The arc of HIV epidemics in sub-Saharan Africa: new challenges with concentrating epidemics in the era of 90-90-90

Abstract

Purpose of review:

To examine the emerging results from the HIV universal test and treat (UTT) cluster-randomized trials in sub-Saharan Africa, discuss how expanding access to HIV clinical services is likely to reshape the arc of HIV epidemics, and consider implications for HIV prevention and control strategies in the coming decade.

Recent findings:

The effect of universal HIV testing followed by immediate ART on community-level HIV incidence remains unclear upon completion of five UTT trials. Only two of the four trials that measured HIV incidence found significant reductions in community-level incidence. Even in these trials, HIV incidence remained above epidemic control levels (≤1 case per 1,000 person-years) despite high levels of ART coverage and viral suppression. These findings may indicate that community-delivered HIV services are not reaching the high frequency transmitters who sustain HIV epidemics and are likely members of marginalized core groups that lack access to health information and services.

Summary:

With expanded access to HIV services in sub-Saharan Africa, HIV epidemics are transitioning from hyperendemic to declining/endemic epidemic phases, characterized increasingly by the re-concentration of HIV in marginalized core groups. To move towards epidemic control, novel HIV service delivery models and new technologies are needed to engage those who continue to drive HIV incidence in this new epidemic phase.

Introduction

The past two decades have seen tremendous progress in HIV prevention and control [1]. Global HIV incidence has declined by almost 50% since its peak in 1996, while globally the number of people living with HIV (PLWH) has grown as access to antiretroviral treatment (ART) expanded and consequently more HIV-positive people are living longer, healthier lives (Figure 1) [2,3]. ART has both treatment and prevention effects, largely eliminating HIV transmission risk among those who are virally suppressed with good adherence [4,5]. The expansion of access to ART in HIV clinics in sub-Saharan Africa – initially restricted to those with symptomatic disease and HIV-infected pregnant women, but more recently expanded to all living with HIV immediately upon diagnosis [6] – is one of the largest global health investments ever committed [7,8].

Figure 1. The HIV incidence:prevalence ratio for sub-Saharan Africa, 1990–2017.

The ratio of incident to prevalent HIV cases (UNAIDS 2018 estimates) has declined substantially as the number of new cases each year has fallen and the number of persons living with HIV as grown because of global expansion to ART services.

In 2014, UNAIDS released its 90-90-90 treatment targets, designed to help end the AIDS epidemic by 2030: assure that 90% of HIV infected individuals know their HIV status, 90% of HIV-infected individuals are on treatment, and 90% of those on treatment are virally suppressed by 2020 [9]. These targets were based on modeling that suggested that universal HIV testing followed by immediate treatment could dramatically reduce population-level HIV incidence and might even lead to HIV elimination over 15–20 years [10]. These treatment targets have gained much traction in sub-Saharan Africa. Many countries in this region measure the success of their national HIV prevention programs in relation to the 90-90-90 targets, and many international donors consider progress on these targets in decisions about continued funding [7].

While the 90-90-90 targets have galvanized resources and focused national strategies, the proportion of persons living with HIV who must be on treatment and achieve viral suppression in order to achieve meaningful decreases in population-level HIV incidence remains unclear, as does the long-term feasibility of approaches to sustain the required coverage. Recent community-randomized trials, designed to address these questions and conducted in a number of sub-Saharan African settings, have delivered mixed results and raised questions about the impact of HIV control strategies that focus primarily on “universal test and treat” (UTT) approaches [11**,12**,13**,14**,15**]. Despite high levels of ART coverage, community-level HIV incidence persisted at levels above epidemic control (≤1 case per 1,000 person-years [16]) and did not consistently decline in these trials. In this article, we examine the emerging results from the UTT trials, discuss how expanding access to HIV clinical services is likely to reshape the arc of HIV epidemics on the continent, and consider the implications of the trial findings and evolving HIV epidemics in sub-Saharan Africa for HIV prevention and control strategies in the coming decade.

The universal test and treat trials

Trial designs

The five UTT cluster-randomized trials were conducted in Botswana, Kenya, South Africa, eSwatini (formerly Swaziland), Uganda and Zambia between 2012 and 2018 (Table 1) [11**,12**,13**,14**,15**]. In each trial, intervention communities received home-based HIV testing and counseling (HTC) with referral of HIV-infected participants for immediate initiation of ART regardless of CD4 count. Four of the five trials were cluster-randomized by community, while the MaxART trial was cluster-randomized by HIV clinic.

Table 1.

Characteristics of the five cluster-randomized universal test and treat trials^*

	Setting & dates	Intervention communities				Comparison communities	National ART initiation changes
Trial		Test-and-treat	Linkage to care	Adherence support	Combo interventions
ANRS TasP	South Africa (Mar 2012 - June 2016)	Home-based HTC every 6 mos; referral to immediate ART.	Called participants who did not attend clinic within 3 mos (amendment)	None.	VMMC & condom referral.	Same home-based HTC + linkage to care + combo interventions as intervention communities; referral to ART per national guidelines.	Jan 2015: ≤350 to ≤500 CD4 cells/μl
SEARCH	Uganda & Kenya (June 2013-Dec 2016)	Hybrid HIV testing: annual 2-wk comm. health fairs (incl. HTC) + home-based HTC for non-attendees; referral to immediate ART.	Point-of-care CD4 count + appt at local clinic (<1 week) + clinic phone number + transport reimbursement	3-mo refills + multi-disease services + clinic phone # + appt reminders + phone contact & home-visit for missed appts	Health education + screening for hypertension, diabetes, & TB + prompt care for conditions detected.	Same hybrid HIV testing approach as intervention communities in the baseline year and year 3; referral to ART per national guidelines.	~Sep 2014: ≤350 to ≤500 CD4 cells/μl ~Sep 2016: Treat all
BCPP/YaTsie	Botswana (Oct 2013-Mar 2018)	Home-based, mobile, and targeted HTC; referral to immediate ART.	Point-of-care CD4, scheduled clinic appt, SMS reminder, active tracing if missed appt	Active tracing if missed appointment.	Community mobilization + expand & strengthened VMMC services + scale-up of PMTCT services.	Standard HIV testing and treatment services with some guidance and improved technical support for government clinics; ART per national guidelines.	June 2016: Treat all
PopART (HPTN 071)	South Africa & Zambia (Jan 2014 - June 2018)	Home-based HTC; referral to: (A) immediate ART or (B) ART according to national guidelines.	Repeat home visit for all PLWH to check ART initiation and retention on ART; accompaniment to clinic and extra counseling provided	Home-based adherence support from CHiPs + phone contact + extra counseling.	Home-based VMMC & PMTCT referral, STI & TB screening, and condom promotion/distribution.	Standard HIV testing and treatment services with some health system strengthening around HIV testing, VMMC, PMTCT, STI treatment, and TB services; ART per national guidelines.	~Sep 2014: ≤350 to ≤500 CD4 cells/μl ~June 2016: Treat all
MaxART	eSwatini (Sept 2014-Aug 2017)	Community mobilization (incl. HTC) 20–30 kms around each clinic; referral to immediate ART	None.	None.	Community education on the benefits of early ART	Same community mobilization (incl. HTC) + education as intervention communities; ART at clinics per national guidelines.	Dec 2015: ≤350 to ≤500 CD4 cells/μl Oct 2016: Treat all

Abbreviations: French National Agency for Research on AIDS and Viral Hepatitis (ANRS); treatment as prevention (TasP); Sustainable East African Research in Community Health (SEARCH); Botswana Combination Prevention Project (BCPP); Population Effects to Reduce HIV Transmission (PopART); Early Access to ART for All in Swaziland (MaxART); HIV testing and counseling (HTC); appointment (appt); voluntary medical male circumcision (VMMC); prevention of mother-to-child transmission (PMTCT); sexually transmitted infections (STIs); antiretroviral treatment (ART); tuberculosis (TB); community HIV-care provider (CHiP); people living with HIV (PLWH).

^*The ANRS TasP, SEARCH, BCPP/YaTsie and PopART trials were all community randomized, while the MaxART trial was clinic-randomized.

There were important differences in the design and implementation of the UTT trials. First, the level of support provided to participants who tested HIV positive in the intervention communities varied across the trials. Active linkage-to-care and follow-up for participants who tested HIV positive was offered to participants in SEARCH [13**,17], BCPP/YaTsie [14**,18], and PopART [15**,19]. HIV-positive participants in ANRS TasP [11**] received linkage-to-care facilitation mid-way through trial implementation, while those in MaxART [12**,20] received no linkage-to-care facilitation. ART adherence support was only offered in SEARCH [13**,17,21], BCPP/YaTsie [14**,18], and PopART [15**,19], and implementation differed from routine home-visits [15**,19] to active tracing of participants who missed scheduled clinic visits [13**,14**,17,18]. Second, the five trials implemented different combinations of additional HIV prevention interventions in their intervention communities. The other “combination HIV prevention” interventions in these trials included various forms of education on health [13**,17] and the benefits of early ART [12**,20], different types of screening and referral for hypertension [13**,17], diabetes [13**,17], sexually transmitted infections (STIs) [19] and tuberculosis [13**,17,19], and referrals to nearby clinics for voluntary medical male circumcision (VMMC) and prevention of mother-to-child transmission (PMTCT) [11**,14**,15**,18,19]. Third, intervention delivery in comparison communities varied across the five trials. In the ANRS TasP [11**], SEARCH [13**,17], and MaxART [12**,20], comparison communities, HTC and participant referral to HIV services were provided through community outreach in homes, at community health fairs, or via mobile services, while in BCPP/YaTsie [14**,18] and PopART [15**,19] the comparison community interventions were designed to strengthen services at local clinics and did not include community outreach.

When interpreting the results of these UTT trials, it is important to note that all six of the sub-Saharan African countries in which the trials were implemented changed their HIV treatment initiation guidelines during these trials [11**,12**,13**,14**,15**]. In 2013, the World Health Organization (WHO) increased their recommended CD4 threshold for ART initiation from ≤350 CD4 cells/μl to ≤500 CD4 cells/μl [22], and updated their recommendation to a treat-all approach in 2015 [23]. The years in which each of the countries changed their national HIV treatment guidelines to reflect WHO recommendations varied slightly, but by the end of all five trials, participants in the controls communities were receiving immediate ART treatment regardless of their CD4 count (Table 1) [11**,12**,13**,14**,15**].

Trial results

The effect of universal HIV testing followed by immediate ART on community-level HIV incidence remains unclear upon completion of the five cluster-randomized trials (Table 2) [11**,12**,13**,14**,15**]. Only two (BCPP/YaTsie and PopART) of the four trials that measured HIV incidence found that UTT delivered a significant reduction [11**,13**,14**,15**]. Of note is that significant increases in ART coverage and in the proportion of the population with undetectable viral load were not consistently associated with significant incidence reductions. SEARCH and BCPP/YaTsie both documented significant improvements in ART coverage and viral load suppression, but only BCPP/YaTsie also achieved a significant reduction in HIV incidence [13**,14**]. Furthermore, in PopART, a borderline significant increase in the proportion of HIV-infected participants with undetectable viral load was observed in the communities receiving the most intensive intervention (Arm A with referral for immediate ART), but this did not translate to a significant reduction in HIV incidence [15**]. Surprisingly, the opposite findings were observed in communities receiving the less intensive intervention (Arm B with referral for ART according to national guidelines) [15**]. In these communities, the proportion of HIV-infected participants with undetectable viral load did not increase significantly, while HIV incidence did fall significantly [15**]. It is noteworthy that both of the trials that did not see significant reductions in community-level HIV incidence (ANRS TasP and SEARCH), were those that adopted community or home-based outreach strategies in their comparison arms (including linkage to care assistance in ANRS TasP), potentially inadvertently decreasing the differences between intervention and comparison groups [11**,13**].

Table 2.

Summary of the results from the five cluster-randomized universal test and treat trials^*

Trial	Sample size	Duration follow-up	No. of clusters	HIV prev	ART coverage	Undetectable HIV VL	HIV incidence	Interpretation
ANRS TasP	28,419	52 mos (36 avg)	22 (11/arm)	30%	53% v. 53% (p=0.69)	46% v. 45% (p=0.21)	2.11/100 PY v. 2.27/100 PYs (p=0.89)	No significant increase in ART or undetectable VL coverage; no change in HIV incidence. Findings are likely attributable to limited linkage-to-care, no ART adherence support, and home-based HIV testing + linkage-to-care in the comparison communities.
SEARCH	150,395	72 mos	32 (16/arm)	4–19%	80% v. 40%*** (p<0.001)	79% v. 68% (p<0.001)	0.73% v. 0.76%**** (p=0.60)	Significant increases in ART and undetectable VL coverage; no change in HIV incidence. Universal testing in the control communities in the baseline year and year 3 may have diminished incidence findings.
BCPP/YaTsie	8,974	60 mos (36 avg)	30 (15/arm)	29%	98% v. 93% (p<0.001)	96% v. 91% (p<0.001)	0.59% v. 0.92%**** (p=0.04)	Significant increases in ART and undetectable VL coverage; significant reduction in HIV incidence. UTT and linkage-to-care are feasible and effective in this context, using this approach. However, despite high coverage, HIV incidence remained above level of epidemic control.
PopART (HPTN 071)**	52,500	54 mos (36 avg)	21 (7/arm)	21%	Arm A: 81% Arm B: 80% Arm C: not measured	A: 72% (p=0.07) B: 68% (p=0.30) C: 60%	A: 1.45/100 PYs (p=0.51) B: 1.06/100 PYs (p=0.006) C: 1.55/100 PYs	A: Compared with Arm C, borderline significant increases in undetectable VL; no significant reduction in HIV incidence. B: Compared with Arm C, no statistically significant increases in undetectable VL; highly significant reduction in HIV incidence. Null findings for Arm A for HIV incidence require further analysis. Even with high ART coverage and undetectable VL, HIV incidence far above level of epidemic control.
MaxART	3,405	36 mos	14 (2/step)	31%	86% v. 80% (p=0.005)	79% v. 4% (p<0.001)	No incidence outcome	Significant increases in ART and undetectable VL coverage; no HIV incidence outcome. Immediate clinic-based ART initiation can increase coverage and VL suppression.

Abbreviations: months (mos); number (no.); prevalence (prev); antiretroviral treatment (ART); viral load (VL)

^*The ANRS TasP, SEARCH, BCPP/YaTsie and PopART trials were all community randomized, while the MaxART trial was clinic-randomized.

^**PopART was a 3-arm cluster-randomized trials: A & B were two different intervention arms, while C was the standard-of-care arm (see Table 2). Viral load outcomes were reported at 24 months; HIV incidence outcome was cumulative at 36 months.

^***among baseline HIV-infected persons not on ART

^****annualized HIV incidence

Despite relatively high ART and undetectable viral load coverage among HIV-positive participants in the intervention communities of three of the UTT trials (SEARCH [13**,24] BCPP/YaTsie [14**] PopART [15**]), community-level HIV incidence remained strikingly high (Table 2). For example, in BCPP/YaTsie, HIV incidence in the intervention communities was 0.6% (6 per 1000 person-years), despite 98% ART coverage and 96% undetectable viral load among HIV-positive participants [14**], which are above the UNAIDS’ 90-90-90 [9] and even 95-95-95 [25] HIV treatment targets. While this community-level HIV incidence was significantly lower than the 0.9% (9 per 1,000 person-years) in comparison communities, it remains much greater than rates suggested to be consistent with HIV elimination or epidemic control [16]. The persistence of HIV incidence at these levels despite high ART coverage may be an indication that community-delivered HIV services are not reaching those central to HIV transmission, the characteristics of whom may be changing as HIV epidemics change in these settings with improved access to HIV treatment and prevention services.

The arc of HIV epidemics

Epidemic phases and changes in sub-population topology

Epidemics of HIV and other sexually transmitted infections (STIs) move through predictable phases that are shaped by a number of factors including the biological characteristics of the pathogen, the behaviors of the populations in which they emerge, the population’s prior exposure history, and the population’s access to and engagement with prevention and control services [26] (Figure 2). Like all sexually transmitted pathogens, HIV’s spread and survival in human populations depend on its reproductive rate, which is a function of its transmission efficiency, duration of infectiousness and the rate of sex partner change in the population [27]. Indeed, in order for HIV to persist in a population, if transmission efficiency and/or duration of infectiousness are reduced by interventions such as ART, pre-exposure prophylaxis (PrEP), VMMC or condoms that reduce infectiousness and/or susceptibility, rates of sex partner change must increase. This means that the characteristics and size of the sub-populations in which HIV is concentrated will change as HIV epidemics are reshaped by large-scale intervention programs and policies.

Figure 2. Characteristics of HIV epidemic phases and patterns.

Epidemics of HIV and other sexually transmitted infections (STIs) move through predictable phases shaped by factors including the biological characteristics of the population, biology and behaviors of the population, and access to prevention and treatment services.

The initial growth phase of HIV epidemics is primarily driven by small subsets of the population or “core groups” [28,29], with sexual, substance use, and/or health care-related behaviors that result in high rates of HIV acquisition and transmission (e.g., sex workers, their clients, men who have sex with men [MSM], and people who inject drugs [PWID]) [30]. In this initial phase, HIV infections are still relatively rare, compared to the entire population size, and many are acute infections, which may exhibit greater transmission efficiency than chronic, prevalent infections that dominate in later phases. In the absence of widespread recognition of HIV even among health experts, and without significant intervention tools or programs focusing on strategies such as behavioral risk reduction, STI treatment, VMMC or ART, biological and behavioral susceptibility are high, likely resulting in near peak transmission efficiencies and infectiousness with the potential to ignite explosive epidemics because of high HIV viral loads in the context of the high proportion of acute HIV infections and lack of access to ART.

The subsequent course of HIV epidemics is heavily dependent on the sexual and social networks that define these core groups and their members. As members of core groups mix sexually with members of the general population, HIV epidemics transition to a hyperendemic phase, a population-pathogen equilibrium with high prevalence rates in which the majority of new infections may occur within stable (but not necessarily monogamous), heterosexual sexual partnerships [31]. As infection becomes more prevalent, particularly in populations with social and financial resources and with effective contact with health care systems, access to behavioral and biomedical interventions expands, with increases in individual and community level behavioral change. However, both access and uptake frequently expand unevenly, reflecting not only resource allocation decisions, but also the fact that the social networks that fuel the spread of prevention information and health-seeking behaviors tend to have limited overlap with the sexual networks that fuel HIV transmission.

When access to HIV testing, treatment, and prevention interventions becomes widely available, HIV epidemics may slow and eventually enter an endemic phase in which incidence re-concentrates among core groups that may differ from those that launched the initial growth phase. These core groups are often characterized primarily by lack of access to health information and services due to poverty, remote geography, young age, stigma, and/or limited social capital for other reasons, but may also include individuals who do not engage in HIV services for other, complex reasons, including inaccurate risk perception and denial. Among those who do access HIV testing and interventions that reduce HIV transmission efficiency or duration of infectiousness, increases in rates of sex partner change and poor intervention adherence may be responsible for HIV incidence that sustains transmission and exceeds the threshold for elimination [26]. Decreased mortality and migration that brings additional infected individuals into the community may continue to fuel the epidemic.

In reality, HIV epidemics do not move smoothly through the growth, hyperendemic, and endemic phases outlined above, HIV incidence rates oscillate over time as a consequence of new interventions, new programs and policies, and/or changing social and political landscapes. However, HIV epidemics in sub-Saharan Africa have largely followed these patterns. For the past two decades, the majority of countries in the region have experienced hyperendemic HIV epidemics affecting large portions of the general population. Consequently, delivery models for HIV treatment and prevention in this region have been designed to reach the population at large, as well as key populations such as traditional core groups and their partners. There has been substantial progress, with many countries in the region realizing significant declines in HIV incidence, marking the beginning of a new, late epidemic phase and demanding reassessment and refinement of HIV prevention strategies, including UTT.

In this context, we hypothesize that in many high prevalence settings there may be a non-linear relationship between ART roll-out and its population-level effectiveness in reducing HIV incidence (Figure 3). At the individual-level, ART with adherence and viral suppression is, in essence, 100% effective at preventing HIV transmission [4,32], but at the population-level the relationship between ART coverage and reductions in HIV incidence may not be one-to-one because all individuals living with HIV do not have the same risk of transmitting the virus nor are the individuals who are diagnosed and treated the same as those who are not engaged in care. When ART coverage is low, HIV treatment frequently goes first to HIV-infected individuals with the most advanced disease progression who may present limited HIV transmission risk due to poor health, or to antenatal clinic clients to prevent maternal-to-child transmission. Healthy HIV-infected individuals with high rates of sex partner change have not usually been the top priority initially. As ART coverage becomes more widely available, the majority of individuals who present to clinics and test HIV-positive will be able to initiate treatment. Here, we begin to see treatment benefits because many of these individuals are sexually active and at risk of transmitting HIV to members of their sexual networks. However, as ART coverage grows via expanded access to clinics or referrals from community-based interventions (e.g., home-based testing [11**,13**,14**,15**]), treatment benefits may begin to plateau because of the challenges inherent in delivering HIV services to those who are high frequency HIV transmitters, many of whom are often hard to reach and/or hard to engage. These individuals – the majority of the last 27% of UNAIDS’ 90-90-90 target [9] – sustain HIV incidence and contribute to a growing fraction of HIV transmissions as effective coverage expands [33], potentially helping to explain why HIV incidence did not fall in some of the UTT trials despite high population-level ART coverage and viral load suppression.

Figure 3. Hypothetical non-linear relationship between ART roll-out and population-level effectiveness in reducing new HIV infections.

At the population-level the relationship between ART coverage and reductions in HIV incidence appear to not be one-to-one, likely because all individuals living with HIV do not have the same risk of transmitting the virus and the expansion of ART coverage often does not include marginalized groups with high frequency transmission.

Country examples

With expansion of HIV treatment and prevention services, HIV incidence in some sub-populations at HIV risk in a number of high-income countries (e.g., White MSM in the United States) has begun to decline and, in some cases, enter endemic phases, while incidence has remained high or increased in more marginalized groups (e.g., Black and Latino MSM in the US) [34,35]. There are suggestions that this transition is occurring now in other populations and settings.

Recent HIV incidence data from the Rakai Community Cohort Study (RCCS), an open population-based cohort in southeastern Uganda, supports the hypothesis that with the expansion of HIV treatment and prevention services, HIV incidence may re-concentrate in core groups that are marginalized or hard to engage [36*]. Since 2013, the RCCS has been providing home-based HTC and referral for ART and other combination HIV prevention services in four high HIV prevalence Lake Victoria fishing communities. Free HIV treatment and prevention services (including new community-based HIV clinics, promotion and expansion of VMMC services, universal HIV testing and treatment, and a community health worker program) were scaled up rapidly. With expanded access to these HIV treatment and prevention services, between 2001 and 2016 ART coverage and viral load suppression more than doubled, and HIV incidence was halved in these communities (Kagaayi J, Chang L, Ssempijja V, et al., unpublished data). However, persistent viremia was documented in almost 15% of community members who attended at least two clinic visits, and HIV incidence remained above levels to achieve epidemic control [16, 36*]. Persistent viremia was concentrated among younger men, never married women, and recent migrants into the four communities, while HIV incidence was associated with low education, previous marriage, two or more sexual partners, and self-reported genital ulcer disease, characteristics frequently seen among members of marginalized core groups [36*].

Data from Kenya suggest similar patterns. National HIV surveillance data show that HIV incidence has steadily declined over the past two decades [37,38], as progress on UNAIDS 90-90-90 treatment targets has increased [39], and other HIV prevention interventions, such as VMMC [40–43], have expanded coverage (Figure 4). In western Kenya, where the HIV epidemic has been most severe, progress has been striking, with reductions in HIV incidence from 11.1 per 1,000 person years in 2011–2012 to 5.7 per 1,000 person years in 2012–2016 [44]. Despite these successes, a quarter of HIV-infected adolescents, and 10% of HIV-infected adults have not achieved viral load suppression [39], and the latest estimate of national HIV incidence (1.5 per 1,000 adults aged ≥15 year in 2018) [37] is still above the ≤1 case per 1,000 person-years threshold for HIV epidemic control [16]. Preliminary analyses of data from a large 2018 population-based impact assessment survey that measured the reach and impact of HIV programs in Kenya also show high levels of HIV status knowledge, ART coverage, and viral load suppression, similar to the national surveillance data (Cherutich P, personal communication). In this survey, however, knowledge of HIV status appears to be lower among men, individuals with ≥2 sexual partners in the past 12 months, adolescents and young adults, while viral load suppression remains lower among adolescents and young adults, despite high linkage to ART among those known to be living with HIV in the various subgroups. Furthermore, in the national surveillance data adult HIV incidence remains high (in the range of 7 to 9 per 1,000 person-years) in some geographic areas, particularly those that are home to substantial rural populations, some of whom are living in poverty, such as in Kisumu, Siaya and Homa Bay [39]. This suggests that current HIV service delivery models may not be reaching substantial numbers of high frequency HIV transmitters who are marginalized due to age, geography, poverty, or other social factors or who choose not to engage with HIV services for other reasons.

Figure 4. National HIV surveillance data from the Kenya Ministry of Health, 2000–2018.

National HIV incidence has been declining in Kenya over the past two decades as progress on UNAIDS’ 90-90-90 treatment targets and other HIV prevention interventions, such as voluntary medical male circumcision (VMMC) coverage, have increased. Data courtesy of Peter Cherutich, Kenya Ministry of Health.

HIV prevention and control in the coming decade

There is growing evidence that high levels of ART coverage and viral suppression – sometimes at levels above UNAIDS’ 90-90-90 [9] and even 95-95-95 strategies [25] – may not translate to major reductions in population-level HIV incidence [33]. The UTT trial results [11**,12**,13**,14**,15**] and cohort [36*] and surveillance data [37–43] from sub-Saharan Africa suggest an epidemiologic “inconvenient truth”: not all individuals contribute equally to HIV transmission. Thus, the sub-population distribution of ART coverage and viral suppression matters, if our goal is reduction or elimination of HIV incidence [45]. In the coming decade, as countries in the region expand HIV prevention, testing, and treatment programs, and limit new HIV infections to a smaller number of communities and sexual and social networks, more HIV epidemics will transition to decline and endemic phases. However, increasing viral load suppression is likely to be hardest among those who matter most to HIV transmission, the core groups of individuals who lack access to or choose not to engage in health services for all of the reasons discussed above. Furthermore, the importance and challenge of reaching and engaging these core groups to achieve epidemic control (and prevent HIV drug resistance) [46], will continue to grow.

The delivery of HIV treatment and prevention services in sub-Saharan Africa, as well as the interventions commonly included in these services, must be adapted to reach those who are central to HIV transmission in this new epidemic phase [47]. If late-phase HIV epidemics are sustained by sub-populations with progressively less effective engagement with health systems because of remote geography, age, stigma, inaccurate risk perception, lower educational attainment, and/or lower social or economic status, new or enhanced strategies to address these barriers and engage members of these core groups in sustainable ways across their lifespans will be increasingly critical. By definition, it is no minor challenge, to reach the “hard to engage.” Public sector resources should be reallocated to focus increasingly on outreach strategies that address these challenges.

Several HIV service delivery models or novel interventions may be particularly effective in late phase epidemics. For example, to spread knowledge about HIV prevention, targeted educational campaigns that use social networks (e.g., peer educators, sexual partners) [48,49] or are based in “hot spots” [50] may be more effective at engaging core group members than community-level educational campaigns (e.g., at markets or schools) [13**]). To increase knowledge of HIV status, expanded hours of operation at both traditional and non-traditional HIV testing facilities (including clinics, pharmacies, and mobile services), the addition of youth-friendly spaces to existing clinics [47,48], or innovative delivery models for HIV testing (including peer-based [51,52] or partner [53] delivery of HIV self-tests, or expedited partner services [54,55]) may better engage individuals who have difficulties accessing traditional health services and consequently increase the frequency of HIV testing – which is especially important for individuals with high numbers of sexual partners or high rates of sex partner change. Finally, to expand ART access and adherence (for both HIV treatment and PrEP), peer referral or peer-based drug distribution models [56], new technologies and applications that support a broad array of local care providers with remote physician oversight [57,58], the delivery of drugs outside of clinics (e.g., pharmacies [57,59] or drug-dispensing ATMs [60]), and emerging long-acting ART formulations [61,62] may enhance coverage in these core groups.

As HIV incidence persists despite high coverage of prevention and treatment interventions, increased investment by sub-Saharan African governments and international donors in research to optimize the delivery of HIV services, enhance HIV surveillance systems to guide optimal targeting, refine strategies such as partner notification to better identify and engage with transmission clusters, and develop new prevention and treatment tools will be increasingly important. Research on development of long-acting antiretroviral agents, multipurpose prevention technologies, and HIV vaccines will be essential, as novel modalities that make prevention and treatment easier to deliver should permit more individuals to access services effectively [63]. In addition, expanded implementation science studies (including economic analyses) of current and new prevention and treatment options should be conducted to determine the effectiveness and cost-effectiveness of novel and targeted HIV service delivery models [64,65]. This research should have a particular emphasis on service uptake among the core groups that currently have limited engagement with the health system. Improving HIV surveillance, with granularity at the sub-population level (including geographic, socio-demographic, and key population breakdowns with special attention to enhanced data among adolescents), analytic capacity, and timely dissemination of results will help policy makers and program managers understand where HIV is re-concentrating in this new epidemic phase, and design HIV prevention and care strategies to best reach these individuals.

Conclusion

The arc of HIV epidemics in sub-Saharan Africa appears to be shifting. With the expansion of HIV prevention and treatment services, new infections are likely to re-concentrate in core groups that are increasingly challenging to reach and engage. We should celebrate this trajectory as evidence of long-awaited progress in taming the generalized epidemics that have ravaged the continent. But we must also acknowledge the profound challenges and call to action that it presents on the path to HIV control and ultimately elimination; a call to reassess, rebalance and, in some cases, reengineer our strategies and resource allocations to engage and support those who are most in need and most central to ending the HIV pandemic.