Evaluation of self-collected vaginal specimens for the detection of high-risk human papillomavirus infection and the prediction of high-grade cervical intraepithelial lesions in a high-burden, low-resource setting

Objectives To compare the performance of self-collected vaginal (V) specimens with clinician-collected cervical (C) specimens for detection of high-risk human papillomavirus (hrHPV) and cervical disease using the Cepheid Xpert HPV, Roche Cobas 4800 HPV and Hologic Aptima HPV assays. Methods Women aged 30–59 years (n = 1005) were recruited at two clinics in Papua New Guinea, and they provided specimens for testing at point-of-care using the Xpert HPV Test, and for subsequent testing using the Cobas HPV (n = 981) and Aptima HPV (n = 983) assays. Liquid-based cytology was performed on C specimens to predict underlying high-grade squamous intraepithelial lesions (HSIL). V specimen results of each assay were evaluated against a constructed reference standard and for detection of HSIL or worse. Results There was substantial (κ >0.6) agreement in hrHPV detection between V and C specimens across all three assays. The sensitivity, specificity, and positive and negative predictive values of Xpert HPV using self-collected V specimens for the detection of HPV type 16 according to the constructed reference standard were 92.1%, 93.1%, 63.6% and 98.9%, respectively; compared with 90.4%, 94.3%, 67.8% and 98.7% for Cobas 4800 HPV; and 63.2%, 97.2%, 75.0% and 95.3% for Aptima HPV. Similar results were observed for all hrHPV types (combined) and for HPV types 18/45, on all three assays. The detection of any hrHPV using self-collected specimens had high sensitivity (86%–92%), specificity (87%–94%) and negative predictive value (>98%) on all assays for HSIL positivity. Conclusions Xpert HPV, using self-collected vaginal specimens, has sufficient accuracy for use in point-of-care ‘test-and-treat' cervical screening strategies in high-burden, low-resource settings.