Reproducibility of a rapid human papillomavirus test at different levels of the healthcare system in Tanzania: the AISHA study.

Background: To inform policy makers in Tanzania if and how best to implement rapid HPV testing, we assessed the interobserver reproducibility of careHPV test at three different levels of the healthcare system in an urban and a rural region of Tanzania. Methods: Women aged 30 to 50 years were screened by careHPV testing in two primary healthcare centers (PHC), two district hospitals (DiH), and two regional hospitals (ReH). Aliquots were retested at regional (ReH) and national referral laboratories (NRL). Reproducibility was evaluated using agreement and kappa index measures. Intralaboratory reproducibility was also evaluated in a set of 10 positive and 10 negative samples. Results: Samples from 1,134 women were locally tested and retested at ReH and/ or NRL. Test results from Dar es Salaam ReH and Kilimanjaro PHC showed clear quality problems including suspicion of contamination during testing or aliquoting. After excluding these samples, 18.8% of 743 women were HPV positive at clinic level. The resulting careHPV reproducibility at different levels of the healthcare system was very good [agreement 95.7%, 95% confidence interval ( CI), 94.0-96.9; kappa, 0.86, 95% CI, 0.81-0.91]. Intralaboratory agreement was also very good across four different experiments, with Fleiss' kappa between 0.87 (95% CI, 0.61-1.00) and 1.00 (0.75-1.00). Conclusions: Rapid HPV testing was highly reproducible between lower and higher levels of the healthcare system in Tanzania; however, performance seems to be operator dependent. Impact: The careHPVtest seems to be a feasible option for cervical cancer screening in an organized, decentralized system and in limited-resource settings if quality assurance measures are in place.