HIV incidence in Sub-Saharan Africa (SSA) remains well above global elimination targets.¹Despite an increasingly large, evidence-based toolkit to prevent HIV transmission, there were over 1 million new HIV infections in SSA in 2018.¹We know from SEARCH²and three other recent SSA population-level studies^3-5that reaching (and exceeding) the UNAIDS 2020 population-level viral suppression target of 73%⁶is insufficient to achieve epidemic control (Table 1). We also know that key sub-populations, in particular youth,^{4, 7-10}men,^{2, 4, 10-12}mobile populations,^13-15and persons with heavy alcohol use^{16, 17}continue to fall short of viral suppression targets, putting them at higher risk for HIV complications, mortality, and onward HIV transmission. Finally, pre-exposure prophylaxis (PrEP), an intervention that has reduced incidence in affected populations in some areas where high rates of viral suppression were already present,^{18, 19}has not been effectively deployed in high risk populations (such as young women) in SSA.^{20, 21}To achieve control of the HIV epidemic and mitigate its consequences, we need new and scalable interventions to improve delivery of prevention and treatment to vital segments of the population that contribute to persistence of new infections.  We describe below why previous approaches that were effective for the majority of people in the ART roll-out do not work for subsets of these “persistent-driver” populations, our proposed solutions, and how a multi-component intervention trial can effectively inform stakeholders on programming and resource allocation to accelerate reductions in HIV incidence.

Barriers to effective HIV prevention and treatment outcomes among key, persistent-driver populations fall into three major themes based on epidemiologic and qualitative data from SEARCH on seroconversions (N=704), PrEP utilization (3,489 starts²²), viral suppression sub-group analyses (N=13,529^{2, 10, 16}), and other published literature. First, the majority of new infections are in youth, and in this growing, at-risk population, social and economic concerns often far outweigh health concerns on a day-to-day basis (individual barriers).^23-26There has been insufficient recognition that at any given moment in time, health is not a preeminent concern, particularly for youth. Indeed, poverty itself may increase “present bias” – the tendency  for short term considerations to outweigh long term ones – contributing to decision-making that harms health.^27-30Second and third, lack of flexibility in how prevention and treatment services, respectively, have been offered to people whose needs and behavior vary from mainstream community members (e.g. mobile populations, drinkers) and may be continually changing (systems barriers) have impeded uptake and impact of these services.^31-35

Modelling and phylogenetic studies suggest that further progress will require simultaneous and coordinated approaches in HIV treatment and prevention among persistent driver populations. Individual, “narrow” focused interventions (PrEP) for adolescent girls are important, but alone will not reach HIV elimination levels. Persistent driver populations are minority subsets of demographic populations of considerable size (e.g. while men are at elevated risk of non-suppression, most HIV-infected men are suppressed); some of the barriers they face (e.g. inflexible health systems) are shared across the demographic groups and can be integrated into a personalized public health approach that combines i) foundational changes in health care delivery systems that are scalable and affordable (e.g., re-structuring treatment visits for youth to prioritize non-health concerns) with ii) the option to access more intensive services for the targeted group of persons for whom basic one-size-fits all approaches are insufficient (e.g. intensified alcohol counseling for heavy drinkers and “travel packs” for mobile populations). Assumptions that strategies to reach these populations will be prohibitively expensive or too complex for programs to implement are unsubstantiated, and stakeholders need well-designed studies incorporating multiple interventions for persistent driver populations, and integrating intensification based on choice, to understand what works and inform resource allocation as they move beyond the “90-90-90” era towards HIV elimination.    

To address these barriers for persistent driver populations to increase prevention coverage for HIV- persons and viral suppression for HIV-infected persons, we used the empirically-validated PRECEDE model³⁶where health decision-making and health-seeking behaviors can be influenced via interventions that predispose, enable, and reinforce desired engagement with prevention and treatment services (see Section 8, below for comprehensive description of our PRECEDE intervention framework). 

Dynamic HIV prevention: Our 3 Dynamic Prevention components address barriers we identified in offering PrEP to >10,000 persons with risk factors for HIV acquisition in the region (Table 2). First, we will engage youth in HIV prevention and reproductive health services by generating demand for these services within existing life skills and vocational programs (BRAC youth development clubs in Uganda and micro-finance clubs in Kenya) via integration of key health and socio-economic services. For youth, embedding HIV services within a broader, multi-sector strategy for individual and community empowerment via economic and social advancement can lead to better decision-making and also increase demand for and engagement with HIV services. Similar approaches have been used in diverse settings and may overcome present bias and other barriers faced by youth living in poverty by generating demand for services while improving future expectations and decision-making via increased opportunity.^37-44Despite a strong evidence base that these interventions can impact self-reported behaviors, and in some cases unintended pregnancies and STIs, and recent encouraging evidence from the DREAMS program (HIV incidence reduction in non-randomized longitudinal analyses⁴⁵), there is little rigorous evidence that these interventions can reduce HIV incidence (as noted in a systematic review of income generation interventions for HIV prevention³⁷) or impact biomedical prevention (e.g. PrEP, PEP, condom use) uptake or adherence from randomized studies. We provide a comprehensive review of key completed and ongoing studies in Table 3. Thus our approach to engage young women and men through these existing programs that offer life skills and vocational training opportunities for economic and social advancement, along with embedded HIV and reproductive health education and expedited referrals for HIV prevention is different and complimentary to the DREAMS program and a potentially efficient way to engage youth that has not been rigorously tested.

HIV prevention services that offer choice of delivery options (PrEP & PEP) and flexibility in their use and location of delivery over time allow for a dynamic, patient-centered model of prevention. This approach is responsive to the reality of an individual’s changing needs, thereby facilitating continued engagement. In SEARCH, PrEP use declined over time in spite of initial enthusiasm and uptake by early adopters.⁵⁴In multiple settings, poor adherence to PrEP remains its Achilles’ heel.^{55, 56}Rather than presenting PrEP in a similar framework to ART (i.e. lifelong use with 100% adherence for efficacy with delivery at HIV clinic), approaches that incorporate flexibility in choice and delivery location, akin to contraception options for women in developed countries, may prove more successful. Indeed, over the past four decades, increased choice in contraception and flexibility in access have been directly associated with increased use.^{57, 58}Framing prevention in the context of healthy choices, rather than an assessment of risk, may also reduce the association of PrEP with “risky” behavior or promiscuity. Offering PEP – an event-driven option that does not require accurate prediction of risk before a potential HIV exposure (akin to a “morning after” pill), and which additionally provides protection for those with less sexual agency – is a critical innovation in flexible access to prevention services. PEP can also act as a “gateway” to PrEP and other prevention modalities (Figure 1), providing infrastructure to add new prevention options as they become available (e.g. the PrEP vaginal ring^{59, 60}). 

Dynamic HIV Treatment: offering choice and flexibility in HIV clinical care and ART access that is tailored to the needs of key populations – including youth, men, mobile, and heavy alcohol-using populations – may result in increased care engagement and viral suppression rates among these populations, compared to differentiated care models as currently conceptualized. We have identified key barriers and facilitators for engagement in treatment for each of these groups, obtained via quantitative surveys and focus group discussions among persons who have not engaged, dropped out of care or not achieved viral suppression in SEARCH and via existing literature (Table 4).^{33, 61-67}Treatment outcomes are most likely to be effective when these barriers are overcome. For example in youth, using the PRECEDE framework we have piloted our proposed intervention that combines (i) a “life stage” assessment framework that addresses the many dynamic changes in relationships, school and employment present in youth (PREDISPOSING) (ii) flexible clinic access (ENABLING) and (iii) rapid viral load feedback (REINFORCING). Details on treatment interventions for other key driver populations are in the protocol (see Section 8, below).   

The HIV epidemic will unnecessarily claim and continue to adversely affect millions of lives unless we deploy innovative strategies that extend the gains beyond those of universal testing and ART. Universal ART must be followed by dynamic prevention and treatment paradigms that allow for heterogeneity in patient need and preference, as well as changing preferences over time – while targeting interventions to those at greatest risk of incident and uncontrolled HIV. Furthermore, HIV cannot be addressed in a vacuum and must be coupled with health services that are seen by community members as a priority - including evaluation and treatment for diseases such as hypertension,^68-70diabetes,⁷¹and heavy alcohol use⁷², delivered in an environment that harnesses human capital among youth (the largest segment of the population in SSA⁷³). In our proposal, we postulate that within this broader framework, we can successfully reduce HIV incidence, population viral load and mortality, while improving community health outcomes. We will rigorously evaluate these effects and their costs using a 2-stage optimization and factorial cluster randomized design.

Title: A Multisectoral Strategy to Address Persistent Drivers of the HIV Epidemic in East Africa (SAPPHIRE)

Sample Size:

Phase A:  3,200 participants (2,100 HIV-negative participants, 650 HIV-positive participants, 450 hypertensive participants ≥15 years of age)

Phase B:  32 communities of ~10,000 persons each, for a total of ~160,000 adults (≥15 years of age)

Participating Sites:  44 communities in rural western Kenya and rural western Uganda (12 in Phase A, 32 in Phase B), that: a) are geopolitical units (i.e. parish in Uganda, or sub-location in Kenya) of ~10,000 persons each (5,000 ≥15 years) within the catchment area of government health clinics; b) have HIV prevalence similar to the region using Ministry of Health data; c) have a letter of commitment to the study from the local political leader; and d) are of sufficient distance from each other to avoid contamination.

Overall Objective: Our overall objective is to determine to reduce HIV incidence and to improve community health with multi-sector, scalable interventions.

Primary Objective: Reduce HIV incidence using innovative strategies for HIV prevention and treatment to simultaneously reach “persistent driver” populations.

Secondary Objectives:

1.    Evaluate and optimize individual intervention component effects, alone (Phase A) and in combination (Phase B), versus control conditions on prevention coverage and HIV viral suppression

2.    Assess the effect of the intervention package on other health outcomes (all-cause mortality, tuberculosis, hypertension linkage and control, heavy alcohol use and mother-to-child HIV transmission) in Phase B

3.    Evaluate behavioral and other mechanistic pathways for intervention effects on proximal mediators of HIV incidence in Phase B

4.    Assess the reach, effectiveness, patient and provider adoption, and fidelity and the maintenance of intervention components in Phase B

5.    Use final study data to inform a strategic and sustainable investment model that maximally reduces HIV incidence and improves community health for the combination interventions tested in Phase B

Study Design: This study will consist of 2 phases:

·         Phase A, in which randomized trials and pilot studies will assess effectiveness, fidelity and cost and improve context-specific “fit” of prevention and treatment interventions. Combining effectiveness with implementation, costing and modelling outcomes, Phase A will optimize intervention packages with context specific adaptations in structured consultation with stakeholders.

·         Phase B, which will evaluate the effects of these optimized dynamic prevention and treatment packages, alone and in combination, on prevention coverage, population-level suppression, and HIV incidence, as well as other health outcomes, in a balanced, community randomized 2x2 factorial design.

Interventions in Phase A:

●     Dynamic Prevention: We are evaluating ways to optimize dynamic prevention delivery with the following interventions: Provision of Dynamic Prevention intervention at 1) outpatient health centers, 2) antenatal/family planning clinics, 3) youth multisector events and  4) via village health workers interacting at households.

●     Dynamic Treatment: We are evaluating services to improve HIV and hypertension care in difficult to reach populations: mobile population services; alcohol use intervention for heavy alcohol users, and hypertension interventions.

Study Endpoints: 

Phase A: The primary endpoints in Phase A are: Prevention coverage (uptake and persistence of PrEP or PEP) and viral suppression (HIV RNA <400 c/mL).  For hypertension intervention, the primary endpoints are  linkage to care and hypertension control.

Phase B: The primary and secondary endpoints in Phase B are:

·         Prevention coverage (uptake and persistence of PrEP and PEP)

·         Viral suppression (HIV RNA <400 c/mL)

·         HIV incidence (cumulative 3 year)

·         Mortality (cumulative 3 year)

·         Tuberculosis (annual incidence rate)

·         Hypertension control (<140/90 mmHg)

·         Heavy alcohol use

·         Mother-to-child HIV transmission (infant HIV free survival at 3 years)

Study Duration: The study’s Phase A and Phase B components will be conducted over a 5-year period.

Sample Size and Power, Phase A

Dynamic Prevention: Assuming a standard deviation in covered time of 35% (based on SEARCH), a two-sided hypothesis test (a=0.05), and conservatively ignoring precision gains from stratified randomization and covariate adjustment, a trial with 200 evaluable persons per arm will have 80% power to detect an approximately 10% absolute effect size (corresponding, for example, to an increase in covered time from 30% to 40%). For primary endpoint measurement for the trials at clinics, we will thus enroll up to 300 participants per arm meeting PrEP/PEP eligibility criteria (1,200 total).  We also enroll up to 250 participants per arm from 16 areas serviced by village health workers (total N=500 HIV-negative persons).  For the youth intervention single arm pilots, we will enroll up to 100 persons/pilot-site to examine feasibility in primary analyses and prevention coverage in secondary analyses.

Dynamic Treatment: Assuming approximately 40% suppression (or 50% linkage in the hypertension trial) in the control arm and a two-sided hypothesis test (a=0.05), a trial of 100 persons per arm will be powered at 80% to detect a 20% increase in suppression (or linkage in the hypertension trial).  For the alcohol intervention trial, we will enroll up to 200 participants living with HIV per arm, for the mobility intervention trial we will enroll 125 participants living with HIV per arm, and for the hypertension linkage intervention we will enroll 125 participants per arm. For the hypertension treatment trial, we will enroll up to 200 participants (100/arm) based on assuming 40% control in the non-intervention group for 80% power to detect an 20% or greater increase in control.   

Sample Size and Power, Phase B:

Assuming an HIV- Cohort of approximately 1,175 persons per community, 90% follow up, mean 10% prevention covered time in the control arm, within-cluster standard deviation of 35%, and a coefficient of variation (k) of 0.15 (conservatively based on SEARCH), with 16 communities per arm we will be powered at 90% to detect a relative increase in covered time of approximately 24%. Assuming approximately 124 HIV-positive persons with year 3 HIV RNA level measured per community, 70% viral suppression at year 3 in the control arm, and k=0.10 (conservatively based on SEARCH), with 16 communities per arm we will be powered to detect a relative increase in suppression of approximately 15%. Under some but not all scenarios, this design will also have power >80% to detect an effect of combined Dynamic Prevention and Treatment on three-year cumulative HIV incidence compared to standard-of-care. Assuming a 3-year cumulative HIV incidence of 1.7% in the control arm of the study, k= 0.17 (informed by SEARCH), HIV test sensitivity of 95% and specificity of 99.9%, and 90% follow up of the HIV- Cohort (total N ~17,000 in primary analysis), we estimate >80% power to detect a relative reduction of at least 35% in the full intervention (N=8 communities) vs. control (N=8 communities) arms of the study.