Evaluation of p16/Ki-67 dual-stain cytology performed on self-collected vaginal and clinician-collected cervical specimens for the detection of cervical pre-cancer

To compare the performance of dual immunostaining of p16INK4a and Ki-67 proteins performed on self-collected vaginal specimens and clinician-collected cervical specimens, and to evaluate the performance of this technique in predicting high-grade disease. Women aged 30–59 years (n = 1005) were recruited at two well-women clinics in Papua New Guinea. Each woman provided both cervical and vaginal specimens that were tested for high-risk human papillomavirus (hrHPV) DNA using the Xpert HPV Test (Cepheid) at point of care. A subset of paired cervical and vaginal specimens (n = 243) were selected to undergo CINTec® PLUS (Roche) p16/Ki-67 dual-stain cytology and liquid-based cytology (LBC). Fifty-five pairs (22%) were excluded from further analysis because the smears were not assessable. Of the 189 remaining paired specimens, 74 pairs (39.1%) were positive for one or more hrHPV genotypes. When comparing results of the dual stain, the overall percent agreement, positive and negative percent agreements and κ value between the cervical and vaginal specimens were 87.8% (CI 82.3–92.1%), 64.6% (CI 49.5–77.8%), 95.7% (CI 91.0–98.0%) and 0.65 (CI 0.51–0.79%) respectively. The sensitivity of the dual stain performed on the cervical specimen to predict high-grade disease, determined by LBC, was superior to that of the dual stain performed on the vaginal specimen: 100% (CI 84.6–100%) versus 68.2% (CI 45.1–86.1%). Although further evaluation may be warranted, these findings indicate that dual-stain testing of vaginal specimens cannot be advocated as part of cervical screening programmes in low- and middle-income countries. However, dual-stain cytology performed on cervical specimens may have a role in quality assurance in such settings.