Community delivery of antiretroviral drugs: A non-inferiority cluster-randomized pragmatic trial in Dar es Salaam, Tanzania

BACKGROUND: With the increase in people living with HIV in sub-Saharan Africa and expanding eligibility criteria for antiretroviral therapy (ART), there is intense interest in the use of novel delivery models that allow understaffed health systems to successfully deal with an increasing demand for antiretroviral drugs (ARVs). This pragmatic randomized controlled trial in Dar es Salaam, Tanzania, evaluated a novel model of ARV community delivery: lay health workers (home-based carers [HBCs]) deliver ARVs to the homes of patients who are clinically stable on ART, while nurses and physicians deliver standard facility-based care for patients who are clinically unstable. Specifically, the trial aimed to assess whether the ARV community delivery model performed at least equally well in averting virological failure as the standard of care (facility-based care for all ART patients). METHODS AND FINDINGS: The study took place from March 1, 2016, to October 27, 2017. All (48) healthcare facilities in Dar es Salaam that provided ART and had an affiliated team of public-sector HBCs were randomized 1:1 to either (i) ARV community delivery (intervention) or (ii) the standard of care (control). Our prespecified primary endpoint was the proportion of adult non-pregnant ART patients with virological failure at the end of the study period. The prespecified margin of non-inferiority was a risk ratio (RR) of 1.45. The mean follow-up period was 326 days. We obtained intent-to-treat (ITT) RRs using a log-binomial model adjusting standard errors for clustering at the level of the healthcare facility. A total of 2,172 patients were enrolled at intervention (1,163 patients) and control (1,009 patients) facilities. Of the 1,163 patients in the intervention arm, 516 (44.4%) were both clinically stable on ART and opted to receive ARVs in their homes or at another meeting point of their choosing in the community. At the end of the study period, 10.9% (95/872) of patients in the control arm and 9.7% (91/943) in the intervention arm were failing virologically. The ITT RR for virological failure demonstrated non-inferiority of the ARV community delivery model (RR 0.89 [1-sided 95% CI 0.00–1.18]). We observed no significant difference between study arms in self-reported patient healthcare expenditures over the last 6 months before study exit. Of those who received ARVs in the community, 97.2% (95% CI 94.7%–98.7%) reported being either “satisfied” or “very satisfied” with the program. Other than loss to follow-up (18.9% in the intervention and 13.6% in the control arm), the main limitation of this trial was that substantial decongestion of healthcare facilities was not achieved, thus making the logic for our preregistered ITT approach (which includes those ineligible to receive ARVs at home in the intervention sample) less compelling. CONCLUSIONS: In this study, an ARV community delivery model performed at least as well as the standard of care regarding the critical health indicator of virological failure. The intervention did not significantly reduce patient healthcare expenditures, but satisfaction with the program was high and it is likely to save patients time. Policy-makers should consider piloting, evaluating, and scaling more ambitious ARV community delivery programs that can reach higher proportions of ART patients. TRIAL REGISTRATION: ClinicalTrials.gov NCT02711293.