Comparison of the performance of paired urine and cervical samples for cervical cancer screening in screening population

Abstract The main objective of this work is to determine the performance of urine for human papillomavirus (HPV) detection in cervical cancer screening in screening population. Paired urine and cervical samples were collected from 2038 women (careHPV group: 1002, cobas4800 group: 1036) in 2015. Urine was tested by a new urine‐based HPV test and cervical samples by careHPV or cobas4800 HPV test. Women were triaged based on cervical results and then referred to colposcopy with biopsy as clinically indicated. In 2017, women were followed up and screened with cotesting strategy, women with any positive would be referred and biopsied if necessary. In careHPV group, the HPV prevalence of urine was 14.1%, and 16.4% for cervical samples. In cobas4800 group, it was 19.1% and 20.4%, correspondingly. The concordance of urine samples compared with cervical samples was moderate (careHPV group: 86.6%; κ = 0.48; cobas4800 group: 83%; κ = 0.46). The baseline sensitivity and specificity for urine against CIN2+ detection were 85.7%, 86.8% in careHPV group, and 69.2%, 82.3% in cobas4800 group, respectively. Cervical samples were 100% sensitive for both tests (careHPV and cobas4800) and 85.2% specific in careHPV group and 81.9% specific in cobas4800 group, respectively. The corresponding cumulative sensitivity and specificity were 68.8% and 87.1%, 58.8% and 81.9%, 87.5% and 85.5%, and 94.1% and 81.4%. Urine demonstrated certain potential in cervical cancer screening and could be an alternative if no better screening strategies available.