of analytic sensitivity.
Design: Laboratory analysis conducted during 1993-1995, using 3 cytologic techniques and cervicography, followed by colposcopic examination of women with any abnormal cervical finding, to detect all high-grade intraepithelial lesions and cancer (reference standard of clinically significant disease). The HPV testing was performed subsequently with masking regarding clinical findings.
Setting: Guanacaste Province, Costa Rica, a region with a high age-adjusted incidence of cervical cancer.
Participants: Of 11742 randomly selected women, 8554 nonpregnant, sexually active women without hysterectomies underwent initial HPV DNA testing using the original Hybrid Capture Tube test; a stratified subsample of 1119 specimens was retested using the more analytically sensitive second generation assay, the Hybrid Capture II test.
Main outcome measures: Receiver operating characteristic analysis of detection of cervical high-grade intraepithelial lesions and cancer by HPV DNA testing based on different cut points of positivity.
Results: An analytic sensitivity of 1.0 pg/mL using the second generation assay would have permitted detection of 88.4% of 138 high-grade lesions and cancers (all 12 cancers were HPV-positive), with colposcopic referral of 12.3% of women. Papanicolaou testing using atypical squamous cells of undetermined significance as a cut point for referral resulted in 77.7% sensitivity and 94.2% specificity, with 6.9% referred. Specificity of the second generation assay for positivity for high-grade lesions and cancer was 89.0%, with 33.8% of remaining HPV DNA-positive subjects having low-grade or equivocal microscopically evident lesions. The higher detection threshold of 10 pg/mL used with the original assay had a sensitivity of 74.8% and a specificity of 93.4%. Lower levels of detection with the second generation assay (<1 pg/mL) proved clinically nonspecific without gains in diagnostic sensitivity.
Conclusions: In this study population, a cut point of 1.0 pg/mL using the second generation assay permitted sensitive detection of cervical high-grade lesions and cancer, yielding an apparently optimal trade-off between high sensitivity and reasonable specificity for this test. The test will perform best in settings in which sensitive detection of high-grade lesions and cancer is paramount. Because HPV prevalence varies by population, HPV testing positive predictive value for detection of high-grade lesions and cancer will vary accordingly, with implications for utility relative to other cervical cancer screening methods.