Creating an African HIV Clinical Research and Prevention Trials Network: HIV Prevalence, Incidence and Transmission

HIV epidemiology informs prevention trial design and program planning. Nine clinical research centers (CRC) in sub-Saharan Africa conducted HIV observational epidemiology studies in populations at risk for HIV infection as part of an HIV prevention and vaccine trial network. Annual HIV incidence ranged from below 2% to above 10% and varied by CRC and risk group, with rates above 5% observed in Zambian men in an HIV-discordant relationship, Ugandan men from Lake Victoria fishing communities, men who have sex with men, and several cohorts of women. HIV incidence tended to fall after the first three months in the study and over calendar time. Among suspected transmission pairs, 28% of HIV infections were not from the reported partner. Volunteers with high incidence were successfully identified and enrolled into large scale cohort studies. Over a quarter of new cases in couples acquired infection from persons other than the suspected transmitting partner.


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.

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].

Laboratory facilities
All laboratories are equipped to perform tests for HIV antibody (rapid test for HIV, HIV ELISA, and p24 ELISA), syphilis and pregnancy. To assure assay comparability across CRC, quality assurance and control (QA/QC) is performed by Clinical Laboratory Services in Johannesburg, South Africa. Laboratory personnel are trained in GCLP, and 8 of the nine research centers in the network have now received GCLP accreditation through Qualogy (www. qualogy.co.uk). The Rustenburg CRC did not adopt PBMC processing facilities, instead sending sample to Johannesburg for processing, and therefore never qualified for GCLP accreditation. Three labs were accredited in 2007 (Nairobi, Kilifi

HIV prevalence studies
Three CRCs (Nairobi, Masaka and Kilifi) enrolled 1,000-2,000 study volunteers, 18 years or older, for a single visit to characterize local HIV prevalence and risk factors. HIV VCT employed two rapid tests (typically Determine and Unigold; with tiebreakers Murex in Masaka, Vironostika in Nairobi, and Instascreen in Kilifi). A brief, site-specific, demographic and risk assessment questionnaire was administered. In Nairobi, volunteers included sex workers and their clients. In Masaka, the goal was to obtain a population-based estimate of prevalence and therefore recruitment included all residents of three rural communities, recruited and provided with HIV counseling and results at their homes. Home based counseling included pre-test counseling and sample draw in the home, with delivery of results within 1-4 weeks. Subsequent testing and counseling (in the incidence studies, below) was done in a local research clinic. In Kilifi, study volunteers were recruited from three populations: 1) randomly selected residents of the Kilifi Demographic Surveillance System (DSS), 2) walk-in VCT clients at the study site, and 3) the referred sexual partners of DSS and walk-in volunteers.

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     In Kangemi, all volunteers were sexually active, but not all volunteers were sexually active in the past 7 days doi:10.1371/journal.pone.0116100.t001 (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). 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].

Early HIV infection study
The Early HIV Infection Study has been described previously [22,23]. Briefly, volunteers identified with incident HIV infection were invited to enroll as early as possible (typically within 2 months of their estimated date of infection (EDI)) and follow-up is ongoing. Viral subtype was determined by sequencing the pol region of HIV from an early sample [22]. Suspected transmitting partners were invited to enroll, and molecular epidemiologic linkages were determined by sequence comparison of pol, gag, gp120, gp41, and/or long terminal repeat regions [24].

Data collection and management
Data from the HIV prevalence studies were managed by CRC staff. HIV incidence and early infection study data were managed by the Perinatal HIV Research Unit (PHRU) in Johannesburg, South Africa, except the incidence studies in Rwanda, Zambia, and some in Uganda, which were managed by the CRC staff. Data were transcribed onto a DataFax (Clinical DataFax Systems, Inc., Ontario, Canada) form and digitally faxed to a central server; manual and automatic queries were received then resolved by the CRC data management team. An epidemiologist reviewed the data prior to analysis.

Data analysis
Inferential statistics include chi square or Wilcoxon rank-sum as appropriate. Predictors of prevalent HIV were evaluated separately for different key populations. Covariates were selected to remain in the multivariable model based on a p-value<0.05. Modeling results are shown as p-values, prevalence ratios, and 95% confidence intervals. Comparisons of HIV incidence across CRC or by gender were done by Cox Proportional Hazards modeling; analyses comparing HIV incidence across time was done using the Stata incidence rate ratio calculator. Only those cohorts with recruitment spread over > 3 years and with adequate follow-up (>400 PY) were analyzed by year. Estimates that vary from published results are due to additional follow-

HIV prevalence and associated risk factors in Masaka, Nairobi and Kilifi
From January to December 2004, 4,428 volunteers were enrolled at three CRCs in a cross-sectional study of HIV prevalence. Prevalence varied from 8% to 16%; women and older volunteers had a significantly higher HIV prevalence at each CRC (Table 1). Marital status was associated with HIV infection only at Kilifi, with divorced/widowed and polygamously married volunteers two to three times more likely to be HIV infected than volunteers who had never been married. In Masaka, residing in the community situated along the Kampala-Kigali trans-African highway was associated with a 50% higher HIV prevalence compared to two villages off the highway. In Kilifi, residents of urban areas were four times as likely to be HIV infected as were rural volunteers. In Masaka, inconsistent condom use relative to never using a condom was associated with HIV, and in Nairobi report of ever having used a condom was associated with HIV infection. Report of genital ulcers was associated with a 1.5 to 2.5-fold increase in HIV in Masaka and Kilifi. In Nairobi, marijuana use was associated with a two-fold increased likelihood of HIV infection. In Masaka, report of a single, steady partner was inversely associated with HIV infection relative to report of no sexual partners in the past 3 months.  (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).
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).

2b. HIV incidence in men who have sex with men
In August 2005, the Kilifi, Kenya, CRC enrolled the first MSM into their HIV incidence cohort study. Through December 2012, 726 MSM have been enrolled in Kilifi and enrollment is ongoing; from January 2006-October 2009, 303 were enrolled in Nairobi; and from April 2009-November 2010, 29 in Rustenburg, South Africa. HIV incidence has been very high in these cohorts, from 5-10 cases/100 person-years (Table 2). Incidence dropped three-fold after the first three months on-study in Kilifi (Table 3, p<0.001); no similar drop was observed in Nairobi. No drop was observed in HIV incidence by calendar time (Kilifi).

2c. HIV incidence in Ugandan fishing communities
Between February and August 2009, 1,000 volunteers were enrolled from five fishing communities near Entebbe and Masaka, Uganda, into an 18-month prospective study. The overall HIV incidence was 4.9 cases/100 PY and did not vary by sex (Table 2) or by recruitment area. No evidence for a reduction in HIV incidence with time on study was observed (Table 3). Relatively low HIV incidence was observed among adolescents (ages 13-17: 2.8 cases/100 PY), however few enrolled (n = 26). Using updated screening criteria based on these results, a further 529 volunteers were enrolled in the Masaka district from January 2012; follow-up is ongoing and HIV incidence is high (6.7 cases/100 PY), particularly among women ( Table 2).

2d. HIV incidence in other key populations
Other cohorts include female sex workers (FSW) in Kilifi [20], younger women in Cape Town [20], and women enrolled from the community and local clinics in Rustenburg [25] (Table 2). In 2010, the Rustenburg CRC shifted recruitment strategies into poorer, peripheral communities where HIV prevalence was higher. Eligibility criteria were also changed to include any woman reporting sexual activity or an STI. Over 300 women enrolled in Rustenburg with these eligibility criteria had an HIV incidence of nearly 9 cases/100PY ( Table 2).   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 (

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 realtime 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][16][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 preseroconversion 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 nonsignificant 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][44][45][46][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.