Background: Co-testing for HPV, the most commonly diagnosed sexually transmitted infection, has recently been incorporated into U.S. cervical cancer screening guidelines. Currently, women who are HPV positive (HPV+) with ASC-US or worse cytology (ASC-US+) are referred to immediate colposcopy. Those with normal cytology may be triaged with HPV16/18 genotyping and referred to colposcopy if HPV16/18+; however implementation of this approach in routine clinical practice has not been evaluated. Here we use co-testing data from the Persistence and Progression cohort at Kaiser Permanente Northern California (KPNC) to evaluate the performance of cytology and HPV16/18 genotyping among HPV+ women 30+ years of age.
Methods: We evaluated 34,242 women who tested HPV+ by Hybrid Capture 2 (Qiagen, Gaithersburg, MD). Cases of cervical intraepithelial neoplasia grade 2 or worse that developed within approximately two years (CIN2+, n=4,154) plus a subset of <CIN2 (n=3,517) were genotyped using the Cobas HPV test, which separately detects HPV16/18 genotypes plus a pool of 12 additional carcinogenic HPV types (Roche Molecular Systems, Pleasanton, CA). We calculated sensitivity and specificity, simulating the performance of cytology and HPV16/18 genotyping for triage of HPV+ women.
Results: Sensitivity and specificity of cytology for CIN2+ were 72.3% and 58.8%, respectively. Among women with normal cytology, the sensitivity and specificity of HPV16/18 genotyping for CIN2+ were 41.7% and 81.1%, respectively. Sensitivity and specificity for the combination of cytology and HPV16/18 detection for CIN2+ were 84.4% and 47.7%, respectively.
Conclusions: Among HPV+ women, a combination of ASC-US cytology and HPV16/18+ genotyping would immediately refer 84% with CIN2+ to colposcopy, leaving 16% who would be referred at 1-year follow-up. Nearly half of HPV+ women who do not develop CIN2+ would still be sent to colposcopy. Management of HPV+ women while avoiding overtreatment remains a key challenge; whether this performance is adequate will depend on cost-effectiveness analyses.