Accuracy of HPV testing on self-collected and clinician-collected samples for different screening strategies in African settings: A systematic review and meta-analysis.

ABSTRACT

OBJECTIVES: Cervical cancer still poses a considerable threat to women in low- and middle-income countries, particularly on the African continent. Self-collection of a vaginal sample promises advantages over the established sampling by clinicians. We aimed to assess the accuracy of self-sampling compared to clinician sampling in order to inform its application in primary care in the African context. METHODS: We searched Pubmed, Livivo, Web of Science, Cochrane Library and African Index Medicus on the 07th of February 2022. The eligibility criteria were reporting (i) self-sampling against clinician-sampling, (ii) study location in Africa, (iii) relevant outcome-measures: (a) Cohen's kappa (b) sensitivity and specificity of self-sampling tests. We combined Cohen's kappa effects, additionally sensitivity and specificity estimates using random-effects models. The study is registered with PROSPERO (CRD42020218081). RESULTS: We included 28 studies in the systematic review and 21 studies in the meta-analysis. Self-sampling was used to test for high-risk HPV infections. Two studies additionally tested for low-risk HPV infections. The pooled Cohen's kappa was 0.66 (95%CI 0.61-0.71). Populations at risk yielded 0.63 (95%CI 0.56-0.71). Target amplification tests based on PCR performed best with a kappa of 0.68 (95%CI 0.63-0.73) compared to isothermal mRNA tests, 0.61 (95%CI 0.51-0.71). Point of care tests performed exceptionally well, 0.73 (95%CI 0.67-0.80). Sensitivities are close to 80% and specificities close to 90% of self-sampling to detect high-risk HPV. CONCLUSIONS: Self-sampling agrees moderately to substantially with clinician sampling in the African context. Point of care tests might be particularly suited for application in cervical cancer primary screening in low- and middle-income countries. Populations at risk should get special attention while using self-sampling. Screening protocols should be established.