Introduction

In 2019, there were 1.7 million new HIV infections globally, far exceeding the UNAIDS 2020 target of 500,000 new infections annually [1]. Oral pre-exposure prophylaxis (PrEP) with tenofovir disoproxil fumarate/emtricitabine (TDF/FTC) is highly effective for HIV prevention [2,3] and could accelerate reductions in HIV incidence in combination with other approaches such as treatment as prevention [4]. In high-income settings where both antiretroviral therapy (ART) and PrEP have been scaled up, declines in incidence have been observed among men who have sex with men (MSM) [5,6]. In sub-Saharan Africa, which accounted for 59% of new infections globally in 2019 [1], there have been remarkable gains in ART coverage, but PrEP rollout is just beginning to expand in many settings. As a result, data on HIV incidence among PrEP users in Africa remain limited outside of placebo-controlled trials and open-label studies of specific risk groups, such as serodifferent couples [7], young women [8], MSM [9], and female sex workers [10].

As PrEP is scaled up in generalized epidemic settings, optimal strategies for delivery (including care provision, refills, and laboratory monitoring) are needed to realize its potential to reduce new HIV infections. Community-based HIV testing with supported linkage to treatment or prevention, including PrEP, provides an opportunity to engage individuals (such as adolescent girls, young adults, and men) who may not otherwise access health services. Moreover, lower-barrier models for PrEP service delivery, including options for out-of-facility community-based visits, may further help to enhance retention of individuals in HIV prevention services.

We offered universal access to PrEP with an inclusive approach to eligibility during population-level HIV testing in 16 communities in rural Kenya and Uganda. We provided rapid PrEP start on-site at health fairs and at clinics and a flexible delivery model with follow-up visits at either clinics or community-based sites for both HIV testing and PrEP refills. We sought to evaluate HIV incidence among PrEP initiators and characterize incident HIV infections on PrEP, including clinical and virologic outcomes.

Methods

Results

Study participants and PrEP uptake

A total of 76,132 individuals ≥15 years old not previously diagnosed with HIV received HIV testing in the 16 study communities (Fig 1) from June 2016 to April 2019. Overall, 74,541 tested negative for HIV, of whom 15,632 (21%) were assessed to be at elevated HIV risk, and 5,447 (35%) initiated PrEP (S2 Table). Among the 5,447 PrEP initiators, 49% were women, 29% were age 15 to 24 years, 16% were age ≥45 years, and 19% were in serodifferent partnerships (Table 1; S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PrEP uptake following population-level HIV testing in 16 communities in rural Kenya and Uganda.

Individuals assessed to be at elevated risk of HIV acquisition (based on serodifferent partnership, HIV risk score, or otherwise self-identified HIV risk) were offered enhanced individual counseling on PrEP. Individuals neither in serodifferent partnerships nor identified by the risk score could self-identify as at risk of HIV acquisition. PrEP, pre-exposure prophylaxis.

https://doi.org/10.1371/journal.pmed.1003492.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of individuals who initiated PrEP in 16 communities in rural Kenya and Uganda.

https://doi.org/10.1371/journal.pmed.1003492.t001

HIV incidence among PrEP initiators

Among 5,447 PrEP initiators, 4,260 (78%) had at least 1 subsequent HIV test after PrEP initiation (S4 Table). There were 25 incident HIV infections over 7,150 person-years of follow-up. The HIV incidence rate was 0.35 per 100 person-years (95% CI 0.22 to 0.49) overall, 0.46 per 100 person-years (95% CI 0.24 to 0.68) among women, and 0.23 per 100 person-years (95% CI 0.09 to 0.41) among men (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Observed HIV incidence rate among PrEP initiators, overall and stratified by sex.

Observed HIV incidence rate per 100 person-years among PrEP initiators in all 16 study communities. CI, confidence interval; PrEP, pre-exposure prophylaxis; PY, person-years.

https://doi.org/10.1371/journal.pmed.1003492.g002

HIV incidence among PrEP initiators compared to HIV incidence among recent matched controls prior to PrEP availability

In 8 of the 16 study communities, we compared observed HIV incidence among PrEP initiators to HIV incidence without PrEP, based on propensity score–matched recent controls from the year before PrEP was available. Median follow-up time among matched recent controls was 0.90 years (interquartile range [IQR] 0.83 to 0.97) compared to 1.61 years (IQR 1.02 to 2.45) among PrEP initiators. Among matched recent controls (in the absence of PrEP), there were 17 incident HIV infections over 1,848 person-years of follow-up; the HIV incidence rate was 0.92 per 100 person-years (95% CI 0.49 to 1.41; S5 Table). Among PrEP initiators in the same 8 communities, there were 11 incident HIV infections over 3,393 person-year of follow-up; the HIV incidence rate was 0.32 per 100 person-years (95% CI 0.15 to 0.53), corresponding to 74% lower HIV incidence among PrEP initiators compared to matched recent controls (adjusted incidence rate ratio [aIRR] 0.26, 95% CI 0.09 to 0.75; p = 0.013; Fig 3). Among women, HIV incidence among matched controls was 1.52 per 100 person-years (95% CI 0.70 to 2.36) compared to 0.40 per 100 person-years (95% CI 0.12 to 0.73) observed among PrEP initiators, corresponding to 76% lower HIV incidence among PrEP initiators compared to matched recent controls (aIRR 0.24, 95% CI 0.07 to 0.79; p = 0.019). Among men, HIV incidence among matched controls was 0.40 per 100 person-years (95% CI 0.10 to 0.90) compared to 0.24 per 100 person-years (95% CI 0.06 to 0.49) observed among PrEP initiators, corresponding to 40% lower HIV incidence that was not statistically significant among PrEP initiators compared to matched recent controls (aIRR 0.60, 95% CI 0.12 to 3.05; p = 0.54).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Observed HIV incidence rate among PrEP initiators compared to HIV incidence among recent matched controls prior to PrEP availability, overall and stratified by sex.

HIV incidence rate per 100 person-years in 8 study communities with propensity score–matched recent historical controls. CI, confidence interval; PrEP, pre-exposure prophylaxis.

https://doi.org/10.1371/journal.pmed.1003492.g003

Excluding 3 PrEP participants (all women) who seroconverted at the week 4 visit (and were thus possibly acutely infected at enrollment), HIV incidence was 0.31 per 100 person-years (95% CI 0.18 to 0.45) overall and 0.37 per 100 person-years (95% CI 0.19 to 0.59) among women. Observed incidence was 78% lower among PrEP initiators compared to matched recent controls (overall aIRR 0.22, 95% CI 0.07 to 0.67, p = 0.0078; among women: aIRR 0.17, 95% CI 0.05 to 0.59, p = 0.0054). Stored plasma was not collected at enrollment to determine whether these individuals were acutely infected at the time of PrEP initiation.

Demographics and adherence levels among participants with incident HIV infection

Among the 25 individuals with incident HIV infection following PrEP initiation, 17/25 (68%) were women (Table 2). Among the 17 women with incident HIV infection, 8 (47%) had a serodifferent partner at PrEP enrollment compared to 1 (13%) of 8 men. At the seroconversion visit, the median age among women was 27 years (range 20 to 62) and among men was 35 years (range 22 to 49). Regarding adherence, 18/25 (72%) participants reported not taking PrEP for >30 days before the seroconversion visit. One participant with incident HIV infection received TDF/3TC as PrEP once (of 10 fills); the remaining participants with incident HIV infection received only TDF/FTC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline characteristics of participants with incident HIV infection after PrEP initiation, overall and stratified by sex.

https://doi.org/10.1371/journal.pmed.1003492.t002

Of the 7/25 (28%) participants who reported taking at least 1 dose of PrEP in the last 30 days, 4 reported intermittent PrEP adherence in the last 3 months (2 had tenofovir levels in hair consistent with taking an average of 4 to 6 PrEP doses/week in the last 4 weeks; 1 had levels consistent with 7 doses/week; and 1 had no hair sample available). One participant had 2-class antiretroviral (ARV) drug resistance (described below) and high adherence (tenofovir levels in hair consistent with 7 doses/week). Three participants seroconverted at the week 4 visit and were possibly acutely infected at enrollment; one reported nonadherence to PrEP, while 2 reported adherence (confirmed by hair levels consistent with 7 doses/week).

Discussion

During population-level HIV testing of over 75,000 individuals in 3 regions across rural Kenya and Uganda, we offered universal PrEP access with flexible service delivery to persons at elevated risk of HIV acquisition. Of more than 15,000 individuals assessed to be at elevated HIV risk, over 5,400 participants started PrEP, and of these, 79% engaged in the PrEP program for follow-up visits. Over 7,150 person-years of follow-up, HIV incidence was 0.35 per 100 person-years among PrEP initiators. In 8 communities with propensity score–matched controls, HIV incidence was 74% lower among PrEP initiators compared to recent controls in the year before PrEP availability. These results provide supporting evidence that, in generalized epidemic settings, universal access to PrEP with flexible service delivery could reduce HIV incidence on top of universal test and treat (UTT) strategies [26]. Moreover, despite concerns from early placebo-controlled PrEP trials among women in which adherence to study product was too low for effectiveness [27,28], our study found lower HIV incidence among women who started open-label PrEP in real-world settings as compared to recent controls.

Our study offered PrEP in the context of universal HIV testing with supported linkage to either treatment or prevention based on HIV status [29]. We used an inclusive approach to define PrEP eligibility for HIV-uninfected individuals (based on serodifferent partnership, a risk score, or self-identified HIV risk). We offered same-day PrEP start (on-site at health fairs), an approach that was generally safe, as our group and others have reported [12,30,31]. We also offered a flexible PrEP delivery model with options for clinic- or community-based follow-up visits for HIV testing and PrEP refills. We posited that this lower-barrier delivery model would allow individuals to engage in taking PrEP who might not otherwise access health systems or facility-based HIV prevention services. Within this flexible model, we found that self-reported adherence was much higher among individuals reporting current HIV risk at follow-up visits compared to PrEP participants overall. As previously reported, among participants who self-reported PrEP adherence and current HIV risk, 66% had tenofovir concentrations in hair reflecting ≥4 PrEP doses per week and 44% had levels reflecting 7 doses/week [12]. Moreover, although many participants stopped PrEP at some point, ongoing program engagement provided opportunities for repeat HIV testing and restarting PrEP at each visit. Indeed, half of those who stopped PrEP later restarted taking it, suggesting that some participants may have been attempting to match PrEP use to periods of risk [32]. As PrEP programs extend eligibility criteria beyond specific risk groups to persons at substantial risk in the general population [21] and expand services outside of health facilities [33] (e.g., at drop-in centers [34]), our model and other low-barrier approaches to PrEP delivery could be adapted and scaled up.

We observed a lower HIV incidence among PrEP initiators as compared with historical controls prior to PrEP availability in the study communities. To our knowledge, our study is the first in sub-Saharan Africa to assess HIV incidence after population-level offer of PrEP, including to individuals outside of specific risk groups. Our results add to evidence from studies in serodifferent couples and younger women that access to PrEP was associated with lower than expected HIV incidence. In a demonstration study that offered both PrEP and ART to serodifferent couples in Kenya and Uganda from 2012 to 2015, HIV incidence was 0.2%, representing a 96% reduction compared to matched historical controls without PrEP [7]. The historical comparator group in the study dated up to 7 years prior and lacked access to ART (per guidelines at the time); thus, the reduction in incidence was due to both ART and PrEP. In another PrEP study that enrolled young women in urban southern Africa [8], HIV incidence was 1.0% compared to an expected incidence of 3.7% based on mathematical modeling [35]. Most recently, in a contraceptive trial among women that added PrEP to the current standard of HIV prevention (including counseling, condoms, and management of sexually transmitted infections) in South Africa, HIV incidence was approximately 50% lower overall (not only among PrEP initiators) after PrEP availability compared to before [36]. Our findings extend the evidence base on the effectiveness of open-label PrEP in sub-Saharan Africa and, to our knowledge, are among the first to demonstrate that offering PrEP at scale can have a substantial impact on HIV incidence.

Among women who initiated PrEP in our study, we found significantly lower HIV incidence compared to matched recent controls. Women remain a critical population for HIV prevention and are disproportionately impacted by HIV both in our study communities and across much of sub-Saharan Africa [1]. Indeed, although half of PrEP initiators in our study were women, they accounted for two-thirds of new infections. Recent population-based studies, including SEARCH, have shown greater reductions in incidence among men than women in the context of scale-up of ART [11,37–39]. Thus, primary prevention approaches, such as PrEP, may be important for reducing incidence among women. Questions have persisted about the effectiveness of PrEP in women in part because in 2 placebo-controlled PrEP trials that enrolled younger women not in stable serodifferent partnerships, adherence was too low to demonstrate efficacy [27,28]. However, PrEP was effective among women in mutually disclosed serodifferent partnerships in a placebo-controlled trial [2] and an open-label demonstration project [7] in East Africa, as well as among heterosexual men and women in a placebo-controlled trial in Botswana [40]. Our results add to the evidence that oral PrEP can lower HIV incidence among women in sub-Saharan Africa, including some of the first evidence among women who are not in serodifferent partnerships.

Our study also provides insights into the impact of PrEP on HIV incidence among men. We found that HIV incidence was 40% lower among men who initiated PrEP compared to historical controls, although this result did not reach statistical significance, possibly due to the small number of new infections in men. Although men have not been systematically prioritized for PrEP in generalized epidemic settings (apart from men in serodifferent couples or MSM), we found substantial interest in PrEP among men, who comprised half of PrEP initiators. One-third of new HIV infections in sub-Saharan Africa occur among men, and primary prevention options for this group, including male circumcision and PrEP, are critically needed for men’s own health [41]. In addition, men are an important link in transmission networks [42,43], and expanding their access to PrEP could accelerate declines in incidence.

The UTT trials demonstrated reductions in HIV incidence when universal HIV testing was combined with robust linkage and access to ART care [26,37,38]. These studies further showed that despite achieving high population-level viral suppression, UTT is not sufficient to reach HIV elimination targets. Our study provides evidence of the added impact of PrEP in communities that had already exceeded the UNAIDS 90-90-90 targets for HIV testing, ART coverage, and viral suppression following UTT. For countries that adopt a UTT approach in areas such as regional hotspots, leveraging the opportunity for HIV prevention, with inclusive eligibility and low-barrier delivery, is supported by our results.

Although we observed lower incidence among PrEP initiators compared to recent controls, PrEP uptake and engagement were lower in youth and mobile individuals, as we have previously described [12]. Thus, additional innovations are needed to reach these critical groups and provide tailored service delivery and adherence support for oral PrEP. Moreover, as on-demand or long-acting prevention modalities (such as the dapivirine vaginal ring [44,45] and injectable cabotegravir [46]) become available, expanding options for prevention may help to achieve further reductions in incidence.

Among the 25 participants who seroconverted after initiating PrEP, rapid ART start was acceptable and feasible and most had excellent virologic outcomes. Earlier PrEP trials in sub-Saharan Africa offered ART start based on CD4 threshold (per guidelines at the time); thus, prior studies characterizing seroconversions on PrEP have largely reported on HIV outcomes in the absence of ART [47–49]. Moreover, few studies have offered rapid ART start (now the standard of care for new HIV diagnoses [50]) after seroconversion on PrEP, and little is known about virologic outcomes in this group. In our study, we offered same-day ART start at the time of positive antibody testing and found that nearly 90% of participants started ART ≤7 days after the seroconversion visit. Among those with HIV RNA testing after ART initiation, 95% achieved HIV RNA <1,000 c/ml with standard first-line ART regimens. At PrEP follow-up visits, our study employed quarterly HIV testing using country-standard serial antibody-based algorithms. This approach may result in delayed detection of acute HIV infection, potentially increasing the risk of acquiring drug resistance while continuing PrEP and impacting virologic response to ART. To our knowledge, our study is among the first to report on virologic outcomes after ART start following seroconversion on oral PrEP in any setting globally [51]. While it is reassuring that most participants who seroconverted in our study achieved viral suppression on standard first-line ART regimens, robust HIV viral load monitoring (including rapid and, increasingly, point-of-care technologies [52]) and access to resistance testing are needed as PrEP use expands.

The only participant who did not achieve viral suppression on repeat testing after ART start was found to have 2-class drug resistance on testing of stored plasma from the seroconversion visit. This participant had evidence of transmitted drug resistance based on NNRTI mutations and NRTI thymidine analogue mutations (which confer low-level TDF resistance and are not commonly selected by TDF/FTC alone), as well as high-level FTC and 3TC resistance (M184V) that was either transmitted or acquired on PrEP. Based on objective adherence data, this participant had evidence of intermittent high-level adherence to PrEP (tenofovir concentrations in hair consistent with daily dosing at multiple visits, including at the seroconversion visit), separated by missed visits (including the 2 visits prior to seroconversion). There are several possible reasons for seroconversion with drug resistance in this case. First, the participant may have been nonadherent to PrEP during periods of risk, leading to seroconversion, followed by adherence to 2-drug PrEP after seroconversion, leading to acquisition of M184V on top of transmitted NRTI and NNRTI resistance. Second, the participant could have been nonadherent to PrEP and become infected with a multi-class drug resistant virus (with all mutations transmitted, including the M184V mutation). Third, if periods of adherence matched periods of risk, seroconversion could have been due to a breakthrough infection in the setting of high-level FTC resistance and low-level TDF resistance, rendering PrEP ineffective [53]. We do not have HIV genotyping data from the likely transmission partner to determine if drug resistance was transmitted or acquired in this case.

HIV acquisition in the setting of high levels of PrEP adherence has been described in several individuals previously; all were men from North America, Europe, or Asia [54–59]. Notably, several of these cases had evidence of resistance only to FTC (M184V) without mutations conferring TDF resistance, suggesting that PrEP breakthrough can occur with FTC resistance alone [54]. For the participant in our study, although breakthrough infection in the setting of high adherence is one possible explanation for seroconversion (and, to our knowledge, would be the first such case described in a woman), as noted above, infection also could have been due to nonadherence during periods of risk. As PrEP is scaled up globally, even with very high effectiveness, we are likely to observe more seroconversions among PrEP users. Further work is needed to systematically characterize reasons for seroconversion on PrEP through the use of pharmacologic adherence metrics (such as measuring PrEP drug levels in hair [60], dried blood spots [61], or plasma [62]) and drug resistance testing.

This study has several strengths. To our knowledge, our study is among the first and largest to assess HIV incidence after offering PrEP at a population level. Over three quarters of PrEP initiators were seen following PrEP initiation, providing among the most complete and longest follow-up (over 7,000 person-years) to date of any open-label PrEP study in sub-Saharan Africa. In addition, our estimate of HIV incidence among recent controls was based on robust population-level HIV testing data in 8 of the 16 study communities in the year prior to PrEP availability. We offered broad access to PrEP using a community-wide HIV testing model previously shown to be similar in cost to other mobile testing approaches [63] and feasible for implementation by community leaders [64].

This study also has limitations. First, we did not compare HIV incidence to a contemporaneous control where PrEP was unavailable. HIV incidence could have been declining in the study communities due to secular trends. Over the 3 years prior to PrEP availability in the study communities, high rates of population-level viral suppression (79%) were achieved, and population-level HIV incidence declined by 0.12 per 100 person-years [11]. Assuming that this trend continued during the PrEP study period, we would expect HIV incidence among matched controls to reduce from 0.92 per 100 person-years (observed 1 year prior to PrEP) to 0.85 per 100 person-years if their follow-up had been contemporaneous. Comparing HIV incidence among PrEP initiators (0.32 per 100 person-years) to this trend-adjusted incidence for matched controls (0.85 per 100 person-years) suggests that HIV incidence would have been 62% lower among PrEP initiators compared to matched contemporaneous controls. Second, our analyses of incidence relied on data from individuals with at least 2 HIV tests. However, repeat testing coverage was >75% among PrEP initiators and >80% among historical controls [11], and there were no notable differences between persons with and without a repeat test (S3 Table). Third, despite propensity score matching and adjustment for known risk groups, it is possible that the risk profiles of PrEP initiators with follow-up testing differed from the selected controls. If individuals using PrEP were at higher risk than their matched counterparts, the estimated change in HIV incidence would be biased toward the null. Finally, because we used antibody-based screening and did not collect plasma samples at PrEP enrollment, we are unable to determine whether 3 individuals who tested positive at the week 4 visit were acutely infected at the time of PrEP initiation. In sensitivity analyses removing these 3 individuals (all women), observed HIV incidence among PrEP initiators and changes in incidence compared to matched controls were similar to the primary results, overall and among women. A limitation of this sensitivity analysis is that the control group did not have repeat testing at week 4, and it is possible that individuals in the control group also had undetected acute infection at the time of baseline testing.

Community-wide HIV testing and universal access to PrEP were associated with lower HIV incidence among persons at elevated HIV risk who initiated PrEP in rural Kenya and Uganda compared to recent controls. Our findings included lower than expected HIV incidence after PrEP initiation among women, for whom declines in new infections have lagged behind men in recent prevention studies without PrEP. Our results suggest that universal access to HIV testing, treatment, and prevention, including rapid provision of PrEP with flexible service delivery, is a promising approach to reduce HIV incidence in generalized epidemic settings. Moving forward, combination approaches to prevention that include comprehensive HIV testing with linkage to oral PrEP, male circumcision, and ultimately, on-demand and long-acting modalities hold promise to further accelerate declines in HIV incidence globally.

Supporting information

Acknowledgments

The SEARCH study gratefully acknowledges the Ministries of Health of Kenya and Uganda, our research teams and administrative teams in San Francisco (California, USA), Kenya, and Uganda, collaborators and advisory boards, and especially all communities and participants involved.