Abstract
While urine-based sampling for human papillomavirus (HPV) is being explored as a simple and noninvasive approach for cervical cancer screening, data comparing HPV genotyping in urine and those in cellular sampling of the cervix and vulva, and their correlation with rigorously confirmed cervical disease status, are sparse. We performed HPV genotyping on voided-urine and clinician-collected vulvar and cervical samples from 72 women undergoing colposcopy. Although urine-based HPV carcinogenic HPV detection was lower (58.3%) than cervical (73.6%) and vulvar (72.1%) detection (P = 0.05 and 0.07, respectively), the agreement of urine HPV with cervical and vulvar HPV was moderate (kappa = 0.55) and substantial (kappa = 0.62), respectively. Urine-based carcinogenic HPV detection had a clinical sensitivity of 80.8% (95% confidence interval [CI] = 60.7 to 93.5) and a specificity of 53.3% (95% CI = 37.9 to 68.3) for diagnosing cervical intraepithelial neoplasia grades 2/3 (CIN2/3) on histology; 90.0% of CIN3 was positive for urine HPV. The corresponding sensitivity and specificity values for vulvar sampling were 92% (95% CI = 74 to 99) and 40.5% (95% CI = 25.6 to 56.7), and those for cervical sampling were 96.2% (95% CI = 80.4 to 99.9) and 40% (95% CI = 25.7 to 55.7), respectively. HPV16 was the most common carcinogenic genotype detectable in 25% of urine, 33.8% of vulvar, and 31.9% of cervical samples overall, with prevalence increasing with cervical disease grade, regardless of the sampling method. Stronger cervical HPV PCR signal strengths were associated with increased frequency of urine HPV detection. In summary, the relatively lower detection rates but comparable clinical performance of urine-based HPV sampling underscore the need for larger studies to evaluate urine-based sampling for cervical cancer screening, epidemiologic studies, and postvaccination HPV disease surveillance.