The epidemiology of HIV and prevention needs among men who have sex with men in Africa

Men who have sex with men (MSM) in Kenya are at high risk for HIV and may experience prejudiced treatment in health settings due to stigma. An on‐line computer‐facilitated MSM sensitivity programme was conducted to educate healthcare workers (HCWs) about the health issues and needs of MSM patients.


SERVICES BARRIERS
The HIV epidemic in sub-Saharan Africa (SSA) is dynamic with regional and temporal variation. The 2012 Report on the Global AIDS Epidemic reports a decline in HIV incidence among the general population (GP) by 25%, a decrease by 40% in HIV-related mortality, and that more than half of people living with HIV (PLHIV) who are eligible for treatment were on treatment in SSA [1]. The UN classifies countries as low level when HIV prevalence in the GP, as measured by HIV surveillance data, is under 1% and key population (KP) prevalence does not exceed 5%, where KPs are defined as female sex workers (FSWs), men who have sex with men (MSM) and people who inject drugs (PWID); concentrated epidemics are where HIV prevalence is under 1% in the GP, but any KP (e.g., FSW) prevalence consistently exceeds 5%; and generalized epidemics are where HIV prevalence exceeds 1% in the GP regardless of HIV status among KPs [1,2].
An epidemic appraisal proposed by Wilson and Halperin and others that characterizes the typology of HIV transmission within the country, rather than simply crude estimates of incidence and prevalence at the national level, called for reclassification of countries based on transmission dynamics [3,4]. They posited that concentrated epidemics are driven by KPs, and they added mixed epidemic settings, where both the GP and KPs play a role in HIV transmission, and generalized epidemics, in which they argued that contribution to new infections from KPs is insignificant. But this is debatable.
UNAIDS estimates HIV prevalence to be 17.9% among MSM in SSA. Yet there are limited epidemiological data for MSM or KPs in general in this region. For programmes to be well aligned, we need to better understand the epidemiologic and behavioural burden and social drivers of HIV within KP groups in all epidemic settings. The articles in this series focus on building the literature on epidemiology, social drivers of transmission and programmatic innovations for MSM in four regions of SSA: West, Central, East and Southern Africa. Respondent-driven sampling (RDS) methods were used to engage networks of MSM in these pioneering studies, which characterize the HIV epidemiology among MSM in Cameroon, Senegal, Malawi and Swaziland as elevated and sustained when compared to men in the GP. This supplement also includes a meta-analysis of prevalence studies from KPs in Central and West Africa, which helps to ground our understanding of MSM within the broader context of KPs.
The extraordinary burden of internalized and external stigma and discrimination creates the paralyzing barriers to MSM access for prevention, care and treatment services that are notable across these studies. Key dynamics of intersecting stigmas of HIV and sexual orientation among HIV-positive MSM are also explored in Swaziland [5]. Structural interventions to address the pervasive stigma in healthcare settings have been called for by the public health community, and this supplement also addresses this issue with findings from sensitivity trainings in coastal Kenya, which can influence and support clinical work with MSM who come to healthcare settings for sexually transmitted infection (STI) and HIVrelated care and treatment. Additionally, innovative strategies for community mobilization engaging peer leaders in small-group safe spaces within stigmatized peri-urban townships are addressed through the article from South Africa.
While the role of sex among men is increasingly described in concentrated epidemic settings, studies from Southern Africa within generalized epidemics, where KPs are conventionally not thought to play a significant role (e.g., South Africa, Swaziland, Lesotho, Malawi, Namibia, Botswana and Zimbabwe), have also shown MSM to have high prevalence of HIV, syphilis and hepatitis B virus, with disease burdens equal to or greater than those of men in the GP. Still, because of the conception of KPs as insignificant in these generalized epidemic settings, any data collection or targeted response is limited [6].
Swaziland has been documented to have the highest HIV prevalence globally. The incidence of HIV appears to have peaked in 1998Á1999 at 4.6% according to UNAIDS estimates, while in 2009 it was estimated to be 2.7%. Recent data from the Swaziland HIV Incidence Measurement Survey (SHIMS) estimated HIV incidence at 2.4% in the total population: 3.1% among women and 1.7% among men [1,7,8]. The 2009 Swaziland Modes of Transmission (MOT) study characterized major drivers of incident HIV infections to be multiple concurrent partnerships before and during marriage as well as low levels of male circumcision [9]. While these drivers were validated through the SHIMS, it is critical to note that like many MOT studies in generalized settings in Africa, there was no known prevalence among FSWs or MSM, so the MOT analysis assumed a low frequency of both practices and therefore assigned them as minor drivers of the epidemic.
In this issue, Baral and colleagues conducted the first cross-sectional study to estimate HIV prevalence and its risk factors among MSM in Swaziland [10]. The HIV prevalence in the RDS sample of 324 MSM was 17.6%, and the odds of HIV prevalence increased by 20% for each year of age. The vast majority (70%) of the sample reported being unaware of their HIV status, and consistent condom use with lubricants with male partners was reported by 12.6% of respondents. The authors note that within their MSM sample, HIV prevalence is consistent with that of an age-matched sample from the GP until age 24Á26 years, when the prevalence of HIV among MSM rises higher than that of other men in the GP Á with HIV prevalence of 43.1% among MSM older than 27 years [10].
These data as well as other recent data showing alarming rates (70.3%) among FSWs have called researchers to rethink the prevention, care and treatment response in Swaziland [11]. Similarly, in Malawi, a high-HIV-burden country in East Africa, Wirtz et al. conducted the most comprehensive MSM study to date in Malawi, where the HIV response has largely focussed exclusively on heterosexual and vertical transmission of HIV and where an estimated 8% of GP men have HIV [12,13]. A sample of 338 Malawi MSM had prevalence of HIV and active syphilis of 12.5 and 4.4%, respectively, after adjusting for RDS approaches. Ninety percent of HIV infections were previously undiagnosed, and about half reported consistent condom use with casual male partners. Among MSM 26 years and older, prevalence of HIV was 28.1%.
West and Central Africa, the most populous regions of Africa, have a mixture of HIV epidemics, and KPs are better understood to play an important role in the overall transmission dynamics in Nigeria, Senegal and Burkina Faso, where KPs consistently show elevated HIV prevalence in comparison to the GP, as reported by Papworth et al. in a meta-analysis of KPs from West and Central Africa [14]. In Cameroon, where the GP HIV prevalence for men is 2.9%, Park and colleagues sampled 511 MSM in Douala and Yaoundé through RDS and estimated the HIV prevalence to be 37% [15]. In Douala and Yaoundé, respectively, HIV prevalence was 25.5 and 44.4% [16]. Like in the other studies, the sample was young, with a median age of 24 years, and as age increased, HIV rates increased at staggering levels. HIV among MSM aged 24Á29 years was 47%, and it was 49.4% for those over 30. About half of respondents did not use condoms consistently with casual partners (48.5%), and even more did not use them consistently with regular partners (64.1%). Similarly striking in Senegal, where GP prevalence among reproductive age men is 0.5%, the article included by Drame et al. reports baseline HIV prevalence among a sample of MSM at 36.0% (43/114) with cumulative HIV prevalence after 15 months at 47.2% (51/ 108) [17].
Due to the criminalized nature of male-to-male sex in all countries where studies from this issue took place, with the notable exception of South Africa (which still experiences high stigma), MSM are often afraid to visit healthcare services; and when they do go, they are reluctant to disclose their sexual histories to healthcare providers for fear of rejection, derision, or other negative reactions. Continued work is needed to develop violence screening, reporting and mitigation approaches within these settings [18]. The authors in this issue demonstrate that access to and coverage of quality HIV services for MSM are still marginal and not sufficient to reverse the epidemic's trend among MSM, and this is aggravated by several factors, primarily stigma, discrimination and limited domestic investment in programmes focussed on MSM [19Á21]. Lessons learned from mature programmes targeting FSWs in the region show that with limited coverage, poor dosage and an inadequate combination of approaches, even proven interventions will be ineffective [22].
Research focussing on MSM sexual risk behaviour in SSA is scant, and even less is known about the stigmatization and discrimination of HIV-positive MSM [23]. Kennedy et al. conducted 40 in-depth interviews from 20 HIV-positive MSM, 16 interviews with key informants and three focus group discussions with MSM community members. Internalized and experienced stigma was high among the men living with HIV, who report that they conceal their HIV status from others. MSM living with HIV reported experiencing greater social isolation and lack of support for care-seeking and medication adherence. Perceived and experienced stigma from healthcare settings led to delayed care seeking and travel to more distant clinics to retain anonymity at home.
The authors argue that mental health interventions, training for healthcare providers and better protection against discrimination are needed for Swazi MSM living with HIV, which corroborates previous findings from South Africa [24].
The association between HIV prevalence and the existence of community-based HIV interventions targeting MSM and other KPs has been well described [25]. Tailored communitybased programmes that provide MSM with high-quality, sensitive services that are socially and economically acceptable and led by the beneficiaries themselves is a promising approach. The current article by Batist et al. expands on this association with the use of safe spaces to remove barriers to service access, including those aimed at training providers and mobilizing communities even within stigmatized peri-urban settings, which led to greater feelings of connection, social support and self-esteem among MSM community members and also led to these spaces becoming distribution points for condoms, lubricants and HIV education.
Also in this supplement, two articles look at a sensitivity training for healthcare workers providing services to MSM [21,26], highlighting the pre-existing attitudes that can manifest during clinical encounters with MSM. Healthcare workers in SSA generally do not receive specific training in working with MSM or other KPs, and they may not be aware of risk factors for HIV transmission or appropriate care and treatment needs. Healthcare worker training has been identified as a priority intervention to support a minimum package of essential services for MSM [27]. In van der Elst et al., the researchers implemented a novel approach to sensitivity training for healthcare workers providing services to MSM [21]. The training consisted of self-directed, publically available online modules followed by group discussions focussed on MSM sexual risks and healthcare needs. Knowledge and homophobia were assessed prior to training, immediately after training and three months post training. There was a statistically significant decline in homophobia sustained after three months post training, with greater reductions for males and those in clinical roles (doctors and nurses), who were also more likely to have higher homophobia scores pre-training. However, it remains to be seen whether these attitudes can be maintained over time without ongoing support [21].
In a subsequent article in this supplement, van der Elst et al. explored topics including the sexual identification of subcategories of MSM, sexual practices and risks for HIV and STI transmission, practices for sexual history taking and sexual health examinations for MSM [26]. Stigma was also a concern for healthcare workers, such as negative judgements from peers or community members for being associated with MSM, and was an ongoing challenge after the training. After completing the programme, healthcare workers expressed greater acknowledgement of MSM patients in their clinics, empowerment to address their needs, and a better understanding of the biological, behavioural and social influences that lead to HIV and STI risk for MSM.
The term ''MSM'' is meant to address all MSM, regardless of their gender or sexual identities. Some MSM self-identify as heterosexual rather than gay, homosexual, or bisexual, especially if they also have sex with women, are married, only take the penetrative role in anal sex, or have sex with men for money or convenience [28]. They may not consider their sexual encounters with other men in terms of gender identity or sexual orientation, or they may more aptly self-identify using local social terms which reference sexual identities, masculinity and femininity, and behaviours. One noteworthy finding within the articles presented in this issue from Malawi and Swaziland was the disconnection between gender identity and sexual orientation [10,12,20]. While nearly all the respondents of both surveys reported that they were either gay or bisexual and had anal sex with men as criteria for eligibility in the study, a sizeable number reported that they were not male. In Malawi, 17.0% reported they were female, and another 2.8% said they were transgender. In Swaziland, 15.7% reported being female, and 1.8% said they were both male and female. It is not clear whether participants actually considered themselves to be women or whether their sexual behaviour caused them to consider themselves not to be men. There is a need for further study to better understand how these terminologies translate into risk and sexual identity profiles while not singling out these individuals for further stigma.
As more data become available for MSM in SSA, including in Southern African generalized settings, MSM needs should be identified and addressed throughout the continuum of HIV prevention, care and treatment. We continue to need evidence-based interventions to identify, create and train healthcare providers as well as community champions, including lawyers, media owners, journalists and religious leaders sensitive to MSM programming, and establish communitydriven programmes while expanding integration within the health system as appropriate.
The articles in this series have shown that throughout SSA, there is a significant and sustained epidemic among MSM, fear of discrimination from healthcare settings, and providerbased and self-stigma which impede prevention, care and treatment [29Á31]. The findings highlight the need to focus on MSM as a critical KP in ''mainstream'' approaches as well as MSM-targeted models. Countries in Africa characterized as having generalized epidemics, where KPs are not considered relevant, must be re-conceptualized based on these findings. All ''generalized'' epidemics are in reality mixed epidemics with ongoing transmission among KPs, and this becomes increasingly clear as GP prevalence rates decline while MSM experience expanding epidemics.
While it is not clear what proportion of new HIV infections are linked to MSM directly and via second-order transmission among their partners, it is clear that without addressing this underserved, stigmatized population, HIV transmission will be impossible to abate. Therefore, the benefits of targeted structural, behavioural and biomedical services for MSM go beyond the individuals to benefit the larger public welfare and security of all countries within SSA as well as globally. After three decades of the fight against HIV, plans to end the HIV epidemic through goals such as the AIDS-free generation and the US President's Emergency Plan for AIDS Relief's (PEPFAR) Blueprint seem possible. Substantial progress has been made, and more will come through vigilance, courage, tolerance and commitment.

Introduction
Sub-Saharan Africa has a very high burden of HIV-1 infection, of which a substantial proportion occurs among populations reporting high-risk sexual behaviour such as transactional sex and anal intercourse [1]. Such populations suffer from stigma and rejection, and they have been neglected by many HIV prevention and care programmes [2]. As a result, most African healthcare workers (HCWs) have not been informed about the risk of HIV transmission with regard to heterosexual or homosexual anal sex. In addition, African HCWs may lack understanding of the many challenges that men who have sex with men (MSM) and other key populations face in healthcare facilities [3].
Societal discrimination on the grounds of sexual orientation has been reported frequently among African MSM, taking the form of sexual, physical and verbal assault [4Á7], and a number of studies have demonstrated an association between reported experience of discrimination and HIV risk or risk behaviour [8]. Similarly, high levels of internalized homophobia among MSM have been reported in Nigeria [8], South Africa [9,10] and Uganda [11], known to be associated with individual HIV risk-taking behaviour [12]. Overt stigmatization specifically from HCWs in the context of HIV testing and care, such as denial of service [4,3,13] and harassment in clinic spaces [14], has been reported as a key element of perceived discrimination, presenting a deterrent to service access [15] or accurate disclosure of behavioural risk [14]. In the absence of resources targeted to groups at high risk of HIV infection, the marginalization of MSM from public HIV prevention and treatment resources can only hamper the effectiveness of national HIV control efforts [6].
Health worker training, social mobilization and community engagement were prioritized as structural interventions in a recent consultation on priority areas for MSM HIV prevention research involving 69 participants from 17 African countries [16]. HCWs have also been called to action to reduce stigma and discrimination, provide integrated services for mental health concerns and substance use, screen MSM routinely for HIV and sexually transmitted infections (STIs) and ensure training for all personnel in clinical settings [3]. As yet, African HCWs lack any evidence-based, culturally adapted training model that is sensitive to MSM needs. This problem likely stems from cultural taboos about anal sex practices, even in opposite-sex couples [17], and strong political, religious and public prejudice against same-sex practices [18].
Since 2005, biomedical research has been ongoing with both HIV-1 negative MSM and MSM living with HIV in coastal Kenya [1,19]. To date, the only incidence data for African MSM derive from our cohort and a related cohort in Nairobi [19,20]. Overall HIV-1 incidence among young MSM in coastal Kenya was as high as 8.6 (95% confidence interval [CI]: 6.7Á11.0) per 100 person-years of observation [19]. The majority of these MSM reported sex work, and large numbers of such men have been identified in coastal Kenya [21]. Similarly, our cohort study of MSM living with HIV showed that 40% had less than 95% antiretroviral therapy (ART) adherence, compared to 29% of heterosexual men and 12% of women who were followed in the same research setting [1]. These findings prompted us to brief health authorities and develop materials to help improve care for MSM in Kenya and elsewhere in Africa.
Internet-based learning (e-learning) as a cognitive tool has increasingly been used in health professions in resourceconstrained low-and middle-income countries [22]. E-learning technologies offer learners control over content, learning sequence, pace of learning, time and often media, allowing learners to tailor their experiences to meet their personal learning objectives [23]. The internet-based HCW MSM sensitivity training described here represents our attempt to deploy meaningful, clinically relevant material to meet this specific learning need within Kenyan HIV services through adaptation of an existing training curriculum to a web environment.
''MSM: An introductory guide for health workers in Africa'' is a paper-based HCW sensitization training first developed in 2010. The content of training was validated and revised through a programme of extensive classroom use in South Africa, and following expert review [24]. The paperbased training guide was electronically converted to a selfdirected electronic format and published online in July 2011, a version of which was adapted for use in this study.
The objectives of this study were (1) to assess the feasibility of facilitated self-directed learning of MSM health issues in coastal Kenya and (2) to evaluate the effect of the training intervention upon HCW knowledge and attitudes.

Study site and participants
Seventy-four HCWs involved in HIV prevention, treatment and care services in coastal Kenya were recruited to participate in the study. We mapped 54 ART-providing governmental and nongovernmental health providers in four districts in coastal Kenya (Kilindini, Mombasa, Kilifi and Malindi). An average of two staff representatives from each health-providing facility were invited to the training intervention, including clinicians and counsellors as well as clinic administrators.
Four ''district AIDS/STD coordinators'' (DASCOs) working within the study districts were trained to lead the MSM sensitization training during a 2-day ''training-of-trainers'' course similar to the training proper. An additional day was used to prepare focus-group topic guides and organizational matters. The study procedures were approved by the ethical review board at the Kenya Medical Research Institute, and all participants provided written informed consent for impact evaluation. HCWs received Ksh2000 (approximately US$24.00) to cover travel expenses and lodging.

The training intervention
The training consisted of two consecutive days and included eight modules which were taken in four sessions (i.e., two computer modules per session). Each session was followed by a group discussion. Each group size comprised 18Á19 participants. DASCOs were supported by four members of the research team (i.e., a community liaison officer, a research counsellor, an MSM staff-fieldworker and a social scientist) and two members of a local LGBTI (lesbian, gay, bisexual, transgender and intersex) organization. HCWs were introduced to the sensitivity training on MSM health issues and learned that the training consisted of computer-assisted learning (http://www.marps-africa.org) and group discussions. The curriculum consisted of the following modules of study: (1) MSM and HIV in sub-Saharan Africa; (2) Stigma; (3) Identity, coming out and disclosure; (4) Anal sex and common sexual practices; (5) HIV and sexually transmitted infections; (6) Mental health, anxiety, depression and substance abuse; (7) Condom and lubricant use; and (8) Risk reduction counselling. Modules were designed to be self-completed in 1Á2 hours each, including multiple-choice questions (median 12, range 9Á16) at the end of each module. A score of 71% correct was required to advance to the next module, and upon successful completion of all eight modules, participants were sent a link to download their course certificate. A postcourse evaluation asked for opinions and suggestions for course improvements, using both closed and open-ended questions.
Discussion topics included the identification of subcategories of MSM and their characteristics, sexual practices of MSM and risks for HIV and STI transmission, factors that make MSM vulnerable to STIs and HIV, risk assessment in counselling MSM, best practice for sexual history taking and sexual health examination with MSM, relevant information on safer sex for MSM, personal values and attitudes towards MSM, and addressed stigma and strategies to improve communication with clients who are MSM. At the end of the training, HCWs discussed a work plan on how to strengthen clinical care and uptake of HIV and STD testing for MSM in their dayto-day practice. Study participants with a clinical role were also requested to keep a journal for three months to document and reflect upon their work practices and personal attitudes towards MSM.

Data collection
Course participants completed an online registration, including socio-demographic characteristics (age, gender and level of education and training), details of working practice (role within, type and location of healthcare organization) and specific experience working with HIV prevention, treatment and care with the most at-risk populations (MARPs) in Africa. To assess baseline levels of knowledge, participants conducted a pre-course 24-item multiple-choice assessment covering key learning outcomes across the course material, and they completed a 25-item Homophobia Scale (HS; adapted from Wright et al. [25]). The same two measures were repeated three months after course completion to assess sustained changes in knowledge of and attitudes towards MSM. Immediate post-course knowledge was assessed using the same pre-knowledge questionnaire upon completion of the eight modules. The results of pre-training and post-training assessments were not communicated to participating HCWs.

Measurement scales
Knowledge scores of course material were divided into the following categories: Poor (B17 questions correctly answered), Good (17Á22 questions correctly answered) and Excellent (22 questions correctly answered). When 17 or more questions were correctly answered, the immediate post-training knowledge was considered adequate.
The HS, which was developed and standardized among college students in the United States by Wright et al. [25], was adapted for use in Kenya. The HS aims to measure thoughts, feelings and behaviours towards homosexuality and MSM, and it consists of 25 statements to which respondents indicate their level of agreement on a 5-point Likert scale (Table 1). Questions were reviewed and adapted by three Kenyan research staff and HCWs with professional and personal experience working with local MSM. The adapted HS is shown in Table 1 and reflects changes in terminology (e.g., ''gay'' was replaced with ''MSM'' and ''faggot'' with ''shoga'' in question 9) to reflect local terminology in current use. I have damaged property of gay persons, such as ''keying'' their cars, was replaced with Homosexuality should be treated as an illness/Homosexuality can be cured (question 17); I would feel comfortable with having a gay roommate was replaced with Homosexuality is un-African/is something brought by foreigners (question 18); and I have rocky relationships with people that I suspect are gay was replaced with Gay men have the same rights to public/taxfunded services as straight men (question 25). Responses to items 1, 2, 4, 5, 6, 9, 12, 13, 14, 15, 17, 18, 19, 21, 23 and 24 were reverse coded (item scores 1 05, 204, 3 03 etc.). The total HS score (HSS) was the sum of all item scores, with 25 subtracted from the total. The range is between 0 and 100, with an HSS of 0 being the least homophobic and 100 being the most homophobic.

Data analysis
Analysis was conducted using Stata 11.0 (StataCorp LP, College Station, TX, USA). Binary and categorical characteristics of study participants, established at baseline, were compared using chi-square tests. Although both knowledge and HS scores before and after training approximated to Gaussian distributions, differences between paired measures were non-normal, and thus unadjusted nonparametric methods were used for analysis. Median differences between preand post-training knowledge and homophobia score are reported with an interquartile range (IQR). A Wilcoxon signed rank test for matched pairs was applied to test the statistical significance of differences between pre-and post-training scores. MannÁWhitney and exact McNemar's tests were used to test differences in scores and binary measures, respectively, by HCW characteristics. Spearman's rank was used to assess correlation between pre-and post-training scores, and knowledge and HS scores at both points. Multivariate linear regression models of pre-and post-training score outcomes were explored, but they yielded no additional insight beyond bivariate analysis.

Results
Seventy-four HCWs were recruited to participate in the training programme, and their characteristics are shown in Table 2. The majority were female, and the mean age of participants was 32 years (range: 23 to 53). Sixty-two participants (84%) worked at a government health facility (hospital or clinic), seven (9%) worked at a local nongovernmental organization (NGO) and three (4%) represented faithbased organizations. Most (74%) were in a clinical role (nurse or clinical officer). Irrespective of job role, 8% had received any previous training on how to counsel MSM clients, and a similarly low proportion (7%) had ever received training on how to counsel on anal sex practices. HCWs who had received training on anal sex practices were more likely to have ever asked their male patients if they had sex with men than HCWs who did not report previous training (86% (6/7) versus 31% (21/67), x 2 pB0.01).
All participants said they would recommend the course to others. Open-ended suggestions for course improvements are presented in Table 3. Study participants recommended that the training should be taken by all health stakeholders dealing with MSM issues and be included in medical training. There was an interest in similar training related to other key populations (e.g., women who have sex with women and sex workers).
Effect of training on MSM sexual health knowledge among healthcare workers Table 4 shows knowledge of MSM sexual health issues among participants before the training course, and upon reassessment three months after the course. Prior to the training course, only 10/74 (14%) had an ''adequate'' level of knowledge of MSM issues (threshold score: 17/24), reflecting a median score of 54% (IQR 49Á63%). Levels of knowledge were similar by socio-demographic and workplace characteristics of HCWs, although it was somewhat lower for HCWs in administrative roles compared to other roles (median 42 vs. 54, MannÁWhitney p00.293).
At the end of training, 70/74 (95%) HCWs had adequate course knowledge (exact McNemar's x 2 pB0.001 vs. pretraining). At three months after the course, 35 (49%) of the 71 HCWs reassessed had retained ''adequate'' knowledge compared to 9/71 (13%) at pre-training (exact McNemar's x 2 pB0.001). This represented a significant increase in the median assessment score of 12% (IQR 4Á21%) between baseline and three-month knowledge assessments (Wilcoxon signed test for matched pairs pB0.001). Significant sustained improvements in knowledge were apparent for all HCW age groups and genders, those with clinical or administrative roles and those from governmental health providers.
Pre-training and three-month post-training scores were negatively correlated (Spearman's rho (0.51, p B 0.001), indicating that improvements in knowledge tended to be highest among HCWs with lower pre-training knowledge. There were no significant differences in the degree of knowledge gain by the gender or age group of HCWs; however, participants in counselling roles achieved significantly lower gains in sustained knowledge than other HCWs (median difference: 0% vs. '13%, MannÁWhitney p00.0163).
Effect of training on personal attitudes toward MSM among healthcare workers Table 5 shows HS scores among HCWs prior to training and at reassessment three months later. Overall, the median HS score prior to training was 68/100, representing extensive agreement with homophobic statements and disagreement with statements indicating tolerance of MSM (see Table 5). Male HCWs had slightly higher HS scores at baseline than female HCWs, while HS scores declined with increasing age group but differences were not statistically significant. HCWs in clinical roles (medical and nursing) had higher HS scores than other staff (median 71 vs. 66 respectively, MannÁ Whitney p 00.116), and HCWs working in government facilities had significantly higher HS scores than HCWs in NGOs (Table 5, MannÁWhitney p00.037).
The majority of HCWs reported lower HS scores three months post-training (80.3%, 57/71) compared to their baseline HS score; in four (5.6%), HS scores were unchanged; and in 14.1% (10/71), HS scores were higher after training than before. Overall, the median decrease in individual HS score after training was 8 points (IQR 2Á15), which was statistically significant. These findings did not change in a sensitivity analysis omitting the three HS questions that were culturally adapted (data not shown). Individual pre-training and post-training HS scores were negatively correlated (Spearman's rho 0( 0.71, pB0.001), reflecting the tendency for HCWs with high pre-training HS scores to exhibit greater decreases in this measure as a result of training ( Figure 1).
Male HCWs and those working in clinical roles and in governmental institutions recorded the most pronounced reductions in HS score subsequent to training, although differences in median reduction comparing HCWs' gender, age group, staff role and institution were not statistically significant. More modest declines in HS score were apparent for counsellors (median reduction after training: 4 points) and staff of NGOs (median reduction after training: 0 points); however, it is notable that these groups reported relatively low HS scores prior to training. Collectively, there was some evidence for correlation between scale of increase in individual knowledge and scale of decline in HS score, and this was of borderline statistical significance (Spearman's rho 0(.21, p 00.087).

Discussion
This formal evaluation of a training course aimed specifically to improve knowledge and awareness of MSM sexual health needs among healthcare staff involved in frontline HIV prevention, treatment and care to adult populations in sub-Saharan Africa. Specific and accurate knowledge relevant to the management of behavioural and clinical risks for MSM clients prior to training was poor.
Whilst this may not be surprising in the face of longstanding neglect of Kenyan MSM within HIV policy and resource allocation and a lack of attention to MSM within medical, nursing and HIV counselling training in Kenya, it draws focus to the challenge of maintaining and extending the professional competence of the existing HIV workforce to match the epidemiological realities of the Kenyan HIV epidemic Á especially since the National AIDS & Sexually Transmitted Diseases Control Programme (NASCOP) requires Kenyan HCWs to document the number and category of MSM using HIV services.
Whilst targeted services may well be necessary for subpopulations of MSM, such as male sex workers, they are unlikely to replace the need for MSM-specific clinical care among general health services. MSM-specific programmes have aroused considerable social antipathy in Kenya to date [26] and may in any case not be perceived as accessible to men who covertly engage in homosexual behaviour [27]. Furthermore, surveillance of key populations, including MSM, and strategic information on service coverage to these groups are now an international requirement [28].
The combination of self-directed, modular computer-based learning supplemented by group discussions facilitated by trainers identified from within the existing workforce may offer a relatively sustainable and mobile model for episodic health professional training in this context. The learning content of this course is freely available as a web resource, yet reliable access to internet services remains elusive and expensive in most parts of the country. Even where it is available, the narrative reflections by participants who undertook this training emphasize the importance of the sanction provided by facilitated group discussions to share and explore personal and professional issues arising from the training content that may well be lost in self-directed learning [29].  The brief training programme described here resulted in significant improvements in knowledge of MSM sexual health issues pertinent to day-to-day prevention and clinical practice, and it was sustained by most trainees until at least three months after training. Increase in knowledge was accompanied by a reduction in negative attitudes toward MSM over the same period. Encouragingly, the positive effect of training upon knowledge and personal attitudes toward MSM was strongest among HCWs who had poor levels of knowledge and/or more extreme negative atti-tudes toward MSM prior to training. That positive changes were most marked among HCWs in clinical roles within governmental settings, which represent the backbone of Kenyan HIV services, is cause for particular optimism. Studies to date of perceived barriers to healthcare access identified by MSM in Kenya [7,30] and elsewhere in sub-Saharan Africa [31,32] have reported denial of service, lack of confidentiality, ignorance and verbal abuse from governmental HIV services as central challenges in accessing sexual and general health services. The finding of this study, albeit preliminary, suggests  16.4.18748 both that members of this workforce are willing to learn about MSM sexual health and that their knowledge and attitude toward MSM are responsive to this learning. This study has a number of limitations. The HS, which was originally developed and validated among college students in the United States [25], required amendment to preserve its face validity in a markedly different research context. Whereas the modified scale was responsive to change with training, and these changes were robust to sensitivity analysis excluding modified scale items, the objective meaning of absolute scores and the convergent and divergent validity of this scale in this population remain to be established. Furthermore, although the assessment of training effects upon knowledge and homophobic sentiment was assessed at an endpoint long after the training itself, the longer term effect of training cannot be assumed from this study. Finally, this study lacked a control group which, ideally, would have consisted of HCWs not receiving the intervention and HCWs only participating in the self-directed learning.
We report qualitative narratives among HCWs returning to their workplace after training but finding little support for new perspectives amongst (untrained) colleagues [29]. In a recent qualitative assessment of counselling challenges regarding MSM that are experienced by Kenyan counsellors and clinicians in coastal Kenya, all felt that lack of training and supervisory support impacted their ability to serve MSM [33]. These findings may suggest that either longer term support of trained HCWs and/or more extensive facility-based training of all staff may be prerequisites to longer term changes in institutional practice. While knowledge of same-sex practices is a first step to improve services to MSM, serving MSM in day-to-day practice will further improve services. Follow-up of health workers trained in this study is planned and may provide insights in current care services provided to MSM at two years post-training. Additionally, although the training itself was conducted by facilitators from governmental services that were specially trained for the role, the study was run by a team who was unusually experienced in working with Kenyan MSM, which may threaten generalizability to other settings.
Finally, the ultimate goals of improving knowledge of MSM sexual health needs and reducing prejudicial attitudes toward MSM in healthcare settings are to enhance the accessibility of population-based, public health services to MSM themselves. Although surely a prerequisite to accessible HIV prevention, treatment and care for MSM, the extent to which changes in the attitudes and practices of healthcare providers are reflected in the perceived and practical accessibility and acceptability of services to MSM themselves is unknown. Further study will be required to establish the effect of this brief intervention on long-term attitudes and professional practices towards MSM, and what practical contribution such strategies might make to addressing unmet HIV-related needs among MSM.

Conclusions
In summary, we developed, implemented and evaluated a brief training intervention addressing knowledge and attitudes toward MSM and their sexual health needs in Kenya. The training, which combined self-directed and facilitated group learning, increased health worker knowledge and reduced homophobic attitudes up to three months after training. Scaling up such interventions offers a straightforward response to the immediate need to support HCWs in offering accessible and informed services to address the largely sexual health needs among MSM in Kenya. 9

Introduction
Swaziland is a small, land-locked, lower-middle-income country that is surrounded by South Africa and Mozambique; it has a population of approximately 1.1 million people and a life expectancy at birth of approximately 48 years [1]. Similar to other Southern African countries, Swaziland has been severely affected by HIV, with over a quarter of its reproductive-age adults (15Á49) estimated to be living with the virus, equating to an estimate of 170,000 people living with HIV [2]. Moreover, the incidence of HIV appears to have peaked in 1998Á1999 at 4.6% [95% confidence interval (CI) 4.27Á4.95], according to estimates by the Joint United Nations Programme on HIV/AIDS (UNAIDS), while in 2009 it was estimated to be 2.7% (95% CI 2.2Á3.1%) [3Á6]. There appear to have been further declines in incidence according to 6054 person-years of follow-up data from 18,154 people followed from December 2010 to June 2011 as part of the Swaziland HIV Incidence Measurement Survey (SHIMS) longitudinal cohort. Overall incidence was approximately 2.4% (95% CI 2.1Á2.7%), with incidence estimated to be 3.1% (95% CI 2.6Á3.7) among women as compared to 1.7% (95% CI 1.3Á 2.1) among men [7]. Indeed, women and girls have been more burdened with HIV than men throughout the history of the HIV epidemic in Swaziland, with the HIV prevalence among women 15Á24 in 2006 being estimated to be 22.6% compared to 5.9% among age-matched men and boys [5].
The 2009 Swaziland Modes of Transmission study characterized major drivers of incident HIV infections to be Baral SD et al multiple concurrent partnerships before and during marriage as well as low levels of male circumcision [8]. These risk factors were confirmed in the SHIMS study, with risk factors for incident HIV infections among both men and women including not being married or living alone, having higher numbers of sex partners and having serodiscordant or unknown HIV status partners [7]. There are no known HIV prevalence estimates for key populations in Swaziland, including female sex workers (FSW) or men who have sex with men (MSM) [9,10]. The 2009 Swazi Modes of Transmission Study indicates that both sex work and maleÁmale sexual practices are reportedly infrequent and assumed to be minor drivers of HIV risks in the setting of a broadly generalized HIV epidemic. However, the prevalence of these risk factors has not been measured in the HIV surveillance systems that are used to inform the Modes of Transmission Surveys [11]. The last several years have witnessed an increase in the understanding of the potential vulnerabilities among these same key populations through targeted studies including MSM in neighbouring countries with similarly widespread HIV epidemics [12,13].
The largest body of data is available from South Africa, where the first study completed in 1983 of 250 MSM demonstrated a high prevalence of HIV, syphilis and hepatitis B virus [14]. More recently, a study of rural South African men found that approximately 3.6% of men studied (n 046) reported a history of having sex with another man [15]. Among these men, HIV prevalence was 3.6 times higher than among men not reporting male partners (95% CI 1.0Á13.0, p00.05) [16]. There have also been several targeted studies of MSM in urban centres across South Africa that consistently highlight a population of men who have specific risk factors for HIV acquisition and transmission and limited engagement in the continuum of HIV care [17Á19]. Relatively recent studies from other countries, including Lesotho, Malawi, Namibia and Botswana, have shown similar diverse populations of MSM [16,20,21]. Diversity among populations of MSM across Southern Africa manifests through diverse sexual orientations and practices ranging from those who are gay identified, with primarily male sexual partners, to those who are straight identified, with both male and female sexual partners [22]. Diversity has also been measured in the range of HIVrelated risk practices among MSM, including understanding of the HIV acquisition and transmission risks associated with unprotected anal intercourse and of the levels of use of condoms and condom-compatible lubricants (CCLs) [23].
To better characterize vulnerabilities and HIV prevention, treatment and care needs among MSM in Swaziland, a crosssectional assessment was completed to provide an unbiased estimate of the prevalence of HIV and syphilis among adult MSM in Swaziland. This study was completed in equal collaboration with the Swaziland National AIDS Program (SNAP) in the Ministry of Health. This study further sought to describe the significant correlates of prevalent infections, including individual behavioural characteristics, and describe social and structural HIV-related factors and risks for HIV infection among MSM.

Methods
Sampling MSM in Swaziland were recruited via respondent-driven sampling (RDS), a peer referral sampling method designed for data collection among hard-to-reach populations [24]. Potential participants were required to be at least 18 years of age, report anal sex with another man in the previous 12 months, be able to provide informed consent in either English or siSwati, be willing to undergo HIV and syphilis testing and possess a valid recruitment coupon.

Survey administration and HIV testing
All participants completed face-to-face surveys and received HIV and syphilis tests on site. Surveys were administered by trained members of the research staff and lasted approximately one hour. The study was completely anonymous and did not collect any identifiable information; we used verbal rather than signed consent to further ensure anonymity. Questions on socio-demographics (e.g., age, marital status and education), behavioural HIV-related risk factors (e.g., HIVrelated knowledge, attitudes and risk behaviours) and structural factors (e.g., stigma, discrimination and social cohesion) were included [25]. HIV and syphilis tests were conducted by trained phlebotomists or nurses, according to official Swazi guidelines. Test results, counselling and any necessary treatment (for syphilis) and/or referrals (for HIV) were provided on site. Participant surveys and test results were linked using reproducible, yet anonymous, 10-digit codes.

Analytical methods
Population and individual weights were computed separately for each variable by the data-smoothing algorithm using RDS for Stata [26]. The weights were used to estimate RDSadjusted univariate estimates with 95% bootstrapped confidence intervals (BCIs). Crude bivariate regression analyses were also conducted to assess the association of HIV status with demographic variables as well as a selection of variables either expected or shown to be associated with HIV status in the literature. All demographic variables were then included in the initial multivariate logistic regression model regardless of the estimated strength of their crude bivariate association with HIV status. Non-demographic variables were included in the initial multivariate model if the chi-square p value of association with HIV status was 50.25 in the bivariate analyses. Most of the demographics variables, however, dropped out of the final model after controlling for other independent variables.
Because regression analyses of RDS data using sample weights are complicated due to the fact that weights are variable-specific [27], RDS-adjusted bivariate and multivariate analyses were conducted using individualized weights that were specific to the outcome variable (i.e., HIV status) [27]. The adjusted odds ratio (aOR) estimates were not statistically different from the unadjusted estimates in the bivariate analyses, although some slight differences were observed in the multivariate analyses. Thus, only the unadjusted odds ratios (ORs) are reported for bivariate analyses, while both are presented in Table 1 for multivariate analyses. All data processing and analyses were conducted using Stata 12.1 [28]. Baral SD et al 16.4.18768 Missing data Eleven out of the 324 participants were excluded from this analysis due to missing data on key RDS-related variables. There were 29 out of 313 participants with missing data on at least one variable used in the multivariate analyses. Only two variables had data missing for more than three participants: age at first sex with another man (n missing 04) and knowledge about the type of anal sex position that puts you most at risk of HIV infection (n missing06). Two of the 29 participants with missing data were living with HIV; thus, the effective crude HIV prevalence used in the multivariate model was 17 Although the total number of cases with missing data is not very small (9.3%: 29/313), the number missing by variable is very small. Due to the small change in HIV prevalence in the analysis sample compared to the complete sample as shown in this article, no effort was made to impute missing data. The 29 cases were excluded in the multivariate regression models.

Sample size calculation
The sample size was calculated based on the ability to detect significant differences in condom use among MSM living with HIV and those not living with HIV. There were no known estimates of condom use among MSM in Swaziland, but previous studies of MSM from nearby countries estimated that consistent condom use during anal sex with other men among MSM is approximately 50% [19]. In addition,  [29,30]. Thus, this study was powered on the assumption that those who have received information about preventing HIV infection from other men would have a 16.5% increase in reported consistent condom use. A power analysis demonstrated that with 80% power, we would require 160 participants. Estimates of appropriate design effects for RDS have varied in the literature, and we used a design effect of 2, planning for the accrual of 324 MSM [31]. This sample size facilitates the detection of significant differences in HIV-related protective practices, such as consistent condom use, and targeted HIV-prevention measures, and is sufficient for key social factors such as experiences with stigma and discrimination.

Ethics
The study received approval for research on human participants from both the National Ethics Committee of Swaziland as well as the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health.

Results
Three hundred and twenty-four men were accrued from six seeds over a range of between 1 and 14 waves of accrual, with the largest recruitment chain including 123 participants. As shown in Table 2 (Table 3). About one-third of participants reported having had both male and female sexual partners in the previous 12 months (35.7%, 95% CI 27.7Á43.6). Approximately one-half of the participants reported always using condoms during sex, although significant numbers of men reported both unprotected insertive and receptive anal intercourse in the past 12 months. Condom use was not significantly different between main and casual male or female partners. Overall, safe sex with other men, defined as always using condoms and water-based lubricants over the last 12 months, was not common, with 12.6% (95% CI 7.6Á12.6) measured to report this behaviour. Safe sex, defined as condom use with all sexual partners over the last 12 months, was significantly higher with female partners (at 40.0% in the crude assessment) than with male partners (p B0.05). Overall, safe sex with all sexual partners was uncommon and was reported by 4.3% (RDS-adjusted 1.3%, 95% CI 0.0Á9.7). Knowledge of basic questions related to safe sex for MSM, including sexual positioning, type of sexual act and lubricant use, was low, with 11.2% (RDS-adjusted 9.1%, 95% CI 5.2Á 13.0) of participants providing correct answers. Table 4 demonstrates levels of service uptake, with evidence of statistically significantly lower levels of access to targeted services focused on preventing HIV transmission via sex between men as compared to sex between men and women (p B0.05 for both). Notably, only about half of the sample was somewhat or very worried about HIV. Just under half of the men who had symptoms of a sexually transmitted infection (STI) were tested in the previous 12 months, with 7.8% (95% CI 3.9Á11.7) diagnosed in this same time frame. About half of the sample had been tested for HIV in the previous 12 months (50.7%, 95% CI 43.2Á59.2), including some who were tested more than one time. Reports of any experienced rights violations related to sexual practices, including denial of care, police-mediated violence and physical or verbal harassment, were reported by about half of the sample, although perceived rights violations related to sexual orientation (fear of seeking healthcare and fear of walking in the community) were more common, with 79.6% (95% CI 73.7Á85.5) calculated to report this. Disclosure of sexual practices to healthcare workers was reported by one-quarter of the sample (25.0%, 95% CI 19.0Á31.0), whereas about half of the participants (44.0%, 95% CI 36.4Á51.7) had reported disclosure of sexual practices to a family member.
HIV prevalence was strongly correlated with age in both bivariate analyses (OR 1.23, 95% BCI 1.15Á1.21) for each year of age and multivariate-adjusted analyses (aOR 1.24, 95% BCI 1.14Á1.35) ( Table 1). Other statistically significant associations with HIV in adjusted analyses included identifying as the female gender, having ever been to jail or prison, having lower numbers of casual partners, being diagnosed with an STI in the last 12 months and having easier access to condoms.

Discussion
In the country with the highest HIV prevalence in the world, this study describes the burden of HIV and associated characteristics among MSM who were accrued using RDS. Interpreting the prevalence of HIV among MSM and its relationship with the widespread and generalized femalepredominant epidemic in Swaziland is challenging on a Baral SD et al While the participants in our study were relatively young, the HIV prevalence was consistent with that of general reproductive-age men until age 24Á26, when the prevalence of HIV among age-matched MSM appears to be higher than that of other men sampled as part of the Swazi DHS study (Figure 1) [2]. Given that relatively few men in our sample reported female sexual partners, their HIV acquisition and transmission risks are likely different from those of other men in Swaziland and potentially more related to anal intercourse. Conversely, Swaziland may be among a small number of countries where even the low acquisition risks associated with insertive penile-vaginal intercourse is counterbalanced by the significantly higher HIV prevalence among women, resulting in significant acquisition risks associated with sex with women. However, the idea that acquisition risk for MSM primarily related to sex with other men is reinforced by the results that condom use was lower with male sexual partners than with female sexual partners. Condoms being used more frequently during sex with women as compared to sex with other men have been observed in other studies of MSM across Sub-Saharan Africa and provide an argument against MSM being a population that bridges the HIV epidemic from within their sexual networks to lower risk heterosexual networks [19,20,32,33].
However, to answer this question, phylogenetic studies and the characterization of sexual networks are needed to better describe patterns of HIV transmission. Participants were far more likely to have received information about preventing HIV infection during sex with women as compared to sex with other men. This lack of access to or uptake of information, education and communication services has resulted in participants in this study having a limited knowledge base of the sexual risks associated with same-sex practices. Primarily, participants incor-rectly believed that unprotected penile-vaginal intercourse was associated with the highest risk of HIV transmission, consistent with earlier studies of MSM across Sub-Saharan Africa. Numerous studies have shown the opposite: HIV is far more efficiently transmitted during anal intercourse as compared to vaginal intercourse [13,34]. There was also limited knowledge related to the importance of water-based lubricants being CCLs, which is especially important during anal intercourse given the absence of physiological lubrication in the anal canal. The importance of CCL was underscored as ultimately being the determining factor in just six study participants reporting safe sex with all partners in this study. Thus, while there is significant provision of general HIV-prevention messaging across Swaziland, there has been limited information focused on educating MSM on how to prevent HIV acquisition and transmission during sex with other men. Data suggest that starting with simple and proven approaches, including peer education programmes, is necessary to educate these men about their risks and protective behavioural strategies [35]. However, these approaches will likely not be sufficient to change the trajectory of HIV epidemics given the high risk of infection associated with unprotected anal intercourse with non-virally suppressed HIV serodiscordant partners. Thus, moving forward necessitates assessing the feasibility of combination approaches that integrate advances such as antiretroviral-mediated preexposure prophylaxis and universal access to antiretroviral therapy for people living with HIV [13]. However, the success or failure in achieving coverage with these HIV prevention, treatment and care approaches among MSM will, in part, be determined by the level of stigma affecting MSM.
It is now broadly accepted that addressing the needs of people living with HIV is vital to protect their own health as well as prevent onward transmission of HIV [36]. In addition, mean and total viral loads in a population have been linked to population-level transmission rates of HIV [37]. Only a quarter of the men living with HIV in this study were aware of their diagnosis, demonstrating the need to increase HIV testing, linkage to CD4 testing, and antiretroviral treatment and adherence support for those who are eligible. A recent systematic review and meta-analysis of self-testing for HIV in both low-and high-risk populations demonstrated that selftesting was both appropriate and associated with increased uptake of HIV tests [38]. This may be especially relevant in the Swazi context, where fear of seeking healthcare was prevalent, suggesting the need to study new strategies to overcome barriers to HIV testing among MSM in Swaziland, including leveraging community networks and potentially self-testing [39]. In this study, being a person living with HIV was associated with lower numbers of casual male partners in the last 12 months. This relationship appeared to be stronger among those who were aware of their status, although it was not statistically significant because of limited numbers. In addition, these data are consistent with earlier research findings that simply being made aware of one's status of living with HIV can change one's sexual practices to decrease onward transmission [40]. This further argues for implementation science research focused on optimal strategies to scaleup HIV testing for MSM in Swaziland [41]. Over one-quarter of participants in this study self-identified as women, and this was independently associated with living with HIV. There is nearly a complete dearth of information related to HIV among transgender people across Sub-Saharan Africa [42,43]. However, where transgender people have been studied, they have been found to be the most vulnerable to HIV acquisition because of increased structural barriers to HIV prevention, treatment and care services and because of increased sexual risks, including unprotected receptive anal intercourse [43]. Given the limited information available about transgender people, transgender was assessed in this study as both a sexual orientation and a gender identity. There was a significant disconnect between these two as no participants self-identified as being transgender. Ultimately, further ethnographic research is needed to better understand the HIV-prevention needs of transgender people in Swaziland.
Having been to jail was also independently associated with living with HIV among MSM in this study. Globally, incarceration has been shown to be an important risk factor for HIV, given the limited access to HIV-prevention services such as condoms and CCLs, the interruption of HIV treatment as well as exposure to higher risk sexual partners [44Á47]. While further research is needed on same-sex practices within jails, there is likely a need to provide HIV-prevention services for men in Swazi prison settings [47].
The methods employed in this study have several limitations. While RDS is an effective approach to characterize asymptotically unbiased estimates intended to approximate population-based estimates of characteristics in the absence of a meaningful sampling frame, there are still several uncertainties in the most appropriate tools for interpretation of these data [48]. Moreover, the sample of men accrued here was relatively young, consistent with recruitment challenges observed in other studies of MSM across sub-Saharan Africa. While we conducted significant engagement with older MSM, fear associated with inadvertent disclosure limited their participation in the study. Only with improved social environments will more information about the needs of older MSM become available in difficult contexts [49]. In addition, while RDS was used to accrue a diverse sample, all of the seeds were connected with Rock of Hope, a newly registered organization serving the needs of lesbian, gay, bisexual and transgender populations in Swaziland. We thus may have overestimated actual service uptake among MSM in Swaziland.

Conclusions
The implementation of the research project was guided by recent guidelines to inform HIV-related research with MSM in rights-constrained environments [50]. While these men had not been previously engaged in research on HIV prevention, treatment and care, the success of this study highlights the fact that accrual of this population is both feasible and informative for the HIV response in Swaziland. Moreover, the interconnected social and sexual networks leveraged for accrual can likely serve to disseminate HIV-prevention approaches via MSM throughout the country. While the epidemic in Swaziland is one driven by heterosexual transmission, the burden of HIV and the HIV prevention, treatment and care needs of MSM have been understudied, and these men have been underserved in the context of large-scale programmes [51]. The data presented here suggest that these men have specific HIV acquisition and transmission risks that differ from those of other reproductive-age adults. Encouragingly, Swaziland has seen declines in the rate of new HIV infections over the last seven years, and these declines are related to HIV testing and treatment scale-up [5]. However, the increase in HIV services likely has had limited benefit for MSM, which may result in a scenario where epidemics of MSM expand in the context of slowing epidemics in the general population Á a reality observed in most of the world [13].  (7) In South Africa, MSM-specific service providers and nongovernmental organizations (NGOs), including the Desmond Tutu HIV Foundation (DTHF), engage MSM through both peer education and the use of safe spaces within township communities [13,16,30,33,34]. In 2008, the DTHF used several of these strategies to recruit MSM in Cape Town for the Global iPrEX study, a biomedical HIV-prevention clinical trial [29,35]. Batist  MSM social networks in Cape Town's townships have been described as including key individuals who establish spaces where other MSM are able to socialize safely [36]. As a result of the iPrEX study, links with these individuals and multiple MSM social networks were formed. This led the DTHF to design and conduct a pilot community-based HIV-prevention programme with MSM in these networks.
This pilot programme aimed to reach MSM in various townships through the use of community-based social activities and meeting groups. The programme was designed to disseminate HIV-prevention information and supplies, and promote the use of condoms and HIV service uptake. This article presents an overview of the project methods, a description of the participants and results from follow-up interviews and focus group discussions (FGDs) conducted with a subset of participants.

Methods
The pilot HIV prevention programme was implemented with MSM in five predominantly black African townships in greater Cape Town. Three structured components were included: (i) group meetings were held regularly with small gatherings of MSM to facilitate knowledge exchange and disseminate prevention supplies, (ii) community-based activities were facilitated to provide opportunities for MSM group bonding and (iii) inter-community activities were conducted to promote integration and diversity. All pilot activities took place over a six-month period between May and October 2012.

Community leader selection and participant recruitment
Townships were selected based on high HIV prevalence, which was identified through previous HIV surveillance data, and on the presence of MSM social networks identified through recruitment for the Global iPrEX study.
From each township, one MSM community leader was identified from previous research [30]. Community leaders participated in the planning and facilitation of all activities, disseminated HIV-prevention information and provided healthcare referrals to MSM in their community. They were at least 18 years old; had demonstrated leadership qualities; were respected, trusted and socially prominent among their MSM peers; and lived in a township where pilot activities were planned. The initial community leader team was selected and trained between January and April 2012.
Self-identified MSM were then recruited to take part in the pilot programme using convenience sampling through peer outreach workers and venue-based contact. All participants were 18 years old or older, were born male, were reported to have sex with men and lived in a township where the pilot was taking place. Each participant completed a self-administered paper questionnaire that collected baseline data on their demographic characteristics, sexual practices, health-seeking behaviour and access to services. Participants were offered voluntary HIV counselling and testing by trained staff and were provided with information about MSM-competent healthcare facilities. Participants who tested HIV positive were provided counselling and referrals. Participant recruitment was completed in 57 days between May and July 2012.

Implementation of the pilot programme
Community leaders received initial two-day training and completed follow-up trainings throughout the pilot. Trainings included education on sexually transmitted infections and HIV but primarily focused on developing leadership skills such as effective communication, managing complicated social situations, strategic planning and goal setting, and encouraging healthy social norms.
Group meetings took place every 1Á2 weeks and were held in private and safe venues in each township. Meetings were semi-structured and included both social and educational components such as debates about current events, training on condoms and water-based lubricant, and discussions on HIV-prevention strategies. Meetings were facilitated by a community leader and staff member but guided mostly by the participants, who were encouraged to take ownership and direction of each meeting. Condoms, water-based lubricant and HIV-prevention information were disseminated during these meetings.
Community-based activities were designed based on participant feedback and used to supplement group meetings in each township. Community-based activities included sports (hiking, netball and soccer), dance competitions, drag pageants and debates. Similar to group meetings, HIV-prevention discussions were integrated into each of the activities. Light refreshments were provided to participants at all meetings and activities.
Finally, inter-community activities, which brought together at least two different MSM groups, were conducted at least once a month. These activities were similar in scope to the community-based activities but were organized to promote knowledge sharing and socializing between MSM from different townships. MSM participants were provided with transport to attend inter-community activities.

Data collection and analysis
Quantitative methods Quantitative data from the baseline questionnaires were analyzed using STATA version 11.0 (StataCorp LP, College Station, TX). Numerical variables were explored using measures of central tendency and distribution [medians and interquartile ranges (IQRs)], and categorical variables were explored using proportions and frequency tables.
Participants were requested to sign an attendance register at each activity. Registers were entered into a secure Excel spreadsheet and linked to the participant's ID. Attendance was measured for each participant and defined as the total number of events attended by the total offered to that participant.

Qualitative methods
After completion of the pilot activities, IDIs with each of the community leaders and FGDs with a subset of participants were conducted in December 2012. A purposive sampling strategy was initially used to equally represent MSM who attended regularly and those who did not. However, many participants who did not attend regularly were unable to be contacted, resulting in the remaining FGD slots being filled by participants who attended more frequently. All FGDs and one IDI were conducted in private facilities within each community, and four IDIs were conducted at the research offices of the DTHF. All FGDs and IDIs were conducted by one of the two trained facilitators and supported by a research assistant who took notes. The FGDs and IDIs were conducted predominantly in English, but participants were also encouraged to use the language they felt most comfortable speaking. A semi-structured interview guide was used to explore participants' perceptions and experiences with community life, project activities, stigma, healthcare services and HIV.
Audio recordings from each FGD and IDI were transcribed, and all participant-identifying information was removed. Qualitative data were analyzed using the framework approach. Predetermined themes based on the interview guide questions were used to structure the initial framework, and a coding scheme was developed to identify emerging themes. Two analysts reviewed transcripts from one FGD and IDI together to establish consistency in coding. After this, the analysts each reviewed the remainder of the transcripts individually. Comparisons and discussion between analysts were used to reach consensus on final themes.

Ethical consideration
Written informed consent was obtained from all participants, who were reminded that they would be able to take part in any community-based activities regardless of their decision to participate in this pilot study. Participants taking part in the follow-up FGDs and IDIs were informed that their responses would remain anonymous and would not affect their involvement in future initiatives from the DTHF or other organizations. They received R50 (approximately US$5.00) as reimbursement for their time and transport. Community leaders were provided with a monthly stipend of R800 (approximately US$90.00) as compensation for transport costs and their time spent in project activity planning and implementation. Ethical approval for this project was obtained from the University of Cape Town's Faculty of Health Sciences Human Research Ethics Committee.

Participant baseline characteristics
In total, 98 MSM consented to participate and completed a baseline questionnaire. The majority of participants were black African (95%, 93/98) and gay identified (82.3%, 79/96). The median age of participants was 24.5 with an IQR of 21Á29. Over half of the participants had received secondary education (64.3%, 63/98), and less than one-third (28.6%, 28/ 98) reported current employment. High-risk sexual behaviours including UAI and transactional sex were reported by MSM in each community. In total, 26% (25/98) of participants reported having had at least one female sexual partner in the last six months. A summary of participant baseline characteristics is presented in Table 1.
Community activities MSM community groups were established in 5 townships, and 57 community-based activities including group meetings and 9 inter-community activities were conducted between May and October 2012. Participant enrolment varied between communities, with 33 participants enrolled from Community A, 24 from Community C, 17 from Community D, 14 from Community B and 10 from Community E. Less than half of the participants (44%, 43/98) had previously engaged in other MSM-focused activities or research prior to the pilot.
Attendance registers were not collected from 7 of the 57 community meetings and from one of the inter-community events due to an administrative error. A median of eight (IQR 6Á9) MSM attended the 50 community meetings with attendance registers, and the eight inter-community activities were attended by a median of 20 (IQR 19.25Á21.5) MSM. Condoms and lubricants were distributed during 23 activities and were available on request throughout the duration of the project. Following their enrolment, 60% (59/98) of participants attended at least one pilot activity. Of those participants, 47% (28/59) attended at least one-half of the scheduled activities. A summary of attendance is shown in Table 2.

Follow-up interviews and focus group discussions
Of the 100 MSM who took part in the pilot activities, 36 also participated in follow-up FGDs, and each of the five community leaders completed an IDI. Efforts were made to include participants with varying degrees of attendance; however, there was substantial loss to follow-up of the participants who had lower attendance. Overall, more than half of the participants from the FGDs attended 50% or more of the scheduled activities.
HIV knowledge, testing and services Many participants described the benefits of receiving MSMspecific HIV-prevention knowledge through the meeting groups, while others reported having already received this information elsewhere: I didn't know everything about preventing HIV or AIDS but once I joined the group I've got more information and then that information I used it . . .. I was worried at first you know until I joined the group and then it influenced me in a kinda way to be strong don't have to be worried since you know that mhm theres so many things which can protect you from getting HIV. (FGD1) Participant attitudes towards HIV testing at local health clinics remained consistently negative because of the insensitive or discriminatory care many had previously received. Despite reactions to local healthcare clinics, participants were aware and made use of MSM-competent healthcare services throughout the duration of the pilot project.

Use of water-based lubricants
Prior to the pilot, participants reported limited access to free water-based lubricant and described using petroleum-based lubricant during anal sex. Many participants described that their use of condoms remained inconsistent, particularly with regular sexual partners. Other participants continually referred to an improved knowledge and use of water-based lubricant as a result of taking part in the pilot activities, specifically the group meetings: But we came to the group and they taught us that you have to use specific lube . . . before we can have sex. (FGD 2) Social support and personal development Participants explained how their feelings of loneliness and social isolation were improved after taking part in the pilot because it created opportunities to socialize with other MSM. This seemed particularly true in communities with little existing MSM activities.
. . . You are able to, you know, be yourself and the sense of getting to be yourself and also giving the feeling that you are not alone . . .. (FGD 1) . . . It's nice when we had events, especially in our communities, because there's nothing happening so Suggestions to improve HIV programme implementation Participants offered suggestions for improving the implementation of the pilot activities. Specifically, they felt that staff changes should be kept minimal since it was challenging to develop relationships with new outreach staff. Some participants also expressed the need for improved efficiency with inter-community activities, specifically highlighting the transport and timeliness of other MSM as key barriers. Overall, participants also shared a willingness to engage their broader community to address stigma and expressed a need for activities to do so by targeting other community members beyond MSM.

Discussion
This article presents the outcomes of a pilot communitybased HIV-prevention programme for township MSM. It is important to note that while participants described changes in their behaviour as a result of the pilot, its aim was not to measure behaviour change. Many factors, including concurrent programmes, may have influenced participants' behaviour [14,15]. Taking this into consideration, participant responses do suggest that this pilot was successful in achieving some of its primary objectives. First, the pilot programme successfully engaged MSM from high-risk networks in five Cape Town townships. Attendance data suggest that social activities and group meetings were a feasible method for reaching certain MSM in this pilot; however, overall attendance varied greatly and included a large percentage of participants who attended no activities. This variability may suggest that the pilot activities did not cater to the interests or needs of all participants, particularly the unique needs of MSM [37]. Other factors that may influence attendance have been described and should be explored further in this context, including feelings of mistrust and community stigma [33,38].
Second, HIV-prevention information and supplies were successfully disseminated to MSM during this pilot. Other studies have described how facilitated social spaces can result in knowledge gain by encouraging the exchange and processing of information between peers [39]. Similarly, MSM in this pilot felt that group meetings created safe environments to learn about HIV prevention with their peers. In addition to improving knowledge, increasing access to water-based lubricants and condoms is also essential for MSM, particularly in communities where limited or incorrect lubricant use has been reported [2]. These findings support previous suggestions to explore the use of small communitybased spaces for lubricant dissemination [40]. Small meeting groups and social activities should be further explored as strategies to supplement current lubricant dissemination strategies for MSM in this setting.
Third, participants reported other meaningful benefits to this pilot, including improvements in their self-efficacy, selfesteem and social isolation. Social isolation, poor self-efficacy and limited social support may play important roles in the individual risk of MSM, specifically condom negotiation and lubricant use [10,33,41,42]. Since this study did not aim to address social isolation or self-efficacy directly, it remains to be seen if any risk reduction occurred through this pilot as a result of diminished social isolation or improvements in selfefficacy. However, these results do support previous recommendations to further explore self-efficacy with township MSM in HIV-prevention programmes, and they suggest that community-based group meetings and social activities warrant further investigation as feasible methods to do so [41].
Additional research is needed to explore community-based approaches for condom use and HIV testing in this setting. HIV testing and condom use are complex behaviours affected by a multitude of factors, including stigma [10,43]. MSM in this pilot were supportive of broader community interventions to reduce stigma, lending further support to current recommendations for future community-based HIVprevention interventions to explore methods that empower MSM to safely and appropriately address stigma within their communities [40].
There are limitations to this pilot study. This pilot targeted black African townships; therefore, these findings cannot be extrapolated to other groups. MSM who did not attend pilot activities were not equally represented in the final qualitative interviews. Their reasons for non-participation may not be adequately included in these findings. Even though participants openly shared suggestions for improving the programme, their responses may have been biased towards discussing positive benefits of the programme in general. The timeframe of this pilot was brief and cannot address the sustainability of these activities in the long term.
Taking these limitations into consideration, this communitybased HIV-prevention pilot programme provides useful insights for MSM-specific HIV-prevention programming that warrant further research. Specifically, small meeting groups and social activities promoted an enabling environment, within the context of larger stigmatizing communities, where MSM were able to receive social support, improve their selfesteem and gain access to relevant HIV-prevention information and supplies.

Conclusions
Results from this pilot programme describe how townshipbased MSM can benefit from facilitated social activities and meeting groups. Results from this pilot programme suggest that these strategies are a viable method for disseminating HIV-prevention information, condoms and water-based lubricant. Furthermore, these groups create a supportive environment in which MSM can learn from each other, explore their sexual identities and overcome potential barriers to HIV prevention such as social isolation and low self-esteem. The use of community-based social activities and facilitated smallgroup meetings should be furthered explored as components to ongoing HIV-prevention interventions for MSM in this setting.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions EB, BB and AS were the implementers of this work under supervision of SB and L-GB. EB and BB led the analysis. The manuscript was written collaboratively between EB and BB, with input from SB, AS and L-GB. SB and L-GB provided ongoing support throughout the process of the entire project.

Introduction
The HIV epidemic in Senegal has followed a pattern distinct from the epidemics observed in Southern and Eastern African countries such as Kenya and Malawi, with a far more concentrated epidemic among key populations such as men who have sex with men (MSM) and female sex workers [1]. The Senegalese government launched an early and comprehensive effort to prevent HIV infection in the general population [2]. This campaign was deemed a success by many and is, in part, likely responsible for the limited HIV epidemic in the country, which reports an HIV prevalence of 0.8% among reproductive age women and 0.5% among men ages 15Á49 [3,4]. More recently, there has been increased study of social factors such as unregulated sex work, stigma and discrimination targeting those at high risk of HIV acquisition and transmission, as well as HIV transmission related to same-sex practices among men [5Á11].
MSM have multiple, intersecting drivers of risk and have had a consistently higher risk of HIV acquisition and transmission since the first cases of HIV were discovered [1,12]. This disproportionate burden of HIV in MSM has also been observed in Senegal. Studies dating back nearly ten years have highlighted this disproportionate burden with HIV prevalence among MSM reported to be 22.4% in 2004 and 21.8% in 2007 [11]. Based on these and other data, Dramé reported that HIV prevalence among MSM is approximately 50 times higher than the prevalence observed among reproductive age adults in Senegal according to the most recent demographic and health survey [13]. Finally, the attributable fraction of HIV infections among MSM in Senegal is high; Van Griensven et al. estimated that nearly a fifth of prevalent HIV infections among men in Senegal are among MSM. Collectively, these data reinforce the need to address the HIV prevention, treatment and care needs of MSM in Senegal [14]. Dramé  The definition for social capital is ''institutions, relationships, attitudes and values that govern interactions among people and contribute to economic and social development'' [15]. The importance of social capital has been increasingly recognized as a major social determinant of health because of its association with health outcomes including chronic disease-related morbidity and mortality and, more recently, sexually transmitted infections [16Á26]. Specifically, limited social capital has been associated with higher rates of exchange, survival and commercial sex, and associated with a higher burden of HIV among MSM in Africa [27Á30]. HIV infection has also been associated with low social capital; this may be particularly relevant for stigmatized groups such as MSM [31Á33]. Development of social capital among MSM has been shown to be limited by enacted stigma [34Á36]. And discrimination and stigma affecting MSM are welldocumented, not only in Senegal, but throughout sub-Saharan Africa and more broadly around the world [6,7]. Niang et al. describe the effect that stigma and discrimination can have on health care-seeking behaviours among MSM in Senegal [6]. When men perceive or experience stigma and discrimination in a health care setting, they are less likely to access health services for STI, resulting in higher rates of untreated STI within sexual networks, thereby mediating HIV transmission [37,38].
MSM face additional challenges in countries where sex between men is criminalized [3]. In Senegal, in 2008, several health promoters working in HIV prevention were arrested under suspicion of being homosexual. These arrests, and the fear of further arrests, had wide-ranging effects on HIV in the community of MSM in Senegal [39]. In response, many nongovernmental organizations who had been working in the area of HIV prevention among MSM went into hiding for their own safety. Those who continued distributing prevention materials such as condoms and water-based lubricants saw a marked decrease in the number of men accessing their services. The decreased numbers of men accessing services aimed at the community of MSM also resulted in a lesser availability of social support among MSM. Furthermore, and importantly, health care providers began to fear providing services to MSM following the arrests. This had grave implications for many HIV-positive MSM, who were no longer able to access treatment, either because their provider would no longer see them, or because they feared arrest if they left their home. Some have said that these arrests set HIV prevention efforts back ten years [39].
Stigma and discrimination affect HIV risk and social capital not only by affecting how MSM access prevention and treatment, but also by curbing the presence of research and prevention projects targeting this group in Senegal. A small number of research projects in West Africa has resulted in a limited understanding of what interventions work in communities of MSM in this region [37]. Interventions are difficult to implement, particularly given the constrained legal environment [39]. Community-based organizations of MSM are those with the closest ties to the community and the greatest ability to facilitate interventions [40]. However, these organizations are often not legally registered [13]. Despite these significant challenges, HIV prevention inter-ventions have been effectively implemented for MSM in Senegal [11].
For a population where so much information is left unknown, a cohort study can provide relevant data including prospectively measured HIV incidence [41]. This research project had two primary foci. The first was to assess the feasibility of implementing and retaining participants in a community-driven HIV prevention study in Senegal. The second focus was to describe the study participants in terms of HIV and STI prevalence and incidence, risk behaviours and indicators of social capital at baseline. Ultimately, 119 men were enrolled in the feasibility cohort study. At baseline, all participants completed an informed consent process, a structured survey instrument and a medical examination conducted by an infectious disease physician. The medical examination included a physical exam and syndromic diagnosis and treatment of STI, or a referral for treatment and follow-up if necessary or preferred by the participants. Participants also provided 10 ml of plasma for testing for HIV, hepatitis B and syphilis, according to the Senegalese national testing algorithm [42]. A subset of the participants also received an exploratory intervention.
Because of the small sample size and high loss to follow up, the outcomes of this intervention are not statistically relevant and will not be discussed in this paper.

Follow up
Thirty-seven participants were lost to follow up between T1 (baseline) and T2 (15 months); 14 of these were HIV positive. At the end of the planned implementation period, the remaining participants (n 0 82) again underwent a process of informed consent and completed the same structured survey instrument. At this time period, T2, 60 participants presented for a second session of biological testing for HIV, hepatitis B and syphilis. Whereas at T1, a partner organization was able to perform biological testing on-site immediately following participant surveys, this coordination was not possible at T2. Participants were required to make an additional visit to a clinic for collection of biological samples; 22 participants were unable or unwilling to conduct this additional visit because of inability to pay transportation costs or other competing issues. Retention support was provided by Enda Santé staff through regular visits or phone calls throughout the follow up period, depending on the wishes of the individual.

Ethics
All human subjects' research conducted in accordance with this study has been reviewed and approved by the Senegalese National Ethics Committee for Health Research.

Analytic approaches
The collected data were linked using anonymous codes. Survey data were entered into SPSS, and monitoring data were collected utilizing Microsoft Excel. All data collected were cleaned and merged into a single database. Inconsistencies found during the data cleaning were reconciled to the original questionnaires or laboratory forms.
These data were analyzed using STATA Version 12 (College Station, Texas). Preliminary analysis was conducted using chi square analysis to determine potential associations of social capital at baseline. HIV incidence was calculated by dividing the number of people who seroconverted between T1 and T2 by the number of participants at risk of HIV acquisition (tested negative at T1 and returned for testing at T2), and multiplying this number by person-time. Because of the small sample size and high rate of loss to follow-up, multivariate regression models were not used.

Results
The cohort consisted of 119 male participants who reported having anal sex with another man in the past 12 months, with ages ranging from 18 to 42 years. The mean age for all participants was 28 years, with half of the participants between the ages of 23 and 32 years. Those who were found to be HIV-infected were older than those who tested HIV negative (p 0 0.05), with an average age of 28.8 (interquartile range: 25, 32), compared to HIV-uninfected MSM who had an average age of 26.5 (interquartile range: 22,29). All had had some contact with community groups of MSM in Dakar, Senegal. One-third of the participants had a primary school education or less (n 0 43, 36.4%), onethird had attended secondary school (n 0 39, 33.1%), 15.3% (n 0 18) had attended university and an equal percentage (n 0 18, 15.3%) had attended Islamic or Arab schools. A large majority of participants were single (n 0 104, 88.1%), and 77.3% reported living with their family (n 0 92). Table 1 summarizes the demographic, behavioural, social and financial characteristics of the cohort.

Retention results
Thirty-seven of 119 participants were lost to follow-up (31.1%), meaning they were unable or unwilling to participate in the study at T2. Fourteen of those lost to follow up were known to be HIV positive. HIV-positive participants were not lost at a significantly different rate than HIVnegative participants (p 0 0.43). No statistically significant differences were found between those lost to follow-up and those retained to Time 2 comparing any of the variables listed in Table 1.
Reasons for loss to follow up include participant death, participants being unreachable via contact information and social networks, or participants having moved outside of Dakar. Of the participants lost to follow up, four are known to have died (4/119, 0.03). Cause of death was not recorded in this study.  Table 2.

Biological results
At the baseline medical examination, 49.2% (n 0 59) of participants were diagnosed with an STI. In the biological testing, three cases of syphilis were diagnosed at baseline (prevalence 0 2.6%), and two cases were diagnosed at follow-up (prevalence 0 3.3%). Forty-one participants tested HIV-positive at baseline (36.0%). All participants returning for biological testing at T2 were tested for HIV, regardless of prior test results. Sixty-one participants were tested for at T2, 40 of whom had tested negative at baseline. Eight new infections were observed at T2 (15 months follow up), equating to an annualized incidence of 16 cases per 100 person-years (95% CI 4.6Á27.4%) ( Table 3).

Discussion
This study attempted to use a community based approach to accrue and retain MSM in Senegal for 15 months while implementing a pilot intervention. Although this study was focused on assessing the feasibility of HIV prevention studies, these data also highlight HIV among MSM as an ongoing public health emergency in Senegal. The high incidence of HIV suggests that this is an ideal population in which to assess novel approaches to prevent HIV acquisition. Moreover, the high prevalence of HIV indicates that this is also an ideal population in which to assess the effectiveness of approaches that address the needs of people living with HIV.
These approaches would aim to reduce viral load as a means of improving the health of PLHIV, as well as decreasing the risk of onward HIV transmission.
Loss to follow up in this study was significant, which poses a challenge to the success of future HIV prevention research among MSM in Senegal. Reasons for the loss to follow up were likely multifactorial, including the fact that limited resources were appropriated for enhanced retention approaches in this study. In addition, there was a surprisingly high mortality among this group of men that, with a mean age of 28, was relatively young. Although cause of death was not recorded, anecdotal discussions with community members suggested that these deaths were HIV-related. This pilot cohort study leveraged community groups to implement the study rather than academic teams with significant experience in managing cohorts. Thus, the study demonstrates that cohorts are possible using this approach, but that participant retention strategies should be more thoroughly incorporated into the research protocol. Further research, including qualitative research, is needed to better understand characteristics associated with being retained in the study, and there is a need to explore appropriate retention strategies, for example, using linked peer navigators or SMS-based appointment reminder systems.
Traditional HIV prevention interventions, including condom promotion and HIV testing are necessary. But data on the high force of HIV acquisition and transmission among MSM, as well as the high incidence presented here, suggest that these interventions alone are not enough [43]. Addressing the needs of people at high risk for HIV acquisition could be achieved by assessing the feasibility of antiviral-driven measures such as topical or oral chemoprophylaxis. There are currently Phase II rectal microbicide studies for MSM which include a site in South Africa, and these may eventually represent an important strategy [44]. Separately, oral preexposure prophylaxis has been shown to be effective among MSM and may represent a relevant strategy for particularly high-risk men with limited condom usage despite exposure to condom promotion programmes [45]. The proportion of participants in this cohort who had previously been tested for HIV was high, 88%, though many had not received their results. This suggests the need to optimize the continuum of HIV care in this population; this should include ensuring that people are first aware of their HIV status, then assessed for treatment eligibility, actively linked to treatment services and provided with adherence support to achieve viral suppression [46]. Given the high prevalence and incidence of HIV, these data suggest the need to evaluate active linkage to care interventions for MSM in Senegal [47]. A recent systematic review of linkage and utilization of HIV medical care among PLHIV in the United States reported several approaches for linkage to care may be efficacious, including counseling, education and health system navigators [48]. This study was comprised of a highly selected and relatively small sample of MSM already linked into community based organizations in Dakar. However, these men are subject to multiple levels of stigma and discrimination, including exclusion from social activities, isolation from broader social networks and a society that has criminalized their behaviour. Thus, effective HIV intervention packages should address the individual biological and behavioural facilitators of HIV acquisition and transmission, but also address the broader structural determinants of HIV affecting these men.
The baseline data suggests a relationship between social capital and HIV risk including sexual practices and, potentially, prevalent HIV infections. Men who had less financial need were significantly more likely to report use of condoms and water-based lubricant. These results are consistent with earlier data from Senegal noting the importance of financial stability integration of social services as part of health services in the country [49]. These data link social capital to HIV-related risks and suggest that addressing sexual risk practices without addressing the social contexts in which they are taking place may have limited benefit [43]. Documentation and anecdotal reports from the past two decades have suggested that the implementation of interventions that address social capital among MSM can potentially effectively decrease marginalization, stigma and the risk for HIV infection [16,17,50Á53]. Although the relationship between social capital and HIV risk is complex, increasing trust and community involvement among this vulnerable population may lead to positive changes in social norms and selfefficacy, and can ultimately lead to lower HIV acquisition and transmission risks [16,17,22,23,26,54,55].
The generalizability of this study to the general population of MSM in Senegal is limited for several reasons. Because of the difficulty of contacting MSM, recruitment was conducted using existing community networks allowing for a representative sample of MSM who are members of community organizations in Dakar. However, this approach potentially excluded those who are the most isolated or those who feel the least desire to become involved in the community of MSM. Thus, using a sampling frame derived from members of community based organizations serving MSM potentially selects for a population with higher social capital than average MSM in Senegal. As earlier mentioned, retention in the study was limited, which did not allow for a statistically powered assessment of the benefit of the intervention. Future studies will need to put a heavy focus on participant retention to facilitate evaluation of the tested packages of interventions.

Conclusions
Moving forward, cohorts of MSM will be needed to characterize the effectiveness of combination HIV prevention approaches in the West African context. The experience of conducting this feasibility cohort study with a pilot financial intervention illustrates the potential feasibility of such studies among MSM in a region where they are known to be at among the highest risk for the acquisition and transmission of HIV.

Competing interests
The authors have no competing interests to declare.
Authors' contributions SB, FM, DD MN conceptualized the study. EC conducted the data analysis and led the writing of the manuscript. FD and DD provided management of the research implementation and field teams and supported writing sections of the manuscript. SB and CB provided technical oversight for the implementation, data analysis and manuscript development.
All authors have read and approved the final manuscript.

Introduction
Globally, men who have sex with men (MSM) have substantially higher levels of HIV infection than men in the general population [1]. This is true even in the generalized HIV epidemics of sub-Saharan Africa, where MSM have more than three times the HIV prevalence of general population adult males on average [1]. Despite the knowledge that MSM are more likely to be infected with HIV across settings, there has been little investigation of the experiences of MSM who are living with HIV in sub-Saharan Africa. Positive health, dignity and prevention is a framework used to highlight health and social justice issues for people living with HIV (PLHIV) [2,3]. The primary goals of positive health, dignity and prevention are ''to improve the dignity, quality, and length of life of people living with HIV; which, if achieved will, in turn, have a beneficial impact on their partners, families, and communities, including reducing the likelihood of new infections'' [2]. This framework builds upon earlier concepts of ''positive prevention'' and ''prevention with positives,'' which highlighted the importance of ensuring the health of PLHIV and engaging PLHIV in HIV-prevention efforts [4Á6]. However, positive health, dignity and prevention situates living with HIV within a human rights framework and focuses on the importance of understanding and addressing structural constraints. It also considers the role of stigma and discrimination, which Parker and Aggleton [7]  describe as social processes related to social inequality, power and oppression through which some groups are structurally excluded in society. Stigma has often been defined based on the classic work of Goffman as the social devaluation of a person based on a ''significantly discrediting'' attribute [8], while discrimination has been defined as behaviour resulting from prejudice [9]. Both stigma and discrimination are common in relation to both HIV and same-sex relationships.
In Swaziland, HIV prevalence in reproductive-age adults is among the highest in the world at 26.1% [10]. UNAIDS classifies Swaziland as a generalized HIV epidemic and, to date, the response to HIV in Swaziland has largely focused on the general population. Recently, the first surveillance of HIV prevalence and associated risk factors among MSM in Swaziland was conducted and showed a high burden of HIV among Swazi MSM, comparable to that of men in the general population [11]. However, same-sex behaviour is criminalized in Swaziland, and little attention has focused on the experiences of MSM who are living with HIV in this setting. Indeed, we identified just one peer-reviewed article focusing on HIV-positive MSM in sub-Saharan Africa. Cloete et al. [12] conducted a survey on HIV-related stigma and discrimination among a convenience sample of both HIV-positive MSM and men who have sex with women in Cape Town, South Africa. The survey found that internalized HIV-related stigma was high among all participants. Overall, MSM reported slightly greater social isolation and discrimination due to their HIV status, but these differences generally did not reach statistical significance.
In this study, we sought to explore the positive health, dignity and prevention needs of MSM who are living with HIV in Swaziland to inform HIV prevention, care and treatment services for this population. To our knowledge, this is one of the first qualitative studies to examine these issues among HIV-positive MSM in sub-Saharan Africa. As such, findings could inform the design and implementation of programmes for MSM living with HIV in Swaziland and similar settings.

Methods
A qualitative approach was used to address the study aims. Methods included key informant interviews, in-depth interviews with HIV-positive MSM, and focus groups with MSM community members.
Key informants were selected if they had experience with MSM and lesbian, gay, bisexual and transgender (LGBT) populations or with HIV-related services in Swaziland. Sixteen key informants were interviewed, including HIV programme planners, policy makers, clinicians and LGBT community leaders. Interviews were semi-structured and employed a field guide to direct the conversation and stimulate probing. Participants were asked to describe the situation of MSM in their communities, their knowledge of existing services for MSM and PLHIV and their suggestions for how services could better meet the needs of MSM.
In-depth interviews were conducted with 20 MSM living with HIV interviewed twice each for a total of 40 interviews. Recruitment was conducted through a variety of settings and organizations, including HIV clinics; PLHIV networks; LGBT and MSM community organizations; and HIV prevention, care and treatment services. Participants were asked about the experiences of MSM generally in their communities; MSM social networks; personal and community experiences with HIV prevention, care and treatment services; experiences with stigma and discrimination; and suggestions for how services, interventions and messages could be better tailored for MSM.
Focus groups were conducted with MSM to gather a broader community perspective on the study topics; HIV status was not asked for reasons of confidentiality. Three focus groups were conducted with 26 MSM (4, 9 and 13 participants in each group). Topics covered were similar to interviews.
All interviews and focus groups were conducted in a private setting in either English or SiSwati and lasted approximately one to two hours. MSM were interviewed by a Swazi familiar with the local LGBT community who received training in qualitative research, while key informants were interviewed by an American masters-level research assistant with qualitative training living in Swaziland.

Qualitative data analysis
Analysis of qualitative data was conducted through identification of recurrent patterns and themes following Crabtree and Miller's five steps in qualitative data analysis, or the ''interpretive process'' [13]. These steps are: (i) describing, (ii) organizing, (iii) connecting, (iv) corroborating and (v) representing. These steps form part of an iterative process which starts by re-examining the goals of the research and considering questions of reflexivity, then moves towards ways of highlighting, arranging and reducing texts to make connections through the identification of recurrent patterns and themes.
All interviews and focus groups were recorded, transcribed and translated into English. Debriefing notes immediately following each interview captured the interview context, theoretical issues, methodological issues and follow-up topics. Weekly meetings were held with all interviewers to discuss emerging themes and identify topics for further exploration to ensure an iterative process. After all data were collected, a full-day data analysis workshop was attended by representatives from LGBT groups, Ministry of Health (MOH) and National Emergency Response Council on HIV and AIDS (NERCHA) representatives, interviewers, clinicians and other stakeholders. This workshop devoted individual time to read de-identified transcripts to identify themes, then group time to categorize and discuss emerging themes and implications. Following the workshop, a codebook was developed by four study team members working together until agreement was reached. Codes were selected based on a priori topics of interest (research questions), themes identified during the data analysis workshop and emergent themes from transcripts. Codes were then applied using the computer software package Atlas.ti (version 5.2, Scientific Software Development GmbH, Eden Prairie, MN). The coded text was read to identify further themes or patterns and memos were created for key themes, which were developed into the findings presented here. Kennedy

Ethical considerations
All participants provided oral informed consent prior to participation, and referrals to clinical and counselling services were provided as needed. Study staff members were trained on sensitivity issues around HIV and MSM. A study advisory board, including representation from the LGBT community, implementing partners and government, reviewed the study protocol and interview guides and provided ongoing advice to the management and execution of the study. Ethical review and approval for this study was received from the Scientific and Ethics Committee of the Swaziland MOH and the Johns Hopkins Bloomberg School of Public Health in the United States.

Results
Dual stigma and disclosure of sexual identity and HIV status The predominant theme across interviews was the significant and multiple forms of stigma and discrimination faced by MSM living with HIV in Swaziland. MSM reported experiencing stigma and discrimination related to both their HIV status as well as their sexual identity.
Same-sex behaviour is both criminalized and heavily stigmatized in Swaziland. MSM reported experiencing significant stigma, discrimination and rejection as a result of their sexual identity. One man, when asked if he had ever experienced stigma or discrimination as a result of being gay, responded, ''A lot, several times, too many times.'' As a result of these experiences, and fear of similar stigma and rejection, many participants said they had not disclosed their sexual identity to anyone except other MSM. ''That is my secret and I'm not planning to tell anyone in my family,'' explained one. Participants worried about negative reactions, rejection and abuse if they disclosed. One man, when asked what would happen if he disclosed his sexuality to his friends or family, responded, ''I would not even dare. It would be like being in a devil's den.'' Others worried more about disappointing their loved ones by not conforming to social norms. One MSM asked, Do you know this SiSwati saying that goes, 'you have to have a heart for the other person'? . . . We always put the next person before [ourselves] . . . So we hardly want to disappoint the next person with being me, being myself and being comfortable with myself and insisting that I should be accepted, you know. We want to always conform [to] what society expects.
However, some participants had disclosed their sexual identity to family members or friends and had found acceptance, often after some initial difficulty.
Men also described stigma related to their HIV status. One participant described ''the abuse we are subjected to'' as ''stigma, you see, that once you are HIV-positive, people think that you have AIDS. And also, that people have not accepted and they still do not know what HIV is.'' Experiences or fear of HIV-related stigma prevented many MSM from disclosing their HIV status to family, friends and sexual partners. Lack of disclosure led to challenges with antiretroviral drug (ARV) adherence, hiding medications and a lack of social support for care-seeking and adherence to care and ARVs.
Participants selectively disclosed either their HIV status or their sexual identity to different individuals based on their anticipated reaction. For example, participants said they might disclose their HIV status to family members as they anticipated receiving some material or emotional support as a result, but they might not disclose their sexual identity to those same family members due to fear of rejection or a negative reaction.
Violence and lack of police protection Violence was also a common experience for MSM. MSM reported violence from a range of individuals. One man noted that some MSM ''are killed for being gay, others are assaulted and others are chased away from home and disowned.'' Due to the criminalized nature of same-sex behaviour in Swaziland, many MSM felt they had no recourse to bring incidents of discrimination or violence to the authorities. Furthermore, many had experienced a lack of police protection as a result of their sexuality. One participant described such an incident: Participant (P): I was actually with a friend of mine in Manzini and we went to the butchery for a braai [barbecue], and when we got there, umm, there were these people who were, like, sitting outside at the car park. They were just rude and they started insulting us and we didn't try to defend ourselves, try to explain anything, and they went on, like, we are gay, we have to be beaten up, the gayness should be beaten out of us. We just ignored them and they attacked one of my friends we were with, they started beating him and he was bleeding. Interviewer (I): Really. P: Like for real, he bled to the point where we had to go to the hospital and we obviously went to lay a charge. And the police were kind of 'occupied', they didn't have the time to go and find these people that have beaten my friend. I: Why did the police act in that way? Did you narrate to them what happened? P: We sure did, but I think it's because we told them how the whole thing started Á they called us names because they say we are gay. And I think also the police could tell that we are [gay], so they thought there was no case there.

Stigma from healthcare settings
The stigma associated with being an MSM was the predominant barrier to accessing healthcare services for MSM living with HIV. Both perceived and experienced stigma in healthcare settings led to a lack of care-seeking behaviour. As one participant described it, When they say 'bring your partner', and then you bring the same sex partner, they are like, 'yah, this is why you are having this [HIV], this is why', and they will be throwing words at you . . . so then you get embarrassed, sometimes you'll decide to leave without being treated, and where are you taking that sickness to?
Another participant, when asked how the needs of MSM differed from PLHIV in general, explained that the main difference was how forthright MSM could be about issues related to their sexuality: I think they are different in the sense that for those who are straight they are open and they communicate easily about sex issues. As for us gays, it's difficult unless you have someone you can talk to and give you advice as to what you can do when you have some health issues. As for people in general, with them it's easy for them to go to hospital, but with us it's difficult. You can't say it's painful in your anus Á what will you say the cause for that is?
This participant continued by noting that this influenced care-seeking behaviour, as he would delay care-seeking or self-medicate to avoid disclosure: I: What happens, so you end up not going there [to the hospital]? P: I just stay at home and you find that this thing becomes complicated. When this thing becomes complicated, you find that maybe you go to the pharmacy and they tell you that this thing is at an advanced stage.
Other men said they travelled long distances to seek HIV care at clinics where they either were not known personally or where they did not experience stigma and discrimination. P: Even at the hospital, they interviewed me, then there were changes and I could tell that they wanted me to reveal what type of person I am. Since then I stopped fetching my drugs there. I now go to another clinic which is far away from home. I drive all the way to fetch my tablets instead of taking them locally. I: Really, why is it so? P: Because I thought there is problem at the local clinic since I am gay. So I decided to change . . . They treat us like small devils, as if we are the one who are spreading the HIV virus.
However in a few cases, MSM did disclose their sexual identity to healthcare providers and reported positive and supportive reactions, particularly from non-governmental HIV testing and counselling sites.
Fear of stigma also shaped the type and nature of counselling that MSM received in healthcare settings, particularly regarding offering services to sexual partners. MSM, as well as key informants, noted that in clinical services such as HIV testing and treatment, providers' questions about HIV prevention generally assume heterosexuality. Providers would ask MSM to bring their wives into the clinic to be tested for HIV. Due to fear of stigma, MSM would often simply state that they did not have a wife, but would not mention their male sexual partners.
Finally, participants reported mistreatment by staff and lack of confidentiality at clinics due to being HIV-positive. These negative experiences were particularly experienced when picking up ARVs, leading one MSM to say that ''the ARVs end up being an inconvenience [rather] than helping you.'' Some men felt that PLHIV in general were treated poorly by healthcare workers. ''You really feel that you are different from other people,'' explained one. However, others felt that at least some healthcare workers provided highquality care to PLHIV, and that MSM were not necessarily treated any differently from other PLHIV.

Mental health challenges
Many MSM said that living with a stigmatized sexual identity and a challenging, stigmatized disease led to feelings of depression as well as self-stigma or shame. ''To be like this to me seems like I was created for nothing on earth,'' said one, ''because there is nobody who is happy about me at home and at school.'' The initial receipt of an HIV-positive diagnosis was emotionally devastating for many participants. Participants described feelings of depression and anger. They also said that others had even more difficult coping. ''Some of them they commit suicide because they can't accept their status,'' said one MSM, ''because no one can accept them as they are gay and positive.'' Some participants said feelings of selfstigma led them to drink alcohol as a coping mechanism.
[After testing HIV-positive], I was very much hurt so much that I decided to devote myself to drinking alcohol. I was drinking every day, and there was not a day that went by without me drinking.
However, over time, many participants said they came to accept their HIV status and learn to cope with the disease. MSM also reported that they had difficulty accepting their sexuality. Some described shame related to having sexual feelings for other men.
Participants reported receiving emotional support from a variety of sources. One MSM said he went to his pastor for support, while another derived comfort from religion but had not disclosed or discussed his life with his church. Only one participant mentioned going to formal counselling services, saying he and his partner saw a private counsellor who knew they were gay. However, most received support from partners, friends or family to whom they had disclosed either their HIV status or their sexual identity.
Preventing HIV transmission to sexual partners and the context of MSM relationships MSM in this study were very aware of the need to prevent onward HIV transmission to sexual partners. Many discussed how they had changed their behaviour after being diagnosed with HIV in order to reduce transmission risk to others by using condoms and reducing the number of partners. However, others reported continued risk behaviour, often linked to alcohol use. As one participant put it, ''most of the time we have sex without a condom it is when we are drunk.'' Poverty and lack of economic opportunities also shaped risk behaviours. Participants reported that some members of Kennedy  the MSM community were not necessarily gay, but engaged in transactional sex with men to support themselves financially. However, the majority of our participants identified as gay, and many said they were in long-term, monogamous partnerships with other men.
Some MSM felt that the clandestine nature of MSM relationships in Swaziland may lead to greater numbers of and more casual types of partnerships. MSM described many of their partners as bisexual or having female girlfriends and wives, possibly to fulfil cultural expectations. Furthermore, MSM said that their relationships are often kept secret and therefore families do not play a role in relationship counselling and peacekeeping as they might for heterosexual couples.
Usually in our community we have short-term relationships. These relationships are caused by the fact that there is nothing bonding those people. And maybe the community, the parents or relatives are not involved in our relationships. And then if I have got a problem with my boyfriend, if I say it's over, it's over . . . you are not able to go tell your parents or relatives . . . if people are informed either way about such people [MSM] in the community, if there is a relationship going on with his parent, the parent will be able to intervene either way, and those relationships will sustain.
Improving positive health, dignity and prevention services for MSM MSM said that societal acceptance and stigma reduction would be the most important way to improve services for MSM living with HIV. As one man stated, ''If we can be recognized and they can know that there are people who are living this kind of life and they can know how they can reach us in terms of programmes and services.'' Participants knew that same-sex relationships were more accepted in neighbouring South Africa and hoped that social norms in Swaziland might shift in a similar direction. They also discussed the organizations working openly for LGBT health and rights in South Africa and noted that the lack of such formal organization in Swaziland limited the ability to develop an effective and appropriate response to HIV for MSM.
Participants held a variety of opinions on how best to tailor existing interventions and services for MSM. Some participants suggested developing special clinics or services for HIVpositive MSM. Others worried that targeted services would reinforce stigma. One potential consideration was including MSM living with HIV as ''expert clients'' to help navigate HIV treatment services. Participants said less about mental health services; just a handful of interviewees said that increasing access to counsellors would be helpful, as existing HIV care and treatment providers were overworked and did not have time to provide in-depth counselling for PLHIV.
Currently, as there are essentially no HIV-prevention services for MSM in Swaziland, participants suggested a ''training of trainers'' model, whereby trusted MSM community members could be trained in HIV-prevention messages particularly relevant for MSM and could then share those messages with others in their community. MSM also suggested continued or expanded distribution of condoms and particularly lubricant to prevent condom breakage.
Several participants, both MSM and key informants, said that healthcare workers should be trained on issues related to MSM. As one key informant explained, ''Even their procedures manuals should have information on how to handle MARPS [most at-risk populations, including MSM].'' Importantly, key informants in this study consistently said that regardless of personal belief, they had an ethical responsibility to provide services to everyone, equally. ''As a [member of the] health sector, my belief is non-discriminatory services to all the members of the population, and issues of legality and everything rest with the Ministry of Justice,'' said one. Another stated, Even though I don't approve of what they are doing . . . as a public health officer, I have to make sure that they have access to health services. I don't have to judge them. I don't have to give my views on what they are doing. But my duty is to make sure that they have access to services . . . whatever their sexual orientation is, they are human beings, they are Swazi.

Discussion
This study is among the first studies to examine the positive health, dignity and prevention needs of HIV-positive MSM in sub-Saharan Africa. We found that a social and structural context characterized by significant and multiple stigmas was key to understanding these needs. Dual stigma related to both sexual identity and HIV status led to selective disclosure or lack of disclosure of both identities, and consequently a lack of social support for care-seeking and medication adherence. Perceived and experienced stigma from healthcare settings, particularly around sexual identity, also led to delayed care-seeking, travel to more distant clinics and missed opportunities for appropriate services. These findings support and extend findings from other sub-Saharan African settings that discrimination reduces the willingness of MSM to access services [14Á16]. The lack of support from friends, relatives and society for same-sex relationships was described as weakening these relationships, leading to greater numbers of sexual partners as well as relationships with women to ''hide'' same-sex behaviours, potentially further increasing HIV risk. This finding similarly echoes research from the United States suggesting that psychosocial health problems may increase HIV risk among MSM, leading to a ''syndemic'' [17]; such findings highlight the need to approach HIV prevention within the context of overlapping health problems [18].
Intersectionality is a theoretical framework that examines the relationship or ''intersection'' between multiple forms of oppression and discrimination due to social categorizations such as race, class or gender [19]. MSM living with HIV experience the dual stigma of being a sexual minority and having a stigmatizing illness. Intersectionality posits that these multiple stigmas are not experienced independently, but that they interact in complex ways to create disparity and social inequality in health outcomes [19]. We found that MSM living with HIV described dual stigma as an Kennedy  overwhelming burden in their lives which influenced multiple aspects of their health and relationships. Considering the needs of MSM living with HIV in this intersectionality framework provides the deepest understanding of their experience.
Intersectionality also highlights the ways in which individual experiences with stigma reflect larger social structures that create and sustain inequality. Participants in our study experienced outright discrimination, stigma and violence against MSM and PLHIV. However, because sexual identity can be concealed, they often encountered situations in which they were assumed to be heterosexual Á assumptions which they did not correct due to fear of discrimination. For example, in healthcare settings, many providers assumed their clients were heterosexual and provided services accordingly. For our participants, these assumptions led to missed opportunities for appropriate counselling services tailored to their individual needs and risks, as well as missed opportunities for offering important services, such as HIV testing and counselling, to their sexual partners. Although the World Health Organization couples HIV testing and counselling guidelines support offering these services to same-sex couples [20], in practice, most couples HIV testing and counselling services in sub-Saharan Africa focus exclusively on steady, heterosexual partnerships and fail to consider samesex relationships. Although individual providers may offer supportive services for same-sex partners, a more comprehensive approach is needed to incorporate training on samesex relationships into couples HIV testing and counselling programmes.
Currently, services for MSM living with HIV in Swaziland are essentially non-existent. This is unsurprising, given the lack of data on MSM and HIV risk in Swaziland until very recently and the criminalization of same-sex behaviour in Swaziland. Research has documented a strong correlation between criminalization of same-sex behaviour and lack of investment in services for MSM globally [21]. However, MSM have unique healthcare needs [22], and even in rightsconstrained settings, comprehensive HIV services for MSM can and should be provided [23]. Our findings suggest the beginnings of political will among healthcare workers, key stakeholders at the government and local levels and the MSM community to provide these services. Key informants in particular reflected on their duty to provide services to all Swazis in a non-discriminatory manner. These beliefs can provide a foundation for establishing comprehensive HIV services, including both prevention and care and treatment services, for MSM. In fact, this research helped, in part, to catalyze the official registration of an NGO, Rock of Hope, dedicated to key population rights in Swaziland, including LGBT rights, which has been invited to engage with the country's key population policy technical working group addressing HIV among MSM and other key populations. The technical working group is under the auspices of the Swaziland National AIDS Programme (SNAP), a programmatic body under the MOH. Other implementing partner organizations providing HIV-related services have indicated they would be open to developing services for MSM. In this changing political and institutional context, there appears to be a genuine possibility of government, NGO and civil society collaboration to develop an effective and comprehensive response to the HIV epidemic among MSM in Swaziland.
This study provides unique information about the needs of MSM living with HIV in a sub-Saharan African context with high HIV disease burden. Conducting multiple interviews with MSM living with HIV and working closely with local LGBT groups increased the comfort level of our participants and their willingness to participate in this study. However, MSM participants were still discussing very sensitive, stigmatized and illegal behaviours, and they may not have fully opened up to interviewers. Data were collected largely from MSM in urban centres due to reliance on existing networks; this may limit transferability of the findings to rural MSM or those without strong MSM social networks.

Conclusions
The intersecting stigmas of sexual identity and HIV status shaped multiple facets of the lives of MSM living with HIV in Swaziland. Intersectionality provides a framework for understanding these experiences and highlights how programmes and policies should consider the specific needs of this population when designing HIV prevention, care and treatment services. In Swaziland, programmes should consider tailored multi-level interventions that address these unique needs at the policy, societal and healthcare delivery levels. At the policy level, the health sector in Swaziland is already initiating important research to examine the epidemiology and service delivery needs of MSM; findings from this research should be incorporated into the national HIV response. For Swazi society in general as well as healthcare providers, interventions to reduce stigma, discrimination and violence against MSM and PLHIV are needed. The health sector should also consider distributing condoms and lubricant for MSM, training healthcare providers in the specific health needs of MSM and engaging MSM as peer outreach workers or expert clients in both prevention activities and clinical services. Finally, further research examining the experiences and needs of MSM living with HIV globally is required to improve comprehensive HIV services for this population.

Introduction
The sub-region of West and Central Africa (WCA) is the most populous of sub-Saharan Africa (SSA), with a combined population of roughly 356 million [1]. The region possesses a distinct cultural, economic and historical diversity. The majority of countries purport French as their national language, while English is the state language for four countries, and Spanish and Portuguese are both spoken within the region. Fifteen of the countries in WCA are classified by the World Bank Atlas method as low income ( US$1025), including Benin, Burkina Faso, Cape Verde, Central African Republic, Chad, the Democratic Republic of Congo (DRC), the Gambia, Guinea, Guinea-Bissau, Liberia, Mali, Mauritania, Niger, Sierra Leone and Togo [2]. Côte d'Ivoire, Cameroon, Ghana, Nigeria, the Republic of Congo, Senegal and São Tomé and Príncipe are categorized as low-middle income (US$1026 to US$4035) [2]. One country in the region is upper-middle income (Gabon), and one is ranked as a high-income country (Equatorial Guinea), mainly due to newly found oil reserves and a population under 1 million [2].
Historically and economically multifarious, the region has not been immune to the HIV epidemic. The first reported cases of HIV emerged in the mid-1980s, and national surveillance bodies such as National AIDS Committees (NACs) were established over the subsequent decade [3]. Early phylogenetic subtyping revealed unique regional dynamics, with both HIV-1 and HIV-2 circulating, and the majority of global cases of HIV-2 found in West Africa. Concurrently, the origins and greatest subtype diversity of HIV-1 were reported in Central Africa [4] (Figure 1). Nevertheless, regional epidemiological reporting has traditionally been immersed in the overall context of SSA. Trends in the HIV epidemic show that SSA possesses the highest burden of HIV, and 69% of the global population of people living with HIV reside within its borders [23.5 million (22.1Á24.8 million)] [5,6]. While these statistics show an important burden of disease on the continent, they mask disparities in HIV epidemics regionally [7]. Countries in East and South Africa report consistently generalized epidemics among reproductive-age adults (ages 15Á49), which is defined through the Joint United Nations Programme on HIV/AIDS (UNAIDS) criteria as HIV prevalence consistently higher than 1% in antenatal clinics [8,9]. Nine out of the 15 Southern African Development Community (SADC) members report national prevalence over 10% [5,6,10]. Reproductive-age adult estimates are as high as 25.9% in Swaziland and 24.8% in Botswana [11]. Comparatively, national prevalence in WCA has remained low or moderate since HIV surveillance reporting began, with current general-population estimates ranging from 0.02 to 4.5% [5,6,12]. Twelve countries in the sub-region report national prevalence under 2% [5]. Consequently, the majority of these countries' HIV epidemics are classified as mixed, concentrated or borderline generalized [6,12].
The international community has recently noted that classifications of the HIV epidemic based on prevalence data often limit understanding of the complexity of transmission and appropriate prevention strategies. However, concentrated epidemics have historically been defined as occurring in countries where HIV prevalence is consistently higher than 5% in at least one subgroup within the population, but less than 1% in antenatal clinics [7,9]. These subgroups are generally considered to be female sex workers (FSWs), men who have sex with men (MSM) and people who inject drugs (PWID) [7,13]. There is less clarity around mixed epidemics, although these are generally agreed to be lowlevel generalized epidemics ranging from 2 to 5% HIV prevalence in the general population, and high transmission rates in subgroups of the population [7]. Based on this, the HIV epidemics in countries in WCA are predominantly mixed or concentrated.
Researchers have suggested that the complexity of the regional dynamics in WCA has not been dissected adequately [12,14Á16]. Underlying drivers such as migration patterns, subtype diversity, significant regional variations of the disease and at-risk populations are understudied [11,12,16Á19]. In an era where the global spread of HIV is on the decline, data are progressively emerging to show sustained or expanding transmission in populations at high-risk for HIV [15,20Á22]. However, national surveillance systems, particularly in low and middle-income countries, remain constructed on populationlevel studies such as the Demographic and Health Survey and antenatal care surveillance data [6,13]. These methods provide a global overview of basic risk factors associated with transmission, but they do not capture data characterizing sex work and other transactional or compensated sex, same-sex practices and drug use outside of alcohol consumption, all of which are demonstrated high-risk factors and contributors to the acquisition and transmission of HIV [11,21,23].
Globally, surveillance shows that groups such as FSWs, their clients, MSM and PWID sustain a higher burden of disease in concentrated epidemics and substantially contribute to new infections annually [4,7,18,22,24]. In settings such as Southeast Asia and Latin America, general-population HIV prevalence remains similar to that of WCA, and a higher burden of disease is observed among key populations. For example, Pakistan and Indonesia report 25% and 35% prevalence among PWID, respectively [25]. Vietnam and Chile report an HIV prevalence rate of 15% and 20% among MSM, respectively [25,26]. Myanmar (Burma) reports a prevalence of 10% among FSWs, and Brazil reports 4.9% [25,26]. All of these reported levels are roughly five to thirty times higher than general-population prevalence in the specific countries listed [25,26]. National-level responses on these continents have included programmes for key populations, and noteworthy advances in the reduction of new infections have been reported over time [27,28]. In contrast, WCA reports partial or sporadic data for key populations and limited governmentlevel policies defining key population treatment and prevention needs [5]. National surveillance and programming in WCA subsequently remain rooted in broad HIV prevention messaging and approaches similar to those seen across East and South Africa such as prevention of mother-to-child transmission (PMTCT) and non-targeted community-based behaviour change programmes [5,7].
Lessons learned from other contexts such as Southeast Asia and Latin America, where limited prevalence of HIV among average-risk reproductive-age adults also exists, require us to examine the epidemiology of the HIV epidemic in WCA [11,29]. This systematic review aims to complete a historic, situational and epidemiological analysis of the burden of disease among key populations in 24 countries located in WCA.

Methods
The US National Library of Medicine's MEDLINE database, one of the most comprehensive sources of healthcare   16.4.18751 information in the world, was searched using the PubMED interface to obtain biomedical markers for any of the three key populations: FSWs, MSM or PWID. The study objectives specified the need for epidemiologic studies that report biological endpoints (HIV prevalence) with defined methods; thus, it was decided a priori that MEDLINE would be sufficient. However, a sensitivity assessment was employed using the same search strategy to explore EBSCOhost CINAHL Plus, PsycINFO, Ovid, SocioFile and Popline, and no additional data points were obtained which met the defined inclusion and exclusion criteria. Google and Google Scholar were searched for contextual information and non-peer-reviewed literature. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA) guidelines were referenced for the development of the search protocol and study reporting structure [30,31].
The medical subject headings (MeSH terms) for HIV and AIDS and key terms relating to ''sex work,'' ''men who have sex with men'' and ''intravenous drug use'' were cross-referenced with terms associated with 16  The inclusion criteria for this study included reported HIV prevalence data for any of the three key populations, as well as clients of FSWs, in any of the 24 countries defined for this review. Publications were included if prevalence was listed in the article with sample size and sampling and HIV-testing methods described, regardless of the overall aim or topic of the study. Date of publication was not used as an inclusion criterion. Exclusion criteria included manuscripts not published in French, English or Spanish. Articles were downloaded and organized using Endnote (version X5), and data collection was finalized in April 2013.
Screening and data abstraction A title and abstract search protocol was utilized based on previously validated methods for systematic reviews [32]. At each step in the search protocol, the titles, abstracts and available data were appraised by two independent reviewers (LA and EP), and compiled and synthesized using standardized forms. During the title and abstract reviews, if either of the two reviewers considered the article relevant, it was included. Articles classified as relevant at the title review stage were downloaded for abstract and full-text evaluation. Data were independently extracted by two reviewers (LA and EP), then compared and consolidated for analysis.
Data, including sampling methods, HIV-1, HIV-2 and dual HIV-1 and -2 (HIV-1/2) infections with sample size and number of participants living with HIV, were detailed and coded by the two independent reviewers (LA and EP). Information was categorized by key population studied, sampling techniques, country or countries, sample size, number of study participants living with HIV and notes. Discrepancies in abstracted data from the two reviewers were assessed by a third reviewer independently evaluating the article (SB), as was the final consolidated database (CH).

Results
Our search generated 995 citations, including 885 unique titles with dates of publication from 1987 to 2013 (Figure 2). Based on the inclusion criteria, 122 full articles were reviewed for data extraction, and 76 of these contained relevant data for at least one of the key populations defined. HIV prevalence data for at least one key population existed in 13 of the 24 countries included in the search (54.2%). Eleven of these countries were located in West Africa, and two countries were in Central Africa (DRC and Cameroon).
The majority of publications were assessments regarding FSWs (78.9%, 60/76), and another 10.5% (8/76) provided HIV prevalence data for their clients. One publication provided prevalence data for FSWs and well as clients of FSWs in Togo [33]. Thus, 90.8% (69/76) of the publications included in this study were related to FSWs, representing 41,270 FSWs across 13 countries and 5,986 clients of FSWs across 6 countries.
Two countries (Senegal and Nigeria) had published HIV prevalence data among MSM, and one seroprevalence study was conducted among male sex workers (MSWs) in Côte d'Ivoire, which was included in the MSM pooled data for analysis [34]. A total of six publications combined for the three countries were found for MSM (7.9%, 6/76), and one publication was available with HIV prevalence data for PWID,    1987,1988,1992,1995,1995,1997,1998,2000,2002 1989,1993,1993,1993,1998,2002,2008,2011,2012 Table 1 show a pooled HIV prevalence for the relevant key population(s) in each country, the 95% confidence interval (CI), and the date(s) of the publications retrieved per country. We include both HIV-1 and HIV-2 infections in the pooled prevalence data for the country, and, when possible, we display the division of HIV-1, HIV-2 and HIV-1/2 infections. The far-left data column in Table 1 displays the overall HIV prevalence among reproductive-age adults (15Á49) per country as reported by UNAIDS' most recent country-level surveillance data [6].  N 0 4,612) in Senegal. The pooled prevalence found among clients of FSWs was 7.3% (95% CI 6.6Á8.0) ( Table 2). Six countries had at minimum of one study reporting prevalence data for this demographic, with publications as early as 1992 and as late as 2009 (Table 1).

Men who have sex with men
While this review revealed a paucity of data for MSM, the pooled HIV prevalence in this review was 17.7% (95% CI 16.5Á18.9) for MSM in WCA (Table 2). No studies included were published earlier than 2005, and all but one were published after 2010. Three relevant Nigerian studies showed a pooled prevalence of 15.1% compared to 3.2% in adults of reproductive age [6,93Á95,106]. Senegal's pooled prevalence was 21.7% compared to 0.5% in the adults of reproductive age [6,104,105,107]. The study conducted in Côte d'Ivoire among MSWs reported 50.0% prevalence among a sample of 96 men in Abidjan [34]. Snowball, convenience, purposive and respondent-driven sampling were the primary recruitment methods used to obtain these data.

People who inject drugs
One study included directly sampled PWID. The study found a slightly higher prevalence of HIV at 3.8% (95% CI 2.8Á4.8), compared to 3.2% in the general population in Nigeria [6,92]. The sample was recruited through respondent-driven sampling and mainly compromised of men (90%) [6,92].

Limitations
This study was conducted as a systematic review to understand the prevalence of key populations in WCA and compare historical HIV prevalence to general-population statistics. Data were obtained from peer-reviewed literature, and while this ensures some quality control, we acknowledge that some relevant data that exist in grey literature and other programmatic data may have been overlooked. Programmatic data were not included in this review as it was not possible to implement a standardized assessment of the quality of the methods used and to ascertain an overview of research sampling and testing methods. However, the grey literature obtained through this review played a key role in the contextual analysis and discussion section of this study. Certain limitations also include the use of only English, French and Spanish, as other publications in other languages may have relevant data not captured in these inclusion criteria.The study among MSWs from Côte d'Ivoire was included in the overall analysis; however, the sampling method directly recruited these individuals from an established sex worker clinic, and thus HIV prevalence may be overestimated in this subpopulation. Also, while the authors noted that the majority of MSWs in the Abidjan area were MSM, they did not collect data on types of partner [34]. The contextual description from the authors is supported by evidence from other contexts where partners of MSWs are male [108,109]. Concurrently, systematic review methods were applied; however, sensitivity analysis and meta-analyses were not utilized. While odds ratios or aggregated comparison data were not generated, the overall analysis provides an overview of HIV prevalence among key populations and details of the epidemiology of key populations since the debut of HIV research in this region.

Discussion
Epidemiologic literature over the past 30 years has demonstrated a consistent and disproportionate burden of HIV among key populations in WCA. From the first published study in 1987 to the most recent in 2013, elevated levels of HIV among FSWs and their clients were consistently reported. In recent years, studies emerged to display an elevated burden of HIV among MSM within the region, although the number of studies in this subpopulation remains limited. Concurrently, there is nascent but growing evidence of the existence of PWID and, consequently, HIV infections in this subpopulation [92].

HIV prevalence
The elevated HIV prevalence among MSM, FSWs and clients of FSWs is important based on the determinants of the HIV epidemic in WCA and even more broadly across SSA. Surveillance has shown that women carry the highest burden of HIV on the continent, with national-level statistics constantly reporting that women have a higher HIV prevalence and incidence than men [13,110]. While programmes are designed to address the various risks associated with female HIV acquisition, the results of this study demonstrate that HIV risks are significantly higher among FSWs than women who do not sell sex in WCA. These results are substantiated by a systematic review of FSWs in low and middle-income countries, which showed that FSWs in SSA have a pooled prevalence of 36.9% (95% CI 36.2Á37.5) with a background prevalence on the continent of 7.42% in females [15]. Globally, FSWs were 13.5 (95% CI 10.0Á18.1) times more likely to be living with HIV than women of reproductive age [15]. Thus, the results of this review and the epidemiology of HIV among FSWs worldwide suggest that inclusion of and significant focus on these women and their clients are of importance to address these populations' high HIV acquisition and transmission risks in WCA [11,72,81,84,96]. On a continent where women are disproportionately burdened with HIV, prevalence of 17.7% (95% CI 16.5Á18.9) among MSM demonstrates a potentially concentrated epidemic in this key population. A prevalence of 7.3% (95% CI 6.6Á8.0) in clients of FSWs is also elevated compared to the general male population of the region and calls into question prevention programmes targeting this population. For clients of FSWs, male acquisition is linked to behavioural risks associated with multiple sexual partners, limited condom use and concomitant infection of an STI, amongst other determinants that are specific to men who engage in transactional sex [16,74,111]. For MSM, recent research has emerged that displays the increased transmission of HIV during anal sex, as well as sexual role versatility during samesex practices that increases individual HIV risks and drives transmission within sexual networks of MSM [21]. Thus, the acknowledgement of a heightened burden of disease in these populations is important for the design and implementation of specialized HIV prevention, treatment and care programmes regionally [16,26].
The heightened HIV prevalence in the MSM community found in these results is not unexpected, although the lack of data in WCA is noteworthy. The high prevalence reported in this review is comparable to other continents, with research indicating that MSM around the world are 19 times more likely to be infected with HIV than their adult male counterparts [18]. Interestingly, same-sex practices in WCA were reported as early as 1996 in a published population-based review [112]. The authors noted that the cumulative number of positive cases had exponentially increased from 1985 to 1995, and the primary modes of transmission were heterosexual practices (73.0%), homosexual practices (0.8%) and mother-to-child transmission (6.0%) [112]. More recent behavioural studies equally noted homosexual behaviour in different demographic studies. In Nigeria, 11.4% of sexually active secondary school students reported same-sex practices, and 12.4% reported anal sex [113]. In two Ghanaian studies in 2006 and 2008, prison inmates reported same-sex practices or identified as homosexual at 30.8% and 29.5%, respectively [114,115]. While sporadic reports of same-sex practices and elevated HIV prevalence have been reported in the region, there is limited targeted programme activity for these men [5,116,117]. What does exist is limited in scale, based on community-driven initiatives, and functioning in highly stigmatized settings [33,117,118].
While HIV prevalence in PWID was found to be relatively low, the Nigerian study provides two important details for programming in WCA. Firstly, while it has generally been assumed that PWID constitute a minimal presence in WCA, the study's ability to generate a sample size of 1459 through respondent-driven sampling indicates that this population does exist. Secondly, while HIV prevalence appears low, we know from other contexts that once HIV is introduced into this specific subpopulation, the possibilities of rapid spread and sustained transmission are great [119,120]. Contextually, policy makers are becoming aware of an increase of drug trafficking in the region, with large quantities of drugs confiscated in the past few years, and the recent conflict in Mali ascribed mainly to this trade [121]. Further supporting evidence of this regional trade was found in behavioural data in prisoners. In the same Ghanaian study in 2006, 41% of inmates reported imprisonment for narcotics; 7.3% had used cocaine, 5.2% heroin and 4.2% phencyclidine [114]. In the 2008 Ghanaian prison study, 35% of 1336 prisoners reported ever injecting drugs [115]. As was seen in Afghanistan as well as Thailand, Cambodia and other Southeast Asian countries, migration, trafficking, drug use and the HIV epidemic are intrinsically linked [119,120,122]. Thus, this is an important population to identify and appropriately engage in WCA in the coming decade of HIV prevention and control.

Historical perspective
This review also indicates that knowledge of HIV prevalence among key populations and the proportion of HIV infections attributable to key populations in WCA are not representative of new or changing dynamics of HIV transmission. In 1995, Djomand et al. noted that the male:female ratio of HIV infection in Côte d'Ivoire had declined over time and the gender ratio had shown females to be 4.8 times more likely to be infected than men in 1988, compared to 1.9 times more likely in 1991 [20]. The authors asserted that this decline displayed that the HIV epidemic was initially concentrated in a core group of FSWs and their male partners, and was potentially expanding in broader populations with less identifiable risk factors, similar to dynamics observed in other regions outside of SSA [20,122Á124]. In 2004, Côté et al. conducted a study of adult males (15Á59) in Accra, Ghana, and attributed 84% of existing cases of HIV to sex work and other transactional sex [125]. A study in 2008 based on Demographic and Health Surveys across four countries in SSA, including Ghana, showed that men who ever paid for sex were more likely to have HIV than men who had not (odds ratio 1.89, 95% CI 1.57Á2.28) [126].
In the capital city of Lomé, Togo, researchers estimated the attributable fraction of current HIV cases to sex work and other transactional sex was 32%, in contrast to only 2% of cases outside of Lomé [18]. Finally, recently in Nigeria, a modes of transmission study asserted that 23% of HIV infection was attributable to key populations, including 10% of new infections amongst MSM [93]. Despite high HIV prevalence among key populations and a high number of HIV in 2009, cases attributable to behaviours such as sex between men and sex work, systematic prevention and treatment programmes for key populations have not been implemented regionally [5]. While prevention programmes for FSWs and their clients have been noted in countries including Ghana, Côte d'Ivoire, Nigeria and Cameroon, the appropriate scale of these programmes and collected surveillance data are limited, and HIV prevention, treatment and care programming for key populations has failed to become a standard of best practices in the region [5]. Economic and regional migration Underlying dynamics of the epidemic indicate external, economic and urban-centred disparities have contributed to the complexity of the HIV epidemic in WCA over time. Domestic and international migration patterns were repeatedly reported and significantly mirrored economic crises and fluctuations in specific countries. For example, a study in Côte d'Ivoire documenting the FSW population that accessed health clinics between 1991 and 1998 noted a major shift in country of origin over time, with Nigerian women surveyed increasing from 2 to 56% between 1992 and 1998, and Ghanaian women decreasing from 82 to 9% in the same time period [29]. Other studies reported the migration of Ghanaian FSWs to other countries in the 1990s and asserted that the significant economic and political crises in the country at that time contributed to this migration [3,35]. The proportion of Liberian FSWs included in the same Ivorian study was shown to have increased from 0% in 1992 to 15% in 1995, and then to have declined to 2% in 1998 [94]. This evolution reflects the first internal conflict experienced in Liberia in the 1990s (1989Á1996) [127,128]. In a study reviewing the spread of HIV among FSWs in four cities across SSA, researchers noted that Cameroonian FSWs were more likely to have migrated internally to urban centres, while in Benin 86% of the FSWs sampled were from another country [41]. The only MSM study to discuss countries of origin was the MSW study in Côte d'Ivoire. Of the 96 MSWs sampled in Abidjan, 7.3% (7/96) reported a different country of origin [34].
The importance of these findings is revealed in the HIV prevalence among immigrants in the various studies. Nigerian and Ghanaian FSWs in the 2002 Côte d'Ivoire study were 1.03 (0.47Á2.23) and 3.69 (2.28Á5.97) times more likely to be infected than their counterparts from Côte d'Ivoire, Liberia and other West African countries [106]. In Lomé, two-thirds of FSWs were immigrants, and Ghanaian FSWs were 1.68 (1.06Á 2.66) times more likely to be living with HIV [126]. Addressing the needs of migrating populations at risk for or living with HIV is crucial, as these populations have less access to health services, are less likely to understand their human rights, and are more likely to contract a disease [129]. These populations are also more likely to be mobile; thus, successful prevention services for immigrant or mobile FSWs could potentially have an important impact in the overall reduction of HIV transmission and acquisition in the region [129].
Concurrently, disparity of HIV prevalence per locality was repeatedly reported in the various studies reviewed. In the same study that cited higher HIV levels among Ghanaian FSWs in Lomé, the prevalence among Lomé FSWs in 2005 was reported at 45.4% compared to 17.7% in the rest of Togo [18]. In two studies in Benin, there was significant spatial variation in the burden of HIV. For example, a study conducted in six cities in 2005 showed prevalence for HIV as high as 48.2% in Parakou, compared to 16.4% in Abomey/Bohicon [36]. A similar study found HIV prevalence in Cotonou, Benin, among FSWs to be 38.5%, compared to a pooled prevalence in three other large cities of the country of 58.9% [35]. Therefore, from an HIV prevention perspective, cross-border initiatives, effective community-based networking and standardized programmes across urban and regional landscapes for key populations are relevant for the WCA region.

Ways forward
Our review makes clear that there is a significant gap in the literature and subsequent HIV programmes for key populations in WCA. This may be ascribed to the application of the HIV response model of SSA to WCA epidemiological and prevention approaches. However, as reports of high HIV prevalence among key populations have existed in the literature since 1987, it also calls into question the structural barriers to healthcare for populations that engage in these defined sexual behaviours in this region. As in other contexts, sex work and other transactional sex, same-sex practice and drug use are either criminalized or highly stigmatized in this region, and public policies have ignored or generally declined to address the specific health needs of key populations [5,130,131]. Research has shown that macro-level policies that impede or deter health service delivery for key populations ultimately increase vulnerability to disease acquisition [23,130,132].
Data presented here provide a useful framework for HIV programming in the region. The inclusion of relevant sexual history and behavioural questions in large-scale surveillance surveys, such as DHS, may also be of benefit in obtaining a better overview of the epidemiology of key populations, both in WCA and worldwide. While the delivery of sensitive questions such as engagement in sex work, transactional sex, same-sex practices and drug use must be carefully administered (ideally not within the household setting), standardized national data collection would go far to inform country and regional policy development in WCA.
Subsequently, emerging data have shown that addressing the epidemic in key populations requires combined behavioural, biomedical and structural approaches [23,133]. Limited condom use with regular sexual partners, unawareness of HIV status and co-infections with genital ulcerative diseases are contributing factors to heightened prevalence [10,21,116]. High prevalence among key populations concurrently has implications for prioritized biomedical interventions [21,134].
While the knowledge that these populations have a higher risk for transmission and acquisition of HIV and other STIs is acknowledged, the method in which prevention and treatment programmes address these risks has yet to be firmly cemented in HIV prevention programming [13]. Researchers in the United States and elsewhere have demonstrated the importance of engaging populations in the continuum of HIV care Á from undiagnosed cases to testing and diagnosis, followed by linkage to ongoing care and treatment [135]. The continuum of HIV care significantly reduces the viral load among people living with HIV and ultimately reduces transmission [135,136]. In two recent studies in the United States, researchers found that due to advances in antiretroviral regimes, with 70Á80% adherence to antiretroviral therapy (ART) by participants, durable viral suppression occurred in most individuals, lowering the possibility for onward HIV transmission [136,137]. The findings indicate that the key to community viral suppression is early diagnosis of the disease, well-developed referral systems to clinical services, and care and support programmes that encourage adherence and access to treatment [136]. This approach has been shown to be effective in contexts with both high and low prevalence, and recent research from South Africa affirms that adequate ART coverage at the community level reduces incidence over time [138]. Thus, prevention programmes are beginning to show that distribution of prevention commodities and messages should be in concert with interventions that address the virology and biomedical aspects of care and treatment [135]. This is even more relevant for key populations who carry a significant burden of disease and ultimately are people living with HIV.
Structural factors acting at the macro-and meso-levels should not be ignored in WCA and are essential when building combination biomedical programmes [23,139]. Criminalization and public policy neglect substantially inhibit key populations' ability to access appropriate, life-sustaining and preventionoriented health services. Policy-level gaps and communitylevel stigma must be addressed if programmes are to adequately confront the needs of these populations [140,141]. Studies from other countries on the continent indicate the stigma experienced within their communities and at health services, significantly deters the uptake at clinical services for key populations [130,142]. Public policies that adequately address the intricate health needs, reduce stigma and discrimination, and facilitate community and provider level HIV care and treatment delivery will highly benefit the overall control and prevention of HIV among key populations in WCA [23].

Conclusions
This systematic review suggests that the concentrated HIV epidemic in WCA more closely resembles the epidemics in Southeast Asia and Latin America than those in the rest of SSA. This not only calls into question the response to the HIV epidemic in WCA but indicates that the region has an opportunity to adapt and develop region-specific prevention and treatment strategies. Targeted, cost-effective programmes that address not only behavioural but also biological and structural risk factors associated with HIV acquisition and transmission key populations should be engaged to reduce the onward spread of HIV. Prevention programmes should model strategies on appropriate programmes that reduce community viral loads, increase uptake of treatment among key populations and address the barriers to healthcare that exist in highly stigmatized settings. Ensuring that programmes rooted in community-based approaches address the continuum of HIV care, from diagnosis to viral suppression, will be a challenge but also a possible victory for HIV prevention and control in WCA.

Introduction
Globally, it has been observed that HIV prevalence among men who have sex with men (MSM) significantly exceeds HIV prevalence in the general population, even in the context of generalized epidemics [1Á3]. Across sub-Saharan Africa, HIV prevalence is estimated to be approximately 5% in the general population and 17.9% among MSM [1]. The few published studies from West Africa consistently report higher HIV prevalence among MSM than in the general population, with HIV prevalence estimates of 13.5% among MSM in Nigeria, 16.3% in Burkina Faso and 21.8% in Senegal [1,2,4Á6]. Individual-, network-, community-level and policylevel factors noted to contribute to the higher risk of acquisition and transmission of HIV and other sexually transmitted infections (STIs) among MSM have been found to be prevalent in Central and West Africa [5,7,8].
With over 550,000 people living with HIV in Cameroon, the prevalence of HIV among reproductive-age adults in A preliminary analysis of data reported here was presented at the 7th International AIDS Society Conference, which was held from 30 June to 3  Cameroon is estimated to be 4.3%, which represents a mature and widespread generalized epidemic [9,10]. In Douala and Yaoundé, the two largest cities of the country, HIV prevalence among reproductive-age adults is estimated to be 4.6% and 6.3%, respectively [10].
MSM were recently listed as a priority group in the Cameroon government's ''National Strategic Plan for HIV, AIDS, and STIs: 2011Á2015,'' along with goals including strengthening HIV-prevention programmes and building capacity for HIV health services that serve MSM [11]. The higher biological risks of HIV acquisition and transmission associated with unprotected anal intercourse (UAI) compared to other forms of sexual intercourse make MSM an important target population for HIV-prevention efforts [12]. However, only one HIV prevalence estimate from programmatic data in Douala is available to date for MSM; in this 2007 study, which used convenience sampling, HIV prevalence was estimated to be 18.4% [13].
Established individual-level risks for HIV acquisition and transmission among MSM in the region that are modifiable include UAI, inconsistent use of condom-compatible lubricants (CCLs), a high number of male partners, drug use and syphilis co-infection [1]. In a recent study, UAI in the past 6 months was frequent among MSM in Douala, as was having one or more female sexual partners [7]. Bisexual concurrency and bisexual partnerships among MSM have been observed in studies in Nigeria, Senegal and southern Africa [5,14,15]. Inconsistent condom use with male and female partners was common among MSM in one Togo study, and in a study conducted in Nigeria, it was associated with prevalent HIV infection, as was having been the receptive partner in anal intercourse in the past 6 months [5,16]. Other factors associated with prevalent HIV infection among MSM in Nigeria and Senegal were older age and having a symptomatic STI [5,15].
Network-level factors that may impact HIV-transmission risk include sexual network size, STI prevalence, levels of peer education, knowledge of HIV status within the population and network tendencies for drug use or transactional sex [1]. Community-level factors that may contribute to HIV risk include high community viral load and suboptimal coverage or uptake of healthcare services [1]. Additionally, the social stigma surrounding HIV, sexual identities and homosexuality in Cameroon may deter MSM from seeking voluntary HIV counselling and testing (VCT) or other health services [17Á 20]. Perceived stigma, including fear of seeking healthcare and refraining from disclosing same-sex practices to a health professional, and enacted discrimination, including denial of healthcare access based on sexuality, were frequently reported by MSM in Senegal and southern Africa, and were associated with increased sexual risk practices and prevalent HIV infection [21Á26]. Similar to most countries in sub-Saharan Africa, sexual relationships between men are both criminalized and highly stigmatized in Cameroon, and prosecution can result in up to 5 years of imprisonment [8]; physical violence from law enforcement is also a reality for some MSM, posing challenges to HIV programming [8,27Á29].
In light of the unique needs of MSM within generalized epidemics, and the limited data available on this vulnerable population in Cameroon, we aimed to describe the socio-demographic and behavioural characteristics of MSM in Douala and Yaoundé, determine the age-stratified HIV and syphilis prevalence in both cities, and investigate the individual-, network-and community-level factors associated with HIV infection among this population.

Study population
This cross-sectional study was conducted in AugustÁSeptember 2011 at two community-based organizations (CBOs) that provide targeted services to MSM: Alternatives-Cameroun in Douala and the Cameroon National Association for Family Welfare (CAMNAFAW) in Yaoundé. The interviewers were MSM community volunteers from Alternatives-Cameroun, Humanity First and CAMNAFAW. The MSM sensitivity trainings for interviewers were conducted at the Association Camerounaise pour le Marketing Social (ACMS) conference rooms in Douala and Yaoundé. Men aged 18 years or older who reported engaging in penileÁanal or oral intercourse with another man in the past 12 months were eligible for the study. Participants were recruited using respondent-driven sampling (RDS) [30], a sampling technique that enables estimation of proportions and regression modelling while controlling for non-random social network structures that bias peer-based recruitment. Seven seeds heterogeneous in sexual identity and sexual role preference were selected through existing community contacts to begin the recruitment process in each city. Upon enrolment in the study, all individuals were given three uniquely coded coupons to refer other MSM to the study. The CBOs worked with the research team to identify the initial seeds, screen study participants for eligibility and interview participants after receiving informed consent.
Sample-size calculations were based on the ability to detect a 15% change in the prevalence of condom use at last anal intercourse over time from 60% at baseline, with a design effect of 2, a significance level of 0.05 and a power of 80%, yielding 241 men for each city.
All participants provided written informed consent. The study was approved by the Cameroon National Ethics Committee, and the secondary analysis of the study data was approved by the Johns Hopkins Bloomberg School of Public Health.

Data collection
Participants completed an interviewer-administered structured questionnaire containing questions on: socio-demographics; network size; sexual behaviours, including condom and lubricant use (always vs. often, sometimes or never); experiences of STI symptoms; access to CBO-run MSM centres (which included outreach services); access to free condoms; VCT experiences; knowledge of HIV transmission, prevention and treatment (a composite score from 13 questions); and perceived social support for condom use (a composite score from eight questions, including support from partners, family and peers). Interviews were conducted in French or English, and they were recorded in French.
After participants received pre-test counselling, approximately 4 ml blood specimen was collected from them by a Global Viral Cameroon phlebotomist and tested to confirm  16.4.18752 HIV and syphilis serostatus, followed by post-test counselling on the same day. Men who screened positive for HIV or syphilis were referred to appropriate health services. All participants were reimbursed 1000 CFA franc (US$2) for completing the questionnaire and an additional 1000 CFA franc (US$2) for each peer referred into the study. All participants received free VCT, condoms and CCLs. Participants were also given access to peer education, support groups and linkage to HIV care.

Laboratory testing
Specimen processing and testing were conducted by staff from Global Viral Cameroon at the field sites. The national HIV surveillance algorithm for second-generation surveillance of HIV, adopted by the Ministry of Public Health of Cameroon, was used to measure current HIV status, including Determine † HIV-1/2 (Inverness Medical, Chiba, Japan) and Human HEXAGON HIV 1 ' 2 (Human GmBh, Wiesbaden, Germany). All indeterminate and positive samples and 15% of the negative samples were transferred to the Global Viral Cameroon Yaoundé laboratory for fourth-generation HIV enzyme-linked immunosorbent assay (ELISA), which detects antibodies to HIV-1/2 and the p24 antigen (whose presence indicates a possible seroconversion). Screening for syphilis was performed according to the national algorithm in Cameroon using Rapid Protein Reagin (RPR; SGM Italia, Roma, Italy) and Treponema pallidum haemagglutination assay (TPHA; Fortress Diagnostics Limited, Antrim, UK). Global Viral Cameroon was responsible for blood specimen collection, laboratory testing and serology data management.

Statistical analysis
ACMS and CARE International-Cameroon managed study data. Questionnaire data were double entered into the CSPro (version 4.0) software, exported into SPSS for data cleaning by ACMS and then exported to Stata/SE (version 11.2) for data analysis.
To minimize biases associated with chain referral sampling, weights were created in Stata/SE version 11.2 using the RDSII estimator to account for the effect of differences in the social network sizes of participants. Weights were based on the transition matrix for the dependent variable, current HIV status. Network size was assessed using the response from the latter of two questions: ''How many men who have had oral or anal sex with men in the last 12 months do you know, who also know you and live in this city?'' and ''among these men that you know personally, how many of them are 18 years and older?'' Homophily (range: (1 to '1) was assessed to evaluate the preferences of individuals to recruit MSM with the same HIV status [31].
Bivariate logistic regression models were used to estimate the unadjusted association between HIV infection and covariates selected based on our knowledge and the published literature. RDS-weighted prevalence and bootstrapped confidence intervals were calculated for all variables explored in regression modelling. Multivariate logistic regression models were built to estimate the adjusted association between current HIV status and covariates, with age forced into all models regardless of statistical significance. The Akaike information criterion (AIC) was used to favour the most parsimonious models. Bivariate and multivariate logistic regression models were also built with RDS weighting. p-values B 0.05 were used to indicate statistical significance. We further compared the associations between binary covariates using the Pearson chi-square test.

Results
A total of 295 men were screened in Douala, of whom 272 participated. In Yaoundé, a total of 246 individuals were screened, resulting in 239 participants. The median number of descendants per seed was 32 (range 6Á99) in Douala and 31 (range 2Á88) in Yaoundé. In Douala, the median number of waves per seed was 6 (range 1Á8); homophily for HIV status was (0.04 among the HIV-negative group and 0.06 in the group living with HIV. In Yaoundé, the median number of waves per seed was five (range 1Á9); homophily for HIV status was 0.004 for the HIV-negative group and 0.06 for the group living with HIV. In both samples, RDS network homophily was close to 0, which may indicate a close approximation to random recruitment. The majority (77.9%; 398/511) reported that they would have given a coupon to their recruiter (an indicator of the reciprocal ties assumption [32]).
Overall, the median age was 24 years (range 18Á51, interquartile range (IQR) 21Á28). In both cities, the majority had completed secondary education and were single. Sixtytwo percent of MSM in the overall sample identified as bisexual, compared with 28.6% who identified as gay or homosexual and 9.8% as MSM or other. Ninety-eight percent of all participants reported having penile-anal intercourse in the past 12 months. Median age of sexual debut with another man was 19 (IQR 17Á22) ( Table 1).
Responses to questions on health service uptake, HIV knowledge, social support and sexual practices are presented in Table 2. Men in Yaoundé were much less likely to access CBO services targeting MSM than men in Douala (33.7% vs. 66.1%, x 2 (1), pB0.001). No difference was observed in ever receiving free condoms (74.2% vs. 68.8%, x 2 (1), p 00.2). In both cities, a large proportion of men reported sex with males and females (46.2%) and experienced STI symptoms in the previous year (34.6%). Inconsistent use of condoms with regular male partners (64.1%; 273/426) and casual male and female partners (48.5%; 195/402) was common, as were condom slippage and breakage (43.7%; 216/494). Ninety percent of MSM who used condoms also reported using lubricant. Of these men, 26.3% (124/472) reported using lotion, saliva, Vaseline or other condom-incompatible lubricants.
As presented in Table 3, crude and RDS-weighted HIV prevalence were 28.6% (73/255) and 25.5% (95% CI 19.1Á 31.9) in Douala and 47.3% (98/207) and 44.4% (95% CI 35.7Á 53.2) in Yaoundé. Age-stratified prevalence is presented in Figure 1. In Douala, only 17 (6.3%) MSM refused to be tested; in Yaoundé, this number was higher (13.4%, n032). An association between having a history of VCT and refusing testing in the study was observed in Yaoundé, although it did not reach statistical significance (15.2% vs. 4.9%, p 00.08). Refusal was not correlated with age, education level, age of sexual debut, condom use, receptive sexual role preference, number of male sexual partners in the past 12 months or

Discussion
The high HIV prevalence and inconsistent use of condoms and CCLs observed in this study highlight that MSM are a   priority population for HIV prevention, treatment and care services in Douala and Yaoundé. Furthermore, these data suggest that HIV risks are not evenly distributed given the significant differences in HIV prevalence between cities and between MSM sub-populations [1]. The individual-level factors found to be associated with HIV infection indicate that future HIV programming and interventions in Cameroon should address both behavioural and structural hurdles relevant to MSM. Consistent with data from other countries of sub-Saharan Africa [15,16,24], condom breakage and slippage and inconsistent condom use were common among this sample. CCLs, which decrease the risk of condom breakage, were also used inconsistently [33], suggesting that increased access to quality condoms and CCLs is essential [34,35]. While maximizing the use of condoms and CCLs is necessary in decreasing HIV risks among MSM, likely it will not be sufficient to change the trajectory of the epidemic given the high transmission probability of HIV infection associated with UAI, as observed in other settings [1,34]. The prevalence of active syphilis was low, as observed in other countries in the region [4,5]; however, a high proportion of participants reported experiencing STI symptoms, highlighting another network-level risk factor potentiating HIV transmission within the sexual network. Increasing the capacity for routine STI diagnosis, particularly for genitourinary infections, and linkage to treatment tailored towards MSM should be incorporated to support HIV-prevention programmes [15,36].
MSM in Douala who reported a preference of being the receptive partner during anal intercourse were more likely to identify as gay and be living with HIV. This not only affirms existing data demonstrating the increased HIV acquisition risk associated with unprotected receptive anal intercourse [37] but also echoes previous studies conducted in African settings in which self-reporting as gay was associated with higher odds of living with HIV compared to other MSM in the African setting [24,38]. Given that antiretroviral pre-exposure prophylaxis (PrEP) and rectal microbicides have been identified as research priorities for African MSM [39], and that rectal microbicides are currently in Phase II trials that are enrolling MSM from the African continent [34], evaluating the feasibility of novel biomedical interventions for sub-populations of MSM in Cameroon with significant HIV acquisition risks may be appropriate [34,40,41]. However, the cost-effectiveness of implementationing such biomedical interventions requires further research [42]. In addition, exploring increased antiretroviral therapy (ART) for MSM living with HIV likely represents an important strategy for preventing the transmission of HIV to sexual partners. However, the limited availability of ART for people living with HIV who are currently eligible for treatment, which has been documented in Cameroon, also needs to be addressed in order for ARTbased strategies for people at risk for the acquisition or transmission of HIV to be effective [43].
A significant proportion of the MSM in our sample were living with HIV by the age of 18Á23, indicating a high risk for HIV acquisition for men under 18 in these settings [24,44]; however, men under 18 have traditionally been excluded from HIV surveillance and prevention programmes [1]. Confidential youth sexuality counselling hotlines, web-based education and social marketing campaigns may be useful in reaching younger MSM with HIV programmes [36,37].
While our study did not include a detailed assessment of social stigma, other studies have demonstrated that stigma limits the provision and uptake of HIV prevention, treatment and care for MSM in the region [18,19,27]. Uptake of services delivered by targeted CBO providers such as Alternatives-Cameroun in Douala was high in our study, suggesting that community-based approaches can spread information-leveraging networks of MSM despite the contextual barriers. There was limited uptake of services in Yaoundé and higher refusal of HIV testing in the study; to the best of our knowledge, MSM-tailored HIV programmes were new and in development at the time of this surveillance project. The historically limited services may partially explain the higher HIV prevalence observed among MSM in Yaoundé as compared to Douala, although these participants also tended to be older and report more male partners, drug and alcohol use, and STI symptoms.
Data on the proportion of MSM living with HIV who were eligible for treatment, or who were actually on treatment, were not collected in this study. However, consistent data highlight the importance of addressing the needs of people living with HIV, including linkage to care, to optimize their own health and prevent onward transmission to other men and to women [45]. In Cameroon, only half of all patients eligible for treatment are estimated to be receiving ART, and ART stock outages at health facilities are frequent [43]. Given the significant stigma and discrimination that have been documented as affecting MSM in Cameroon, MSM living with HIV may be at higher risk of being unaware of their diagnosis or not achieving viral suppression [8,20,46]. MSM community groups have long been known to play essential roles in the HIV response, and the data collected here suggest that community-driven approaches should be scaled up to increase uptake of VCT and support linkage to HIV care, treatment and adherence support for those eligible [47,48].
The cross-sectional design of this study does not allow us to assume causality of the associations present in the data. There are several limitations to the generalizability of the   HIV prevalence estimates reported in this study, which included individuals who reported receptive or insertive anal intercourse in the past 12 months. The generalizability of the results for MSM living in smaller urban centres and rural settings is unknown given that recruitment occurred in two large cities. Similarly, as our sample was predominantly young and educated, the results may not pertain to older MSM or individuals with lower educational status. Future studies to address these gaps could be conducted. The modest sample size may have reduced our statistical ability to detect other associations [49]. Due to the high refusal of HIV testing during the study in Yaoundé (13.4%), we were unable to assess the potential for bias in the HIV prevalence estimate from this city. However, RDS network homophily was close to 0, which may indicate minimal recruitment bias based on HIV status. Data on self-reported HIV status and the percentage of undiagnosed men were not available, which limit our interpretation of the association between knowledge of one's own HIV status and behavioural factors such as inconsistent use of condoms and CCLs. This requires further investigation in future studies. Although non-significant, the positive association between having been tested and refusing testing may suggest that individuals who are already aware of their HIV status may be underrepresented in our study.

Conclusions
These data provide results that can be integrated into HIV programmes for MSM in Cameroon and highlight the importance of targeted HIV prevention, treatment and care services that address all levels of HIV risk. Coordinating behavioural, biomedical and structural interventions, and supporting the work of local CBOs, will be keys to ensuring that HIV-negative MSM receive regular VCT and appropriate prevention services, and that MSM living with HIV are effectively engaged in the continuum of HIV care. Success in the continuum of HIV care necessitates addressing the barriers to the uptake of care, such as concerns about confidentiality and healthcare-related enacted and perceived stigmas [8,20,36]. Protecting the dignity and rights of MSM in healthcare settings and beyond allows for a safe environment for individuals to receive optimal care to protect themselves and their partners [27,29]. Monitoring the success of the next generation of HIV-prevention approaches will require innovative implementation science exploring changes not only in individual-level risks, community viral load and HIV incidence, but also in social and policy-level factors including stigma, discrimination, violence and criminalization.
Experiences of Kenyan healthcare workers providing services to men who have sex with men: qualitative findings from a sensitivity training programme  [8], with the aim to support more inclusive health services for MSM [9]. Implementation of Kenya's AIDS policies requires the ability of healthcare workers (HCWs) to deliver appropriate and sensitive services to MSM patients. Effective HCWs must have accurate knowledge of the sexual health issues of MSM, non-prejudicial attitudes and behavioural skills to treat MSM patients [10]. However, HCWs in Kenya, as elsewhere in sub-Saharan Africa, rarely receive specialized training on how to provide care for MSM [11].
To address this gap in training service providers, Kenya's National AIDS and STI Control Programme (NASCOP) developed an education training programme to strengthen HCWs' skills and capacity to provide non-judgemental counselling and HIV healthcare services for MSM. The training programme incorporated two learning modalities: a computerfacilitated training programme covering eight modules [MSM and HIV in sub-Saharan Africa; stigma; identity, coming out and disclosure; anal sex and common sexual practices; HIV and sexually transmitted infections; mental health, anxiety, depression and substance use; condom and lubricant use; risk-reduction counselling] in addition to facilitated group discussions among programme trainees about the programme content and relevant clinical experiences working with MSM. Both learning modalities offer complementary approaches to educational training. Computer-facilitated training modules can offer a standardized and disseminable approach to improve HCWs' knowledge and health service delivery skills for MSM patients [12], especially in settings such as Kenya where access to formal medical education is constrained. Supplementing the computer-facilitated training with opportunities for peer discussion and support among HCWs can potentially enhance the transfer of standardized learning to the workplace [13]. van  We conducted a preliminary pre-post-evaluation of HCWs who participated in the programme [14]. Quantitative findings showed improvements in MSM-related knowledge and reductions in discriminatory attitudes towards MSM. Effects were most pronounced among HCWs who had low levels of knowledge and/or more extreme negative attitudes towards MSM at baseline, and among HCW in clinical roles within governmental settings.
This article reports data from qualitative focus groups with participating HCWs, conducted prior to and three months after completion of the programme. The objectives of this analysis are to explore: (i) how HCWs characterized their professional challenges in serving MSM patients prior to the programme, (ii) how HCWs described the impacts of programme participation on their personal attitudes and professional capacities and (iii) how the computer-facilitated educational training programme can be improved.

Participants and intervention procedures
The study was conducted between October 2011 and March 2012 in four districts in coastal Kenya: Kilifi, Kilindini, Malindi and Mombasa. To recruit trainee participants, NASCOP issued announcements to 49 health facilities providing antiretroviral treatment in the four targeted districts. Announcements described the study as a two-day residential programme involving computer-facilitated training and group discussions on HIV and MSM. Volunteer participants completed informed consent procedures, and those who enrolled received 2000 Kenya shilling (approximately US $24) for travel and lodging adjacent to the training facility in Kilifi.
Participants were 74 HCWs from the four target districts. Fifty were females and 24 males, including 22 clinicians, 43 nurses and counsellors, and nine were administrators/ managers. The average age was 34. All participants identified as Kenyan, 84% as Christian and 15% as Muslim. Eighty-six percent had no previous training about MSM or anal sexual practices. Three participants (two females and one male) were transferred to health facilities outside the study area after the initial training and could not participate in the follow-up focus groups.
A total of four groups were convened to participate in the two-day residential training (one group per district), with 18Á19 participants per group. During Day 1, participants received a general overview of the programme, and each participant then independently self-administered the first four modules of the standardized, computer-facilitated training. Modules were designed to take up to two hours to complete. At the end of each module, participants answered a series of multiple-choice questions (up to 16 questions); to advance to the next module, participants were required to achieve a minimum score of 71% correct. After every two modules, participants engaged in a group discussion to reflect on the information and identify barriers and facilitators to improve on HIV prevention and other services for MSM in Coastal Kenya. A member of the research team facilitated group discussions. During Day 2, participants completed the final four modules and group discussions. At the end of Day 2, participants were asked to discuss work strategies for improving the quality of clinical care and HIV/ STD testing for MSM patients in their districts. Research team members included an MSM counsellor, a community liaison officer, a senior research counsellor and a social scientist; teams were supported by two MSM members from a local non-governmental organization. Research team members received a comprehensive three-day training on the intervention objectives and procedures, including didactics and role-play opportunities for discussion and problem solving.

Focus group discussions
Eight focus group discussions (FGDs) (each comprising 9Á10 participants; two focus groups per training) were conducted with participating HCWs prior to the training and were repeated three months following completion of the training. Focus groups were semi-structured and facilitated by a member of the research team, with a co-facilitator present to observe and take notes. Discussion topics included: identification of subcategories of MSM and their characteristics; sexual practices of MSM and risks for HIV and STI transmission; practices for sexual history taking and sexual health examination with MSM; risk-reduction counselling for MSM; personal values and attitudes towards MSM; strategies to improve communication between HCWs and MSM patients. Most discussions were conducted in English, although participants were also encouraged to speak in Kiswahili depending on their preference and language skills. All discussions were audiotaped, transcribed and entered into NVivo. FGDs conducted in Kiswahili were translated into English.
Analyses of qualitative data followed the ''framework approach'' described by Ritchie and Spencer [15], which involves systematic coding to identify and define concepts emerging from the data, mapping the concepts, creating typologies, finding associations between concepts and seeking explanations from the data. Data were coded by two independent research team members to ensure that interpretations of quotes were consistent and that data analysis was rigorous and transparent. The main concepts emerging from the data included: secondary stigma, professional training and service barriers to MSM patients; types of and justifications for social discrimination towards MSM in Kenyan culture; invisibility and silence about homosexuality in Kenyan culture; and subjective theories about the origins and nature of homosexuality. Differences among coders were resolved by group discussion involving other members of the research team.
The study procedures were approved by the ethical review board at the Kenya Medical Research Institute. All participants provided written informed consent for the FGD.

Results
Discussion of MSM-related attitudes, beliefs and behaviours before training Secondary stigma For most participants, secondary stigma was a dominant concern. Secondary stigma refers here to negative judgements from peers and community members for being associated with MSM. Participants cautioned that professional trainings van  You know MSM, as he had mentioned, are regarded as outcasts. Therefore, if you offer to treat them in your clinic, the community will perceive it as . . . the clinicians are also MSM.
Owing to this fear, many HCWs described minimizing the amount of time with MSM patients. For example, one participant described having a basic willingness to serve MSM patients, but would allocate the shortest time possible: The fear of being associated, that's what is making us spend as little time with MSM clients when they come to our facilities. You will hurriedly clear him out.
However, fear of secondary stigma was not consistently expressed by all members of the discussion. A small subset of participants who had previous education and sensitization on MSM prior to the training reported comfort in attending to MSM patients. Consequently, these HCWs had become MSM patient advocates and educators in their clinics prior to engagement in this research study: I was trained . . . on issues to do with MSM. Last week, I met an MSM client who was HIV positive. It was in one of our departments and the nurse was like, '. . . you are the person who deals with these kind of clients'. I told her to refer the client in my office . . . Actually, I had to take [my colleague] for an MSM training. Her attitude has really changed and she is a now a different person.
Inadequate professional training and resources Participants acknowledged having little or no education about MSM health. Indeed, prior to the training programme, many HCWs expressed a sense of denial about the existence of MSM. For example, one reported that: I tend to reason differently when it comes to MSM. I sometimes tell myself, no, this doesn't exist; this is not possible.
Across multiple discussions, others questioned whether MSM are present in their local communities: Some of us are really green, we just hear stories on internet that some men are having sex with other men but we have never had an interaction with the MSM. MSM are unheard of in the place I come from.
HCWs who acknowledged the presence of MSM patients in their clinics described feeling inadequately prepared to provide services. Those with prior experience consulting MSM patients described specific challenges in diagnosing and treating rectal STIs, and argued for more appropriate guidelines: Of late, it's only a few individuals who have been trained in our facility. We don't have a guideline, yet we see them daily. We have no idea on how to manage infections affecting men who have sex with other men . . . Most of the medical personnel are not sensitized on issues to do with anal STIs and they are also not indicated in the STI charts. They only specify about urethral discharge, cervicitis, urethritis in men, PID [pelvic inflammatory disease] etc. It doesn't mention the anus.
Lacking the knowledge, skills and treatment guidelines for rectal STIs, HCWs often relied on guesswork and assumptions. Participants recognized the likelihood of underdiagnosing or misdiagnosing rectal infections transmitted through anal sex.
And when we are counselling or probing them about sex, we only ask them, 'Do you usually have sex?' When they say yes, we don't probe further to know the type of sex i.e., we just assume it is heterosexual. The medics are also not trained and if an individual comes with an anal complaint, they assume that it is haemorrhoids and refer them for surgery.
HCWs described how limitations in assessment forms reinforce the invisibility of MSM in their clinics. By not collecting information about same-sex behaviour or anal sex practices, these topics are reinforced as taboo issues that warrant silence and discomfort.
Most of the tools and the working conditions are not accommodative for this line of sexual orientation. I have never seen a tool in the CCC [comprehensive care centre] or the TB clinic asking for the clients' sexual orientation. So it's like, 'I don't need to know of what you do' . . . Therefore, the tools should be designed to capture the sexual orientation of a person so that the health workers can have a feel that it is a part of the health issues and not a gossip.
Additional resource limitations for treating MSM were discussed. HCWs reported on the inconsistent supply of lubricants for use during anal sex, and also described how the physical structure of the health facility hinders their ability to provide privacy and confidentiality for sexual health consultations.
The MSM usually come to the clinics and ask for the lubricants or condoms but you will find that the lubricants are not available; it's only the condoms. I think there is no confidentiality because of the way our health facilities have been structured, van  i.e., someone can bump in while you are attending to a client. You could be talking of sensitive issues but other staffs won't bother. They will sit on the other side and do their stuff. So the client might not be free to open up.
Personal and social homophobia Many HCWs acknowledged holding prejudiced views towards MSM. A number of participants commented on how negative judgements towards MSM may influence the provision of services.
We perceive them negatively and feel that they don't deserve our services. Some health workers don't like to examine them. They claim that such infections are self-inflicted.
HCWs reflected on the influences of culture and religion on their treatment of MSM patients. When reminded of their professional obligation to provide effective services to all patients, they described internalized barriers that must be overcome.
I find it abnormal for a man to have sex with another man. It is both culturally and religiously unacceptable . . . Voices from religion or the community tell me that it is wrong. Professionally, I will have to handle that shock and look at possible ways of helping this person.
Participants reported a tendency to exhibit subtle forms of stigma and discrimination towards MSM patients, such as by maintaining body distance. Other times, HCWs explicitly showed disparaging treatment: When they seek medical assistance in our facilities, the same providers will shout, 'Look at him, he is telling me that he is having an anal STI; can you leave my room'. Instead of treating them with respect, they end up drawing their colleagues' attention.
However, some HCWs challenged those who expressed personal prejudice towards MSM. Participants who had prior exposure to MSM sensitization argued that HCWs have a professional duty and societal obligation to provide non-prejudicial services to MSM.
We as health workers feel that MSM issues need not to be discussed, they are regarded as outcasts. How then would we come up with a constructive discussion about people whom we feel should not be in the society at first place? In my opinion, I think this is the biggest obstacle. If we accept these people and treat them as our clients, then it will be of great help to the society. Post-training discussion of HCWs' attitudes, beliefs and behaviours Recognition of MSM in Kenya A pervasive theme in post-training focus groups was the explicit recognition of MSM in Kenya. Many reflected on how their prior denial of MSM behaviour, and their previous belief that anal sex among men was negligible in Kenya, had inhibited their capacity to provide services. Participants felt ''empowered'' by the training to address HIV and other health needs of MSM, as one stated: I didn't ever believe that MSM were in existence but the training empowered me with a lot of knowledge and information on how to probe about issues of anal sex.
Participants described how the training enhanced their understanding of the complex interplay between homophobia, community denial of MSM and HIV transmission. Some advocated to local colleagues for the acceptance of MSM and educated them about the biological and behavioural circumstances that place MSM at heightened risk for HIV infection. One participant described: I went and gave the feedback to my colleagues immediately after the training and some were as if they have never heard such a . . . They used to hear about it but they were not sure whether it was a real, whether such people exist. Therefore, I had to make them understand that the practice is in existence and that's nature.
Professional responsibilities as a health provider During follow-up focus groups, participants described their professional responsibility to treat all patients with equity and respect. They endorsed a basic value of professionalism and treating MSM patients to the best of their ability. For many, this required a suspension of personal judgement in order to provide effective care: As a professional, I am not supposed to segregate them, whether I support homosexuality or have a different perception or judgment. As a clinician, my duty is to treat without imposing my values on the patient. That's the positive thing I got from [the training program] and it's what I'm doing now. Some described witnessing discriminatory actions towards MSM in their facilities or observing breaches in patients' confidentiality. They reflected on how these experiences could foster distrust of HCWs and discourage MSM patients from seeking care when needed, thus perpetuating a cycle of HIV transmission. There was widespread consensus among group members that a concerted effort must be made to establish trusting rapport with MSM patients, and take extra care to employ discretion at all times. As one participant articulated: I think the problem is that, the individuals we have attended to still want to see if they can trust us, if we can respect their privacy . . . As for now, it will take time because they are trying to internalize on our missions towards them and they will come out once they are convinced that you don't have an ill motive towards them.
During the follow-up focus groups, HCWs were asked to reflect upon and share their experiences, that is, work practices and attitudes towards MSM in their respective health facilities, and to reflect on strategies to change van  discriminatory actions towards MSM in their health facilities. Many participants stressed the importance of separating personal and religious values from professional ethics for the sake of HIV prevention in Kenya. While some felt the training had helped to normalize same-sex relations, others adamantly affirmed their aversion to MSM practices, but felt that they could compartmentalize their values to achieve the greater national public health goal.
The key message is almost the same. We are concentrating in breaking the transmission cycle among special groups, neglected groups. The bottom line is: we are not promoting but trying to help.

Sophisticated knowledge of risk in MSM
During the follow-up FGDs, participants exhibited a multifaceted understanding of the biological, behavioural and social influences that place MSM at risk for HIV. They described a better understanding of the processes through which unprotected anal sex contributes to HIV and STI transmission in both men and women, and the ways in which condoms and lubricants help to reduce risk. Moreover, many participants identified quality health education and counselling for MSM patients as integral to HIV prevention efforts in Kenya.
Participants generally recognized the societal pressures on MSM to conceal their sexual orientation, which MSM often mitigated by engaging in heterosexual relationships. They discussed the ways in which discrimination and lack of counselling and support services have hampered access to vital health services for MSM. The stigma endured by MSM in Kenya was consistently identified as an impediment to treatment, and many participants emphasized the need for HCWs to be thorough when examining MSM patients, who might not readily disclose their sexual practices: I think it is good to do an examination as far as STI is concerned. A client might tell you that he is having a problem in his private parts. Such a client will openly tell you the exact location of the problem when you take the initiative to examine him. Even if they go, they tend to be reluctant to disclose to clinicians that they are having anal infections. They end up getting the wrong medication and suffer in silence.

Ongoing challenges
Participants reflected on the challenges they will continue to face in affording appropriate health services to MSM. Many HCWs noted that time constraints and heavy workloads hinder their ability to deliver sensitive health services that MSM patients might require. Despite their desire to provide comprehensive health services to their MSM patients, some of the participants felt this was not always possible in practice: Sometimes, as much as you would like to give all the attention to the client, there is a workload issue as other patients will be waiting. You may want to give the best, but the patients and the workload are too much.
Secondary stigma was considered an ongoing challenge, and HCWs tasked themselves to confront discrimination and stigma towards MSM expressed by their professional peers. Education, institutional support and other monitoring mechanisms were mentioned as powerful means for mitigating the effects of secondary stigma on service delivery to MSM patients, but all HCWs concurred with the fact that ''it begins with openness, respect and understanding.'' HCWs emphasized the social challenges in targeting MSM for HIV preventative care. The marginalization of MSM, the belief that homosexuality runs contrary to cultural values and the fear of secondary stigma and resistance from fellow health professionals were regarded as impediments to the provision of care for MSM. As one participated stated: Personally, I can say that my values have changed, though not 100%. I am not sure of the exact percentage, but I have positively changed. As much as I would like to live and exercise my changed values, there are still so many challenges in the society. I would like to give comprehensive care to MSM, but the society is too negative about them. This is a very big blow, given the fact that I am the only changed person.
In light of this, many participants noted the need for duplication opportunities for HCWs not yet trained on MSM sensitivity issues. They unanimously remarked that the on-line sensitivity course is very beneficial for skill development and in combination with follow-up group discussions allows for interpreting learning and connecting it to daily practice.
All participating HCWs advocated for community-wide sensitization campaigns to reduce stigma and encourage awareness of HIV risk in MSM, expressing the need for the community at large to engage in ongoing and productive dialogue in the struggle against HIV in Kenya.

Discussion
This analysis provides qualitative insight into HCWs' attitudes and experiences with MSM prior to and following a computer-facilitated MSM sensitization training programme [15] that will assist in amending the health workers' e-learning sensitization course in future. Primary concerns expressed at baseline included fear of secondary stigma, lack of professional education about MSM, and negative influences of personal and social prejudice towards MSM. The nature of discussions changed following the programme, in which participants acknowledged the presence of MSM in their clinics, endorsed the need to treat MSM patients with high professional standards, and demonstrated sophisticated awareness of the social and behavioural risks for HIV among MSM. HCWs advocated for continuing the training and inviting more health professionals to participate, but cautioned that exclusively targeting MSM in the programme title could deter participation. HCWs also commented on the need for ongoing community dialogue about MSM, but recognized that community-level change will take time.
The attitudes and beliefs expressed by participants before versus after the training reveal many of the challenges to van  service provision for MSM patients. In general, participants' personal beliefs about MSM and their endorsement of stigmatizing attitudes appear to have transformed following the programme. However, participants expressed ongoing concerns about secondary stigma and the influence of their professional peers' negative judgements towards MSM patient and, by association, towards themselves. Professional peers' negative and stigmatizing attitudes can potentially dilute the effects of the training on HCWs. Efforts to train larger cohorts of HCWs, establish networks of trained HCWs across different health clinics and change of institutional norms towards MSM patients may be necessary to counter the effects of secondary stigma and achieve sustainable improvements.
Limitations to this research must be acknowledged. First, due to the nature of qualitative methodology, participants' responses might be influenced by social desirability and peer influences. Second, the findings reported here do not permit temporal, causal, or quantitative inferences, but indeed correspond with programme evaluation data reported in a related paper [14]. Third, due to the voluntary nature of participation, attitudes expressed by HCWs in this sample might not be representative of their peers and colleagues. Fourth, due to the active role of Kenyan health administrators in supporting this programme, the findings might not be replicable in areas where such support is lacking.

Conclusions
This is the first known qualitative evaluation study of an MSM sensitivity training in Africa, which suggests that an online MSM sensitization training combined with group discussions can be a promising approach to improving health providers' awareness, attitudes and beliefs about the health needs of MSM patients. Quantitative evaluation results, which show similar findings, are reported in a companion paper [14]. Further research is needed to evaluate the programme in a controlled study, and examine the implementation processes associated with successful programme delivery. Perspectives and service delivery outcomes from MSM patients would enhance understanding of the impact of this training on patient interaction. A particular strength of the intervention was the incorporation of two complementary training modalities Á computer-facilitated training and group discussions Á to provide didactic content as well as opportunities for group reflection, feedback and support. In general, participants noted a transformation in their personal attitudes and endorsement of stigma towards MSM following the training. However, their comments revealed the continued challenges to providing services to MSM in the context of broader societal homophobia and secondary stigma among their peers; their comments also highlighted challenges in recruiting larger groups of HCWs into the training due to anxiety around secondary stigma. Findings reported here can inform further adaptations of the training, particularly those domains that might influence HCWs' willingness to participate and respond to the training (e.g., by emphasizing professional responsibilities of all health providers) and that diminish the effects of secondary stigma (e.g., by providing opportunities for ongoing support among trained HCWs). Findings underscore the need to view HCWs as an integral, but not absolute, component in addressing HIV and other health adversities among Kenyan MSM. Trained HCWs might benefit from continued opportunities for peer support, to counter feelings of professional isolation and motivate engagement in best practices. As participants noted, multi-component programmes and long-term commitments are necessary to achieve the goal of providing appropriate, effective services to MSM. Results: Crude HIV and syphilis prevalence estimates were 15.4% (RDS-weighted 12.5%, 95% confidence interval (CI): 7.3Á17.8) and 5.3% (RDS-weighted 4.4%, 95% CI: 3.1Á7.6), respectively. Ninety per cent (90.4%, unweighted) of HIV infections were reported as being previously undiagnosed. Participants were predominantly gay-identified (60.8%) or bisexually identified (36.3%); 50.7% reported recent concurrent relationships. Approximately half reported consistent condom use (always or almost always) with casual male partners, and proportions were relatively uniform across partner types and genders. The prevalence of perceived and experienced stigma exceeded 20% for almost all variables, 11.4% ever experienced physical violence and 7% were ever raped. Current age 25 years (RDS-weighted adjusted odds ratio (AOR) 3.9, 95% CI: 1.2Á12.7), single marital status (RDSweighted AOR: 0.3; 95% CI: 0.1Á0.8) and age of first sex with a man B16 years (RDS-weighted AOR: 4.3, 95% CI: 1.2Á15.0) were independently associated with HIV infection. Conclusions: Results demonstrate that MSM represent an underserved, at-risk population for HIV services in Malawi and merit comprehensive HIV prevention services. Results provide a number of priorities for research and prevention programmes for MSM, including providing access to and encouraging regular confidential HIV testing and counselling, and risk reduction counselling related to anal intercourse. Other targets include the provision of condoms and compatible lubricants, HIV prevention information, and HIV and sexually transmitted infection treatment and adherence support. Addressing multiple levels of HIV risk, including structural factors, may help to ensure that programmes have sufficient coverage to impact this HIV epidemic among MSM.
Keywords: HIV; men who have sex with men (MSM); behavioural risks; stigma; Malawi; prevention.

Introduction
Recent years have witnessed an increased awareness of the high burden of HIV among men who have sex with men (MSM) across the globe [1Á3]. Emerging research suggests a greater transmission efficiency of HIV through receptive anal intercourse that is approximately 18 times higher than that of penile-vaginal sexual contact, increasing the risk among MSM for acquisition of HIV during sexual intercourse [4,5]. National HIV strategies and funding priorities, however, remain inequitable in many countries [6,7], particularly where structural factors, such as the criminalization of homosexuality, play critical roles in the level of research and programming available to MSM [8,9].
The HIV response in Malawi has focused on the prevention of heterosexual and vertical transmission of HIV to counteract the observed HIV incidence rates of 2Á4% among adults in the 1990s. Today, the epidemic remains a generalized one, with an estimated 8.0% HIV prevalence among adult men [10]. Like neighbouring countries, assessments of specific risk factors for the acquisition and transmission of HIV, including transmission among MSM and other populations such as female sex workers, have been limited in the country [11]. Criminalization and stigmatization of homosexuality, as in other settings [8,12,13], are likely underlying factors for the limited targeted research and programming in the Malawian context.  [14]. This study documented HIV prevalence at approximately 21% [14], individual risk for HIV infections associated with increased age of the participant and inconsistent condom use [14] and high levels of violence and perceived stigma [15].
Another exploratory study examined socio-demographic and sexual behaviour characteristics among 97 MSM in central and southern Malawi. Although HIV prevalence was not assessed, the study found evidence of high-risk behaviours such as inconsistent condom use (32.5%), transactional sex (23.7%), low exposure to HIV messaging (17.5%) and a low history of HIV testing (58.8% ever tested) [16]. Although these studies were the first and only to elucidate the socio-behavioural factors among MSM in Malawi, they were rapid assessments and served to highlight areas for future research and prevention.
In response to the global epidemic of HIV among MSM, combination prevention packages have been put forth as a key method to curb the HIV epidemic among MSM [17,18]. To inform the content and scale of a combination HIV prevention intervention (CHPI) for MSM in Malawi, we conducted this study to estimate HIV prevalence, characterize associations of prevalent HIV infections, and evaluate barriers and facilitators to uptake of HIV prevention services among MSM in Blantyre, Malawi. Research was conducted in collaboration with a community-based organization, the Centre for the Development of People (CEDEP), and the Malawi College of Medicine, University of Malawi.

Study population and setting
This cross-sectional assessment was conducted from August 2011 to March 2012 in Blantyre, Malawi. Eligibility requirements for participation included being born male, being aged 18 years or older, being fluent in Chichewa or English, having reported anal sex with another man in the last 12 months, having no prior participation in this study, and providing informed verbal consent to participate. Study activities were conducted in private rooms of CEDEP's study site and implemented by staff from CEDEP, which provides HIV prevention activities for MSM in Malawi, and the Malawi College of Medicine. All staff members were trained in confidentiality and human subjects protection, qualitative and survey research and respondent-driven sampling (RDS) methods.

Recruitment method
Participants were recruited via RDS, a chain recruitment method often used to achieve more representative samples of hard-to-reach populations [19]. Recruitment began with 10 purposively selected seeds who were each provided with three study-specific coupons with which to recruit peer MSM from their social network into the study. Initiation of seeds was staggered over the duration of the study, taking into consideration potential propagation failure by some seeds and eventual die-out of the chains. Seeds were recruited from the pool of MSM who were involved in local HIV prevention programmes or had participated in prior formative research, and they were selected to represent a range of characteristics, including age, education, employment and sexual identity. Individuals who were recruited by seeds and enrolled in the study were then provided with three study coupons for further recruitment of peers. This process continued until the target sample size was reached. Participants were reimbursed K1500.00 (US$5.00) for transportation costs for participation in the study and K500.00 (US$1.50) for recruitment of each peer into the study. A full description of traditional RDS methodology can be found elsewhere [20]. Netdraw software (Analytic Technologies) was used to monitor RDS recruitment [21].

Sample size
The sample size calculation was powered on the assumed 85% effectiveness of condoms in preventing the transmission of HIV during intercourse [22]. Thus, we assumed that approximately 30% of the sample would be consistent condom users and that they would be 85% less likely to be living with HIV than the 70% who are not consistent condom users. Based on previous research, we estimated that the HIV prevalence in the population would be about 20%, equating to 27% among non-consistent condom users, 4% among consistent condom users and a 30% population prevalence of consistent condom usage. We used a design effect of 1.5 [23], power set at 80% and a significance level of 95%, which resulted in an effective sample size estimate of 345 participants for which we had targeted 350 MSM.

Measures
Participation included a structured survey instrument and a biological assessment of HIV and syphilis. Trained interviewers administered surveys in the Chichewa language, following pilot testing. Measures included sociodemographic characteristics, substance use, mental health and depression symptoms, sexual relationships and disclosure of orientation or sexual practices to family and peers. Measures of sexual practices included practices with men and women, including anal, oral and vaginal sex; number of sexual partners and partner characteristics; concurrent relationships, defined as ''two sexual partnerships at the same time or two ongoing sexual partnerships (male and/or female genders)''; and transactional sex (purchased or sold). We measured HIV knowledge and prevention, including aspects of condom and condom-compatible lubricant use; HIV testing and counselling exposures; and access to and uptake of health services. Human rights measures included experiences of physical and sexual violence, experienced and perceived stigma and history of imprisonment. Recall periods were lifetime, last 12 months or both, and they are specified in the results tables. The development of survey questions, recruitment methods, coupons and study procedures was informed by formative research that was conducted in MayÁJuly, 2011 [24]. Wirtz  Biologic specimens Following completion of the interview, participants proceeded to HIV and syphilis testing. A trained nurse from the College of Medicine conducted HIV testing, blood specimen collection and pre-and post-test counselling. Blood-based rapid HIV tests were conducted simultaneously using the Determine † HIV-1/2 and Uni-Gold rapid HIV tests (manufactured, respectively, by Inverness Medical, Chiba, Japan; and Trinity Biotech, Bray, Ireland). Participants received their HIV test results and post-test counselling within 15 minutes of collection. Separate specimens were collected for confirmatory testing of discrepant or indeterminate HIV rapid tests using Western blot in accordance with Malawian National Guidelines [25]. Approximately 5 ml of whole blood was collected for TPHA (treponema pallidum haemagglutination test) syphilis testing (Bio-rad, Hercules, CA, USA). Resource constraints prevented the use of the nontreponemal test, which would differentiate active from past syphilis infections. Confirmatory HIV and syphilis tests were analysed at the Malawi College of Medicine laboratory in Blantyre. Participants returned within one to two weeks to receive their syphilis test results. Participants testing positive for HIV and/or syphilis were referred to the local hospital or to the Johns Hopkins antiretroviral therapy and sexually transmitted infection clinic located at Queens Hospital. Participants were provided with information about local health centres that had, as part of the study, received training for the provision of services to MSM. One trained team member (EU) provided counselling services to MSM participants as needed.

Analysis
Johns Hopkins University conducted secondary data analysis of collected data. The principal outcome of interest was HIV diagnosis with predictor variables that included demographics (education, age, number of children and marital status), socioeconomic variables, lifetime residence in urban or rural locations, recent sexual behaviours, human rights exposures, HIV prevention methods, health-seeking behaviour and laboratory markers of syphilis infection. Variable-specific individualized weights, which take into account estimates for individual degrees, were computed by a data-smoothing algorithm using RDS for Stata [26]. The estimated weights were used in univariate RDS-weighted analyses. HIV status individualized weights were used in the bivariate and multivariate RDS-weighted analyses. Bootstrapped 95% confidence intervals (CIs) were computed using 1000 iterations for the estimated descriptive statistics [23]. Homophily, a measure of to what extent respondents prefer to recruit from their own group rather than at random, was estimated where appropriate and presented in the results in Table 1.
To develop the statistical model, we first carried out bivariate analysis to assess the association of HIV status with the control variables (Table 4). Demographic variables were included in the multivariate logistic regression model regardless of the estimated strength of their bivariate association with HIV status. Selected non-demographic variables were included in the multivariate model if the chi-square p-value of association with HIV status was 50.25. Some variables such as HIV testing were not included in the multivariate model due to collinearity. The final model, presented in Table 4, includes demographics and variables left in the final model following goodness-of-fit tests. All statistical analyses were conducted using Stata 12.1 [27]. Results provided in the text report RDSweighted estimates (unless otherwise specified), while tables display unweighted and RDS-weighted estimates as well as 95% CIs for weighted estimates. Table 4 presents the results of bivariate and final multivariate analyses, including unweighted and RDS-weighted odds ratios (ORs) and adjusted ORs (AORs) for the final multivariate model.

Human subjects
Research activities were reviewed and approved by the Malawi College of Medicine Ethics and Research Committee and for secondary analysis by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

Results
A total of 338 MSM (including original seeds) were recruited via RDS and enrolled into the study, reaching 19 waves of recruitment. Out of 10 seeds, five recruited participants; one recruitment chain was responsible for the recruitment of 70% of the study population. Three recruitment chains are reflective of later seed initiation. A total of 706 coupons were distributed with a return rate of 48%. The majority of participants reported recruitment by a friend (60.5%) or sex partner (32.3%). Median MSM network size was 8 (range 1 to 800). Figure 1 displays the RDS recruitment diagram, highlighted by HIV diagnosis. We used this method to monitor recruitment and to assess whether HIV diagnosis inhibited recruitment, which appeared not to be the case.
Participants' median age was 25.1 years (range: 18 to 49). Based on RDS-weighted estimates, 51% were unemployed and 21.6% had ever been in jail or prison. Eighty per cent identified as male gender. Sixty-one per cent identified as gay or homosexual, and 36.3% reported bisexual identity. Sixteen per cent were married or cohabitating with a woman. Table 1 displays sociodemographic characteristics.
The crude prevalence of HIV infection in this population was 15.4%, with an RDS-weighted estimate of 12.5% (Table 1). The majority, 90.4% (unweighted), of these infections were previously undiagnosed; these participants had either selfreported as negative or reported never being tested for HIV infection. Positive syphilis diagnosis was low at 4.4%. Table 2 presents sexual practices, partner characteristics and social exposures. Only 18.1% had ever disclosed sexual practices or orientation to their family, and equally few (18.9%), had ever disclosed to a health provider. Participants reported a mean of four male partners (range: 1 to 50), and 31% reported having female partners in the last 12 months. Half of the population reported concurrent sexual relationships, and, among those in a relationship, 61.3% believed their partner was also involved in a concurrent relationship. Prevalence of perceived and experienced stigma and discrimination exceeded 20% of the population for almost all variables, 7.0% were ever raped and 11.4% had ever experienced physical violence.
Responses to questions on knowledge of HIV risk, prevention methods and practices are reported in Table 3.  16.4.18742 Approximately half of the participants with casual male partners (n 0256) reported using condoms always or almost always with casual male partners; frequencies were approximately similar across partner type (e.g., casual or main) and partner gender. Approximately 44.3% had never been tested for HIV. Among those ever tested for HIV infection, 45.5% (unweighted) had not been tested within the last year.  Several sociodemographic variables were associated with HIV infection in the bivariate analysis (Table 4). These included current age 25 years (RDS-weighted OR: 8.1, 95% CI: 2.9Á 22.2), single marital status (RDS-weighted OR: 0.2, 95% CI: 0.1Á0.4) and having more than one child (RDS-weighted OR: 5.3, 95% CI: 1.8Á15.6). Age B16 years at first sex with a man was associated with HIV infection (RDS-weighted OR: 1.7, 95% CI: 0.4Á7.5). Considering water-based lubricants to be the safest lubricant (RDS-weighted OR: 0.9, 95% CI: 0.2Á3.6) and use of water-based lubricant (RDS-weighted OR: 0.6, 95% CI: 0.2Á2.0) were also marginally protective. The final multivariate model included age, marital status, number of children, knowledge of risk related to positioning (insertive or receptive anal intercourse), lubricant type used, age of first sex with another man, history of rape, number of male anal or oral sex partners and other known confounders such as employment, education and syphilis diagnosis. Of these, current age 25 years (RDS-weighted AOR 3.9, 95% CI: 1.2Á12.7), single marital status (RDS adjusted AOR: 0.3, 95% CI: 0.1Á0. 8) and age of first sex with a man B16 years (RDS adjusted AOR: 4.3, 95% CI: 1.2Á15.0) were independently associated with HIV infection.

Discussion
This cross-sectional study, the most comprehensive yet conducted among MSM in Malawi, describes the high prevalence of HIV infection as well as the limited uptake of HIV prevention, testing and care services among MSM in Blantyre, Malawi.
HIV prevalence was high among MSM, and nearly all HIV infections were among men who reported being unaware of their status of HIV infection. Only slightly more than half of the population reported ever having been tested, and only half of those were within the last year, potentially explaining this level of undiagnosed HIV infections. Knowing one's status is increasingly more important for HIV prevention. Novel HIV interventions, including pre-exposure prophylaxis for HIVuninfected men [28,29] and early treatment for people living with HIV [30], represent a new generation of HIV-statusdependent interventions. Awareness of one's HIV status has also been associated with decreased self-reported prevalence of high-risk sexual practices that are associated with HIV transmission [31]. Recent US Centers for Disease Control guidelines have suggested more frequent testing (every 3 or 6 months) based on individual assessment of sexual risk behaviours [32], representing a strategy which may also be relevant for MSM in Malawi. Young age of first sexual intercourse with a man ( B16 years) was independently associated with HIV infection in this population, with almost four times greater odds of HIV infection compared to the referent group. This association may suggest biologic susceptibility during physical development, high-risk sexual behaviours and lack of access to or low use of condoms at a young age, and/or an association with duration of sexual activity [33,34]. Likewise, the association of prevalent HIV infection with older current age may be due to higher cumulative risk exposures for acquisition of HIV.
However, estimating the duration of sexual activity is challenging as sexual behaviours are not static, but vary across the life course and as partnerships change [34]. While study-related factors such as low power and potential misclassification of behaviours may partially explain insignificant findings, broader factors such as high background prevalence of HIV in the MSM population [35], biologic susceptibility of rectal mucosa [36] and network-level characteristics may also be more determinative in driving HIV transmission and acquisition risks among these men [35,37]. Nonetheless, this study described a population reporting high-risk behaviours, suggesting the need to ensure accessibility to HIV prevention interventions across ages [38Á40]. These behavioural risks, combined with the high proportion of undiagnosed HIV infection in this study, also suggests there is a high likelihood of someone with a high viral load within a sexual network potentially driving onward transmission [41,42]. Future research among MSM in Malawi to better characterize different risk strata among MSM, including reported sexual practices and sexual network characteristics, is needed to better tailor the content of interventions and enable the identification of infection. While addressing the unique needs of the individual is fundamental, stigma and discrimination have been reported commonly as structural barriers to the uptake of services [43,44]. Experienced and perceived stigma as well as other physical and sexual violence were common among MSM in this study, consistent with earlier quantitative and qualitative studies in Malawi [15,24]. Stigma has been shown to limit health-seeking behaviours and use of HIV prevention methods, disclosure of sexual practices to health providers, and providers' liberty to provide services to MSM [14,15,24,45,46]. The need to keep male-male partnerships hidden may lead to more frequent, short-term relationships and increased highrisk behaviours [24]. Such responses to stigma and social pressures may explain the high prevalence of concurrency, the high-risk sexual practices reported in this study, the proportion of men who are married or cohabitating with women and the protective effect of single marital status in this analysis. Addressing these social issues is a necessity for improving access to and uptake of effective HIV prevention interventions [8].
Taken together, these data demonstrate that MSM are an underserved and important population for targeted HIV prevention interventions; MSM may specifically benefit from the CHPI that we subsequently developed based on the quantitative results presented here. Mathematical models have shown that high levels of coverage among MSM (i.e., 60Á 80%) are required to change the trajectory of the HIV epidemic among MSM, and such findings are likely to be relevant in Malawi [2,47,48]. To address low coverage of prevention options among Blantyre MSM and the limitations of single interventions, comprehensive packages of interventions that include behavioural, biomedical and structural approaches may be the most effective approach to reducing HIV among MSM [17]. Such interventions may be feasible in Malawi and may have the same positive impact on sexual transmission that has been observed in other settings, including countries where same-sex practices are criminalized [30,47,49,50].
The method of intervention delivery is critical to the success of HIV prevention programmes in the context of complex social environments. The success of RDS recruitment suggests that interventions leveraging existing peer networks, which have demonstrated efficacy in other settings [51,52], may serve as a feasible approach to providing and supporting HIV prevention interventions for MSM in Malawi. Addressing stigma in healthcare settings may improve provider-patient relationships, facilitate disclosure and meaningful discussion of risk practices, and foster linkage to HIV testing and care [53]. While the subsequent feasibility assessment of the CHPI programme for MSM in Blantyre will be informative for understanding how a comprehensive package may address individual social and behavioural risks for HIV infection, broader social acceptance of MSM may take time and remains a crucial step towards improving the health status of MSM and thus all Malawians [8].

Limitations
The cross-sectional nature of this study limits the investigation of temporal associations and thus the causality of the exposures and HIV-related outcomes. Additional limitations are related to the ability to fully assess correlates of prevalent HIV infection through behavioural surveys, which may have limited the significance of findings in this study. This may also be amplified by the potential response bias related to asking sensitive questions of a highly stigmatized population. We attempted to address these limitations to the fullest extent possible, including using lifetime and recent recall periods, developing survey questions based on formative research and prior research studies among MSM, and taking measures to ensure the confidentiality and privacy of participants and inform them of these privacy control measures. This study provides equipoise for prospective cohorts of MSM to better characterize HIV incidence and, ultimately, appropriately powered HIV prevention and implementation science studies to assess effective strategies in HIV risk reduction.
There are limitations associated with the use of RDS methodology [54]. Specifically, there is debate around appropriate interpretation of the measures of association and optimal strategies to handle variance in studies using RDS. For example, use of water-based lubricants appeared to be independently protective in the model that did not adjust for RDS, but this association is no longer significant with the introduction of the increased variance associated with RDS adjustment in the model. Despite these analytic challenges, RDS represents a relevant sampling strategy to obtain a diverse sample of a hidden population in the absence of a sampling frame or a sufficient number of established venues [19].

Conclusions
This study presents an assessment of individual, sexualnetwork and structural factors and their relationship with prevalent HIV infections among MSM in Blantyre, Malawi. The burden of HIV is high among these men, with the vast majority apparently unaware of their HIV status. Approaches rooted in engagement in the continuum of HIV care will be central moving forward in Malawi [55]. Addressing stigma and discrimination should also represent a core programmatic and policy element of the HIV response, to ensure that these efficacious approaches are translated into effective ones and to optimize the health of MSM living with HIV in Malawi while preventing onward HIV transmission.