Introduction

The African continent is disproportionately affected by the HIV pandemic. Despite representing only 12% of the world’s population, sub-Saharan Africa bears nearly 70% of the world’s cases of HIV [1] and the epidemic there is complex [2]. The greatest diversity of HIV subtypes is found across Africa, primarily subtypes A, C, D and their recombinants, and this varies by region [3]. Modes of transmission and affected populations are equally diverse, and can change over time. While the epidemic is primarily heterosexual [4], other key populations including men who have sex with men [5] may also contribute to HIV transmission regionally. Policy makers and clinical trialists depend on epidemiology data to guide effective prevention planning, service delivery [6, 7] and design and conduct of prevention trials [2, 8, 9].

HIV surveillance systems have often relied on cross sectional surveys of HIV infection [10]. While prevalence data may provide some indication of risk groups and can provide guidance for service delivery, they fail to identify recently infected persons. Though there is some progress being made on assays to estimate incidence from prevalent samples [11], these assays have yet to be proven valid, particularly in the African context [12, 13]. Furthermore, the increasing availability of antiretroviral therapy throughout Africa is allowing persons with HIV to live longer, making prevalence data even less relevant [14]. Research towards understanding and preventing HIV transmission should ideally focus on incident cases of HIV. Large-scale HIV incidence cohorts remain the gold standard for collecting these data.

Beginning in 1999 the International AIDS Vaccine Initiative (IAVI) undertook the funding and development of in-country laboratory and clinical capacity at nine research centers in Kenya, Uganda, Rwanda, Zambia and South Africa to support HIV vaccine trials. This work included observational epidemiology studies designed to recruit key populations suitable for trials, provide critical experience to the research teams, and inform the design of future HIV prevention trials. In 2004 this network began multiple, large-scale studies of HIV prevalence, incidence and early infection. Here we present the findings of these observational epidemiology studies.

Methods

Clinical Research Network

This work was conducted at nine sites in South Africa, Zambia, Kenya, Uganda and Rwanda (Fig. 1). All clinical research centers (CRC) are affiliated with existing institutions, locally staffed and most are led by local clinician-scientists who are recognized experts on HIV. To ensure that studies are conducted according to the International Conference on Harmonization Good Clinical Practices (ICH-GCP), GCP training is provided. HIV testing and counseling training and support are provided to counseling staff through a continuous improvement process. Specialized training addresses the needs of key populations such as men who have sex with men [15] or married/cohabiting couples [16, 17]. Study clinics are equipped to provide some care to volunteers (e.g., diagnosis and treatment of sexually transmitted diseases) and local health care resources available for referral have been assessed and strengthened where possible [18].

Download:

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

Figure 1. Map of the collaborating research centers of the IAVI Africa HIV Prevention Partnership.

MRC/UVRI: Medical Research Council/Uganda Virus Research Institute, IAVI: International AIDS Vaccine Initiative

https://doi.org/10.1371/journal.pone.0116100.g001

HIV incidence studies

Methods for each cohort have been published separately (Table 1). Briefly, each CRC performed community mapping, community-sensitization exercises, and/or prevalence studies, and based on the results and their own area(s) of expertise, each team selected risk screening criteria to enroll key populations. Teams reviewed enrollment criteria and HIV incidence in real time to confirm enrollment of true key populations. The cohorts included: 1) HIV negative partners of co-habiting, sexually active couples with an HIV infected, antiretroviral therapy (ART)-naïve partner (Zambia; Rwanda; Entebbe, Uganda) or irrespective of ART status (Masaka, Uganda); 2) sexually active sex workers and their clients (Kenya); 3) members of fishing communities aged 13–49 years and reporting at least one of the following in the past 3 months: more than one partner, a new partner, a sexually transmitted infection (STI), absence from home for more than 2 nights, regular sex partner who is HIV infected (Uganda; in January 2012, criteria were modified to ages 18–49 years, ART-naïve, report of alcohol/recreational drug use, report of unprotected sex; more than one partner was dropped); 4) Men who have sex with men (MSM; Kenya and Rustenburg); 4) younger participants recruited in Cape Town (≥16 years) and in the first fishing community study (Uganda, ≥13 years); and 5) geography as a screening criterion, as in the recruitment of adult members of several study villages (Masaka) and women living in peri-urban, economically disadvantaged regions with high HIV prevalence (Rustenburg).

Download:

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

Table 1. HIV prevalence and multivariate predictors of prevalent HIV infection in Masaka, Kilifi and Nairobi.

https://doi.org/10.1371/journal.pone.0116100.t001

Volunteers were followed quarterly or monthly. Cohort sizes were selected to provide a robust estimate of HIV incidence in light of the challenges of recruiting marginalized populations, and ranged from 250–1,000 volunteers. At every visit, volunteers were given HIV counseling and testing by both rapid test and p24 ELISA to detect HIV infection prior to antibody seroconversion (see references in Table 2). Rapid tests were typically performed by venipuncture and done in parallel. Condoms were provided, as was lubricant for the cohorts where anal sex was prevalent. Sexual behavior questionnaires were administered quarterly in Kenya, South Africa, Rwanda and Zambia; and every six months in Masaka. Cohorts with HIV incidence persistently <2% are not presented here (e.g., [19, 20]). Volunteers with incident HIV infection were invited to enroll in the early HIV infection study, 2006–2011, and had their CD4 T cell counts and viral load done at regular intervals. These data were used to inform referrals for antiretroviral therapy provided per the national guidelines. CRCs maintained strong relationships with local health care facilities to assure volunteers would be kept aware of and referred to as needed additional care and treatment as it became available (e.g., medical male circumcision) [21].

Download:

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

Table 2. Enrollment and HIV incidence among at-risk volunteers enrolled for HIV prevention preparatory studies stratified by CRC and, where appropriate, sex.

https://doi.org/10.1371/journal.pone.0116100.t002

Results

2a. HIV incidence in discordant couples

From 2002 to 2011, 1,856 couples discordant for HIV status were enrolled in Kigali (Table 2). HIV incidence was similar among women and men (p = 0.16). HIV incidence was nearly 2.5 times higher in the first three months of follow-up compared to subsequent visits for men (p = 0.003, Table 3), but for women this only achieved borderline significance. HIV incidence did not drop significantly with calendar year in Kigali (Table 3).

Download:

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

Table 3. HIV incidence stratified by duration of time in study follow up, and (where data were available) by study year.

https://doi.org/10.1371/journal.pone.0116100.t003

From 1995 (2004 in Ndola) to 2011, 3,800 HIV-discordant couples were enrolled at two CRCs in Zambia (Table 2). HIV incidence was similar across the two CRCs (p = 0.71) and was significantly higher in women compared to men in both Lusaka (p = 0.003) and Ndola (p = 0.01). HIV incidence in the first three months of study follow-up was significantly higher than in subsequent visits for women at both CRCs (Table 3), but the same was true only for men in Lusaka. With the exception of Lusaka women, HIV incidence did not vary significantly by year (Table 3), although a trend was also borderline significant among men in Ndola.

From 2006 through 2011, 1,574 couples were enrolled at both Ugandan CRCs, including 152 couples in Masaka who contributed follow-up while the HIV infected partner was on ART (Table 2). HIV incidence did not vary by sex (Table 2) but was significantly lower in Entebbe than in Masaka (p = 0.02). Masaka was the only CRC that enrolled discordant couples with the positive partner on ART, and HIV incidence was significantly lower in couples on ART compared to those not on ART (p = 0.009). In 2009, enrollment criteria changed in Masaka. There was no evidence of a drop in HIV incidence by time on study (Entebbe and Masaka) or by calendar year (Masaka) (Table 3).

2. Epidemiologic linkages in the early HIV infection cohort

From February 2006 through December 2011, 613 volunteers with incident HIV infection were enrolled. The cohort comprised 255 (41.5%) women and 359 (58.5%) men, of whom 92 (25.6%) were MSM; all but three from Kenya. Most volunteers (460, 75.0%) were enrolled from HIV-discordant couples cohorts, 219 (47.6%) were women.

For 407 volunteers, suspected transmitting partners were enrolled, 393 (96.6%) of whom had sample to test HIV transmission linkages. The proportion of volunteers linked by HIV sequence similarity to their suspected transmitting partner ranged from 67.1% in the Copperbelt (Kitwe and Ndola) to 100% in Entebbe but these differences did not quite achieve significance (Table 4, p = 0.08, Fisher’s exact test), and they did not vary by the sex of the incident case (p = 0.27). In Kigali, a non-significant trend suggested women with incident HIV were more commonly linked to their partners than were men (36/41 [87.8%] vs. 38/52 [73.1%], respectively, p = 0.08).

Download:

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

Table 4. Number of volunteers with HIV sequence data who enrolled with their suspected partner and were epidemiologically linked.

https://doi.org/10.1371/journal.pone.0116100.t004

Discussion

From 2004 to present, IAVI has coordinated HIV observational epidemiology studies at nine CRCs in preparation for HIV prevention trials. Laboratory, clinical, and data management procedures were standardized to ensure comparability across CRCs. Research protocols typically began with 1) cross-sectional HIV prevalence research, followed by 2) HIV incidence cohorts, and then 3) early HIV infection cohorts. 4,428 volunteers were initially enrolled in Kenya and Uganda, and HIV prevalence was observed to be higher than the national average, even in the general population survey in Masaka. Nine CRCs enrolled over 10,000 HIV-negative volunteers in prospective studies of HIV incidence, which ranged from 2% to 10% annually. This work included the first prospective studies of HIV incidence in African MSM [20, 26]. Over 600 volunteers with incident infection were enrolled from these cohorts into the early HIV infection study. Among those with an identified HIV-positive regular sexual partner, about 28% of HIV transmissions were not from that partner.

Prospective studies remain the gold standard for obtaining reliable HIV incidence data [6] needed for informed decision making, from national level program planning to clinical trial design. However, prospective incidence studies are expensive, time consuming, and require real-time review of data to maintain focus on a study cohort that reflects true risk of HIV infection. While a goal of the 2004 prevalence studies was to guide selection of key populations, some populations with elevated HIV prevalence had low HIV incidence. This included data from rural villages in Masaka district [19] where a previous HIV prevention trial had been conducted [27], heterosexual men in Kenya [20], FSW in Nairobi [20] and HIV discordant couples in Kilifi (no observed HIV incidence, unpublished data). In Masaka, the investigative team shifted focus to HIV discordant couples and, later, to residents of fishing communities. In Kenya the focus changed to MSM when men came forward asking to enroll as sex workers. Among MSM along the coast, the Kilifi team recorded HIV incidence as high as 35 cases/100PY among those who report sex exclusively with other men [26] highlighting both the success of careful epidemiology and the extreme need for intervention. Collateral benefits from our work in MSM and sex workers include the development of health care provider training modules to strengthen service delivery to these marginalized groups [15], and the improvement of our outreach and educational activities within the larger community. Following an attack on our Kilifi clinic that took place in 2010 [28] an emphasis was placed on engaging a wide range of public officials and public health stake holders, from religious leaders to government and police officials, and a community liaison officer has been hired whose role is to manage this process of outreach and education. Each CRC successfully identified key populations both suitable for HIV prevention trials and in need of specialized program planning to reduce HIV transmission [15–17].

Among those cohorts where higher risk was observed, we recorded heterogeneity associated with volunteer sex, time on study, and calendar year. Women are frequently cited as at significant risk for HIV acquisition in Africa [29], and in five of the seven cohorts with men and women the incidence point estimate reflected this (Table 2), although this achieved statistical significance only in Zambia. Of the more than a dozen cohorts presented in Table 3, five (men in discordant couples at Lusaka, Ndola, and Kigali; women in Ndola; and MSM in Kilifi) recorded a reduction in HIV incidence after the first study follow-up visit. While pre-seroconversion HIV infection at enrollment may artificially inflate HIV incidence at the first follow-up visit, we employed the p24 antigen assay at every study visit thereby excluding some volunteers at enrollment prior to antibody seroconversion. Nine cohorts in table 3 had adequate follow-up time to consider HIV incidence as a function of calendar year. Only female members of DCs in Lusaka showed a statistically significant reduction in HIV incidence. Of the 16 cohorts reported in Table 3, 5 demonstrate a significant drop in HIV incidence, 9 show a non-significant drop in the incidence point estimate, 1 shows no change, and 1 shows a non-significant increase in the point estimate. We feel the most likely reason for the drop in HIV incidence was the regular counseling and testing and prevention messages, however, some cohorts were simply not powered to detect this drop (e.g., MSM in Nairobi, women in Cape Town and Rustenburg saw a non-significant drop of ~3 cases per 100 person years. This would otherwise represent a very large drop, but in our case, it remains non-significant in a 200–300 person cohort). Additionally, the different stages of the epidemic in different countries and cohorts may explain why we observe no significant changes in incidence, as perhaps some mature epidemics may be less susceptible to intervention. Cohort recruitment may change subtly over time as new clients seek HIV testing at the CRCs. Secular trends in the source populations such as increased awareness of HIV prevention techniques may also be prevalent. Counseling and prevention messages improve as health care providers get to know their stakeholders better. Treatment and care changes as antiretroviral therapy guidelines change, affecting risk of HIV acquisition in certain populations (e.g., HPTN 052 and the recommendation that ART be initiated in HIV-discordant couples). National level program planning requires incidence trends over time to best assess whether interventions are working or not. Clinical trial design (particularly sample size) should always factor in anticipated decline in incidence during the time of study as a result of national program interventions and that of the study itself. Taken as a whole, these data suggest that in our cohorts HIV incidence tends to be lower in men than women, declines with time on study, and may drop with calendar time. It is with caution that one should generalize findings in research cohorts to the broader community; however this is similar to overall trends in the African epidemic [2, 29].

Volunteers with incident HIV infection from our discordant couple cohorts provide an interesting lesson on risk that applies both to the clinical trialist and the public policy maker. Some of the highest HIV incidence was observed in these cohorts, with incidence estimates up to 10 cases /100PY. The HIV incidence among discordant couples who are not receiving quarterly behavioral counseling interventions is likely to be higher—possibly as high as 20–25% annually [30, 31]. Following the results of HPTN 052, which demonstrated a 96% reduction of HIV transmission within discordant couples when the volunteer with HIV infection received ART early [32], considerable attention has been focused on “test and treat” (i.e., putting newly diagnosed HIV-positive persons on ART as quickly as possible) as a means to reduce HIV transmissions [33, 34], and discordant couples represent a priority [35]. While others point out that real-world adherence, funding, and logistics all provide significant hurdles to this lofty goal [36], our data highlight another challenge: up to 30% of infections in our discordant couple cohorts originated from a partner other than the HIV infected steady partner. Indeed, in HPTN 052, 7 (17.9%) of the 39 transmission events were linked to someone other than the volunteer’s enrolled HIV-positive partner [32]. Although “test and treat” may eventually be a practical reality for much of the world’s population living with HIV, until that time discordant couples remain an important key population for public health interventions and for study.

We enrolled thousands of volunteers at risk for HIV infection, hundreds of volunteers with incident HIV infection, and identified key populations in need of public health interventions who might be suitable to participate in the next generation of HIV prevention trials. This work has spearheaded improvements to the health care delivery for marginalized groups such as sex workers and MSM [15] and reinforced the importance of testing and counseling couples together [30, 37, 38]. Specimens from these studies have also been the basis for dozens of ongoing collaborations ranging from estimating the seroprevalence of potential vaccine vectors [39] to characterizing the transmitted virus [40]and the host immune system following infection [41, 42]. IAVI has also tested 13 HIV vaccine products in 25 early-phase clinical trials in Africa [43–47], conducted a multi-site study to establish adult safety laboratory reference ranges against which to characterize adverse events in these clinical trials [48], and enrolled over 1,800 asymptomatic volunteers with prevalent HIV infection in a search for broadly neutralizing antibodies [49]. Collateral benefits extend beyond these CRCs as GCLP and assay-specific training are made available to laboratory technicians outside this partnership.

These large prospective observational epidemiology studies provide valuable data for prevention trial design and conduct, prevention planning, and service delivery. Over the past nine years, IAVI has identified and enrolled thousands of volunteers from key populations, observing a high HIV incidence despite regular HIV testing and counseling. The benefits to the public health, scientific and local communities are tremendous.

Acknowledgments

Assembling a research network of this size and scope is tremendously challenging and the authors thank all the staff, investigators, community members and study volunteers who participated. We are grateful to Family Health International which supported some of the cohort study work in Rustenburg [25] and to the European and Developing Countries Clinical Trials Partnership that helped fund some of the cohorts in the Ugandan fishing communities [50]. We also thank the staff at the Perinatal HIV Research Unit, Contract Laboratory Services, and IAVI’s Human Immunology Laboratory who worked very hard to assure these studies progressed efficiently and effectively. We also thank Wasima Rida for her help with advice on and review of the statistical analysis. This report was published with permission from KEMRI.

These data have been presented in part at The South African AIDS Conference in Durban, South Africa, 18–21 June 2013 (HIV incidence data).

The IAVI Africa HIV Prevention Partnership includes the following persons who contributed to the studies presented herein:

Kigali, Rwanda (Project San Francisco): Etienne Karita, Principal Investigator; Susan Allen, Principal Investigator; Roger Bayingana, Investigator; Kayitesi Kayitenkore, Investigator

Nairobi, Kenya (Kenya AIDS Vaccine Initiative): Omu Anzala, Principal Investigator; Gaudensia Mutua, Principal Investigator

Kilifi, Kenya (Center for Geographic Medicine Research—Coast & Kenya Medical Research Institute): Eduard J. Sanders, Principal Investigator; Peter Mugo, Investigator

Medical Research Council (MRC)/Uganda Virus Research Institute (UVRI) Uganda Research Unit on AIDS, Entebbe, Uganda: Anatoli Kamali, Principal Investigator; Rogers Twesigye, Study Coordinator, John Byabagambi, Study Physician; Florence Babirye, Nurse

Medical Research Council (MRC)/Uganda Virus Research Institute (UVRI) Uganda Research Unit on AIDS, Masaka, Uganda: Anatoli Kamali, Principal Investigator; Eugene Ruzagira, Study Coordinator, Agnes Bwanika, Ubaldo Bahemuka, and Freddie Mukasa Kibengo, Study Physicians; Peter Hughes, Laboratory Manager; Vincent Basajja, Community Liaison Officer

Lusaka, Ndola and Kitwe, Zambia (Zambia Emory HIV Research Project): William Kilembe, Principal Investigator; Susan Allen, Principal Investigator; Shabir Lakhi, Principal Investigator; Mubiano Inambao, investigator;

Rustenburg, South Africa (Aurum Institute): Mary H. Latka, Principal Investigator; Gavin J Churchyard, Investigator; Petra I Kruger, Investigator; Heeran Makkan, Study Coordinator; Candice M Chetty-Makkan, Study Coordinator; Ben Makhoana, Community Liaison Officer; Tiro Dinake, Nurse; Matsidi Malefo, Senior Research Assistant; Ireen Mosweu, Research Assistant

Cape Town, South Africa (Desmond Tutu HIV Foundation): Linda-Gail Bekker, Principal Investigator; Keren Middelkoop, Investigator; Surita Roux, Investigator