Living Network Meta-Analyses to Evaluate Effects of HIV Self-Test Distribution Strategies on HIV Testing Uptake

Background: HIV self-testing (HIVST) has the potential to reach underserved populations in need of HIV testing services. Given the range of possible distribution strategies available to those implementing HIV self-testing, we undertook a systematic review and network meta-analysis (NMA) to evaluate the comparative effects of self-test distribution strategies on HIV testing uptake. Methods: We searched electronic databases, clinical trial registries and conference abstracts for randomized controlled trials (RCTs) that compared HIVST distribution strategies with each other or with traditional HIV testing by health care workers (HCWs) at health facilities or in the community, between 4 June 2006 and 4 June 2019, and applied PRISMA guided systematic review techniques and GRADE methodology for network meta-analyses. We grouped HIVST distribution strategies into six categories based on who distributed the self-test and where and how the test was distributed. We used Bayesian Hierarchical random logit models to generate network estimates and rank comparative effects of each strategy on HIV testing uptake. To account for heterogeneity and avoid intransitivity we divided networks by regions and conducted sensitivity analyses and meta-regression. We plan updating the review every six months or at the time of publication of important new studies. PROSPERO registration number: CRD42018088078. Findings: Searches yielded 14,333 records, resulting in 24 RCTs for inclusion. There was substantial heterogeneity of setting, target population and HIVST distribution strategies. For HIV testing uptake in the sub-Saharan Africa network, partner HIVST distribution (predominantly distribution to male partners) ranked highest (1st, 78% probability) compared to five other HIV testing strategies: traditional facility-based HIV testing (RR: 2.39, 95% CrI 1.65-3.45), facility-based HIVST distribution (RR:1.42; 95%CI: 0.87-2.31), health worker community HIVST distribution (RR:1.68; 95%CI: 0.90-3.14), health worker community traditional HIV testing (1.74; 95%CI: 0.60-5.08) and peer community HIVST distribution (RR:1.82; 95%CI:1.00-3.27). Facility based HIVST distribution ranked 2nd (46% probability) and traditional facility-based testing ranked lowest - 6th (64% probability). In the North American-Asia and Pacific region network where the focus was primarily on men who have sex with men (MSM), web-based and mail HIVST distribution ranked highest (ranked 1st with 93% probability) compared to two testing strategies: traditional facility-based self-testing (RR 1.53, 95% CrI 1.1-2.41), and facility-based HIVST distribution (RR: 1.15; 95%CI: 0.57 -2.00). All HIVST strategies ranked higher than traditional HIV testing services at health facilities in terms of HIV testing uptake. This analysis was limited by few studies contributing to network estimates, substantial qualitative heterogeneity of testing strategies and the use of self-reported outcomes which resulted in findings being downgraded to low or very low certainty evidence in the GRADE system. Interpretation: HIVST increases testing for underserved groups (specifically men). Distribution by sexual partners in sub-Saharan Africa and through mail among MSM populations in North AmericanAsia and Pacific region, combined with direct HIVST distribution by HCWs at health facilities, could be considered for implementation in these settings. Taken together, these analyses demonstrate the application of network meta-analysis techniques to explore HIVST implementation approaches in RCTs. A living approach will incorporate research findings as they emerge and enable the public health community to reach firm conclusions about HVST distribution strategies rapidly. Funding: NIH: K24 AI134413; Bill and Melinda Gates Foundation: OPP1215984 Declaration of Interests: Authors have no competing interests.