Improving the Sensitivity-Specificity Balance of Human Papillomavirus Testing on Self- and Clinician-Collected Samples in South Africa
PURPOSE
Human papillomavirus (HPV) testing on self-collected samples may increase coverage of cervical cancer screening, but previous studies have observed lower specificity of HPV testing in self- versus clinician-collected samples. Here we investigate strategies to improve the sensitivity-specificity balance of a round of HPV testing on self-collected samples.
MATERIALS AND METHODS
Women living with and without HIV, age 30-65 years, were recruited in South Africa. Self-collected vaginal samples and clinician-collected cervical samples were tested with Xpert HPV, an assay that detects the 14 high-risk HPV types in five separate channels: (P1) HPV 16; (P2) HPV 18, 45; (P3) HPV 31, 33, 35, 52, 58; (P4) HPV 51, 59; and (P5) HPV 39, 56, 66, 68. All women underwent colposcopy with histology sampling, and diagnosis of cervical intraepithelial neoplasia grade 2 or greater (CIN2+) was determined by adjudicated pathology. The AUC and related performance parameters were calculated using logistic regression with the cycle threshold (Ct) values of the channels as predictors.
RESULTS
HPV prevalence in women without and with HIV was higher in self-collected (25.1% v 61.5%) than in clinician-collected samples (16.2% v 48.4%). The optimal model to predict CIN2+ used Ct values from the three channels that detect HPV 16, 18, 45, 31, 33, 35, 52, and/or 58. AUC was superior for testing on clinician-collected (0.908) than on self-collected samples (0.878; P = .0261) in women without HIV, as well as for women living with HIV (0.868 v 0.819; clinician v self; P = .0002). Alternate approaches to handling multiple types and sequential testing approaches did not allow self-testing to achieve equivalent performance to testing on clinician-collected samples.
CONCLUSION
Using more stringent Ct cutoffs on the three channels that detect the eight highest-risk HPV types can improve the sensitivity-specificity balance of a round of screening in both self- and clinician-collected samples. Although performance of HPV testing on self-collected samples is excellent, performance parameters are better on clinician-collected samples.