Article Text
PDF
Abstract
Objectives Men who have sex with men (MSM) are at high risk of HIV-1 acquisition and transmission, yet there remains limited data in the African context, and for men who sell sex to men (MSM SW) in particular.
Methods We enrolled 507 male sex workers in a Nairobi-based prospective cohort study during 2009–2012. All participants were offered HIV/STI screening, counselling and completed a baseline questionnaire.
Results Baseline HIV prevalence was 40.0% (95% CI 35.8% to 44.3%). Prevalent HIV infection was associated with age, less postsecondary education, marijuana use, fewer female partners and lower rates of prior HIV testing. Most participants (73%) reported at least two of insertive anal, receptive anal and insertive vaginal sex in the past 3 months. Vaginal sex was reported by 37% of participants, and exclusive MSM status was associated with higher HIV rates. Condom use was infrequent, with approximately one-third reporting 100% condom use during anal sex. HIV incidence was 10.9 per 100 person-years (95% CI 7.4 to 15.6). Predictors of HIV risk included history of urethral discharge (aHR 0.29, 95% CI 0.08 to 0.98, p=0.046), condom use during receptive anal sex (aHR 0.05, 95% CI 0.01 to 0.41, p=0.006) and frequency of sex with male partners (aHR 1.33/sex act, 95% CI 1.01 to 1.75, p=0.04).
Conclusions HIV prevalence and incidence were extremely high in Nairobi MSM SW; a combination of interventions including increasing condom use, pre-exposure prophylaxis and access to effective treatment is urgently needed to decrease HIV transmission in this key population.
- Africa
- HIV
- Males
Statistics from Altmetric.com
Introduction
Key populations at risk of HIV acquisition typically include female sex workers (FSW), men who have sex with men (MSM), HIV-uninfected members of HIV-discordant couples and people who inject drugs (PWID). The Kenya Modes of Transmission study estimated that an approximately 16% of new HIV cases in Kenya involved MSM transmission (https://openknowledge.worldbank.org/handle/10986/3044). While much HIV research in sub-Saharan Africa has focused on heterosexual transmission in the context of a generalised epidemic, and despite criminalisation and intense stigma that surrounds homosexuality in many African countries, awareness of a sizeable HIV epidemic in African MSM is now emerging,1 including reports in African medical journals.2 ,3 In many African countries where data are available, MSM HIV prevalence is similar to or higher than the PWID population.4 Due to the lack of research and low levels of advocacy for African MSM populations, the need for MSM-oriented interventions and calls for improved country-level HIV surveillance have recently been highlighted as important research priorities.5–,7
Despite the paucity of published studies, aggregate MSM HIV prevalence in sub-Saharan Africa was recently estimated at 17.9%, second only to MSM HIV rates in the Caribbean.8 MSM cohort studies in South Africa, Malawi, Kenya and Botswana have reported HIV prevalence rates ranging from 10.4% to 43%.9–,13 Kenyan MSM cohorts, in particular, exhibit some of the highest HIV prevalence rates in Africa,8 many folds higher than the general population.4 Although HIV incidence estimates for African MSM populations are few, a recent cohort study of MSM in coastal Kenya reported an HIV incidence rate of 8.6 per 100 person-years; among MSM who reported sex exclusively with other men, this was 35.2 per 100 person-years,14 one of the highest HIV incidence rates reported worldwide (reviewed in Ref 8).
Disconcertingly, many HIV-positive MSM were unaware of their serostatus,12 reflecting lower levels of HIV awareness that have been reported in other marginalised populations in Africa.15 Risky sexual practices, including unprotected intercourse and use of non-water-based lubricants, have been commonly reported.12 ,13 Additional data on African MSM and HIV vulnerability would help to inform targeted prevention programmes, but lack of quality studies in the region leave a dearth of evidence on which to make these decisions.14 ,16
Within MSM cohorts, some participants appear at greater risk of HIV acquisition than others. Several studies have described men reporting transactional sex work, potentially placing them at even greater risk of HIV infection.11 ,17 ,18 However, populations of MSM who sell sex (MSM sex workers, MSM SW) form a subgroup of MSM that has rarely been described in literature. To date, the few published studies of MSM SW populations in Africa have described very high HIV prevalence of >40%.19 ,20 More than one-third of MSM in this cohort were unaware that HIV could be transmitted by anal sex.19
Given the stigma associated with MSM behaviour and the reported low rates of disclosure to health professionals among African MSM,21 it is crucial for additional research in this group. A better understanding of HIV risk among African MSM could have important public health implications leading to interventions to prevent HIV in this population. Here we present data on HIV prevalence, incidence, risk factors and sexual behaviour in a recent Nairobi-based MSM cohort.
Methods
Study cohort
Participants were recruited to a clinic in central Nairobi during 2009–2012 as part of an open cohort. The majority of participants was active in sex work (ie, selling sex to men) at baseline, and hence will be described as MSM SW. This clinic offers a range of services, including health education, provision of condoms and lubricants, condom demonstration and HIV/STI testing and treatment. Institutional review boards at Kenyatta National Hospital and the University of Manitoba approved this study.
Recruitment for the clinic followed a combination of approaches, including hot spot-based and snowball recruitment (for more details, see online supplemental methods). Participants who enrolled at the clinic were scheduled for quarterly visits, and as need arose. HIV/STI testing was encouraged (but not required to receive services), and clinic personnel administered questionnaires at baseline and each visit. Key demographic and behavioural variables are summarised in table 1.
Baseline associations of HIV status in Nairobi MSM SW
Statistical analyses
HIV prevalence and incidence rates in this cohort, and their 95% CIs, were calculated using the Mid-P exact test in OpenEpi V2.3.1 (http://www.openepi.com/OE2.3/Menu/OpenEpiMenu.htm). We compared baseline cross-sectional enrolment data using Mann–Whitney tests (continuous variables) and χ2 tests (categorical variables). Separate logistic regression models were used to identify associations of HIV status and being an exclusive MSM at baseline (defined as those reporting 0% of their sex partners in the past 3 months as women), with stepwise backward selection of significant associations using the likelihood ratio method. Variables entered into the model included all those at p<0.1 in univariate analysis. Adjusted ORs (aOR) and their 95% CIs from multivariate logistic regression models are reported. A number of sensitivity analyses around model building were run and can be found in online supplemental tables S1 and S2.
Survival analyses examining baseline factors associated with HIV incidence were carried out using Cox proportional hazards regression analyses, including all HIV-uninfected MSM SW. Participants were censored at the time of HIV infection (primary endpoint) or last HIV negative visit. Behavioural and demographic variables from table 1 were evaluated separately in univariate analyses; all variables associated with HIV acquisition at p<0.1 were included in the stepwise multivariate analysis. Raw and adjusted HRs and 95% CI are reported, with the final model removing variables not associated with HIV using backward stepwise selection and likelihood ratios. All statistical analyses were carried out using SPSS V.20 for Mac (see online supplemental material for more details).
Results
Baseline cohort characteristics
At baseline, the majority of the cohort (87%) reported being active in sex work. Of the initial sample, >99% had a valid HIV test result (n=507/510) at the time of analysis. HIV prevalence in this cohort was 40% (203/507) at baseline (95% CI 35.8% to 44.3%). Antiretroviral therapy and HIV care was offered to all HIV-infected participants as per Kenyan national guidelines.
Associations of prevalent HIV infection at baseline
We identified several univariate associations of HIV infection at baseline (table 1). HIV-infected men were older (median 28 vs 26 year, p<0.001) and involved in sex work for a longer duration (median 4 vs 3.8 years, p=0.021); these variables were correlated (r=0.38, p<0.001). Almost all men (99%) reported Kenya as their birthplace. The majority (59%) had a client in the past week. HIV-infected men were less likely to be married to women (64% vs 73%, p=0.05). The cohort was relatively well educated, with more than 80% having secondary education or better, and nearly a third reporting postsecondary education. Higher education levels were associated with lower HIV prevalence at baseline (p<0.001), with postsecondary education almost twice as common in HIV-uninfected men (39 vs 22%). Alcohol use was reported by the majority (75%) of men. Bhang (marijuana) use was reported infrequently (11%) and associated with a slightly lower HIV prevalence at baseline (p=0.051).
About two-third (64%) of participants reported one or more male partners in the past week, and 7% reported three or more male partners during that time. Approximately one-third of men reported using condoms ‘always’ during either insertive or receptive anal sex; condom use was slightly increased in HIV-uninfected participants, although this was not statistically significant (p=0.105). Condom use during different types of sex and partner types was highly correlated (not shown). More than half (59%) of men reported using a condom during their last sex act, with no differences based on HIV status. No associations of sex type, partner volume or condom frequency and HIV seroprevalence were observed. Both anal receptive and insertive sex were common in this cohort (discussed in more detail below). Sex with women was also relatively common, with 38% reporting insertive vaginal sex in the past 3 months and 16% reporting a female partner in the past week. Vaginal sex in the past 3 months was associated with a trend towards lower HIV prevalence (33 vs 41%, p=0.071).
A history of previous STI was common, including genital ulcer disease (27%), urethral discharge (43%) and painful anal sores (28%). These were all associated with a higher HIV prevalence at baseline (p=0.011, 0.044 and 0.006, respectively). Data for specific STI pathogens were not routinely collected, aside from syphilis, which was not common (two active infections based on serology), and urethral gonorrhoea culture, which was available on only a small subset of participants. Seeking treatment at a private hospital for STI was more common in HIV-infected men (31 vs 22%, p=0.016). HIV-positive men were less likely to have been HIV tested prior to cohort enrolment (80 vs 90%, p=0.004).
We determined the independence of associations of HIV prevalence using stepwise logistic regression, including variables that were associated with HIV (p<0.1) in univariate analyses (table 2). Associations of HIV prevalence that remained significant included age (p=0.018, aOR 1.05/year, 95% CI 1.01 to 1.09), bhang use (p=0.027, aOR 0.40, 95% CI 0.18 to 0.90), history of prior HIV testing (p=0.042, aOR 0.51, 95% CI 0.27 to 0.98), having female partners in the week prior to enrolment (p=0.007, aOR 0.38, 95% CI 0.19 to 0.77), ability to demonstrate condom use (p=0.043, aOR 1.69, 95% CI 1.02 to 2.79) and postsecondary education (p=0.001, aOR 0.31, 95% CI 0.16 to 0.63). Secondary education was associated with a lower HIV prevalence compared with primary education or less, but this was not significant (p=0.172). These data highlight several differences between participants on the basis of baseline HIV status.
Adjusted ORs (aOR) and 95% CIs from stepwise logistic regression models of variables associated with HIV status at baseline (table 1) in a Nairobi MSM SW cohort
Sexual practices at enrolment
Most men reported engaging in at least two of receptive anal sex, insertive anal sex and insertive vaginal sex, including 19% who reported all types (figure 1). More than one-third of men (37%) reported both anal receptive and anal insertive sex but not vaginal sex. In total, 12% and 11% of participants reported only having receptive or insertive anal sex, respectively (and not vaginal sex). Approximately three-quarter (73%) of participants practiced two or more types of sex, and 37% practiced two or more types that included vaginal. A comparison of baseline attributes reveals many differences in men who exclusively have sex with men compared with men who also have sex with women (see online supplemental table S3). Approximately two-third (68%) of participants reported ‘ever’ having oral sex with clients, and 9% reported ‘always’. Vaginal sex was associated with higher rates of insertive anal sex (OR 2.27, 95% CI 1.17 to 4.38) and lower rates of receptive anal sex (OR 0.47, 95% CI 0.28 to 0.79). In summary, these data suggest that more than one type of sex is commonly practiced in this cohort and that a significant proportion of men also have regular sex with female partners.
Self-reported sexual behaviour in Nairobi men who have sex with men (MSM) in the 3 months prior to enrolment in an HIV/STI care and prevention programme. On the x-axis is the number of participants reporting one or more of a combination of three sexual behaviours in the 3 months preceding cohort entry: insertive anal sex (‘insert’), receptive anal sex (‘recept’) and insertive vaginal sex (‘vag’).
Prospective associations of HIV risk
We next determined HIV incidence and risk factors for acquisition. A total of 181 initially HIV-uninfected men (60% of the initial HIV seronegative cohort) were followed for 256 person-years, and 28 new HIV infections were observed during this period (ie, an HIV incidence of 10.9 per 100 person-years, 95% CI 7.4 to 15.6). In Cox proportional hazard models, univariate associations of HIV acquisition included younger age at baseline, younger age at sexual debut, condom use during receptive or insertive anal sex, and frequency of sex with male partners (table 3). Other variables including history of urethral discharge, painful anal sores, marijuana use and receptive anal sex in the past 3 months showed weaker associations (p<0.1) with HIV risk and were included in multivariate models.
Prospective associations of HIV acquisition for key variables in Nairobi men who have sex with men (MSM)
In multivariate analysis, predictors of HIV infection included history of urethral discharge (aHR 0.29, 95% CI 0.08 to 0.98, p=0.046), condom use during receptive anal sex (aHR 0.05, 95% CI 0.01 to 0.41, p=0.006), frequency of sex with male partners (aHR 1.33/sex act, 95% CI 1.01 to 1.75, p=0.04) and age at sexual debut (aHR 0.89/year, 95% CI 0.78 to 1.01, p=0.077). Reporting receptive anal sex in the 3 months prior to enrolment was also associated with increased HIV risk; 100% of those who acquired HIV reported this type of sex compared with 69% of those who remained uninfected (p<0.00001). For this reason, this variable was not entered into multivariate models. Restricting the cohort to those who reported receptive anal sex provides a higher incidence estimate of 14.5 per 100 person-years (95% CI 9.7 to 20.9).
Condom use during insertive and receptive anal sex was both strongly protective, even though receptive condom use was a stronger predictor in multivariate models (see online supplemental figure S1). Condom use during these types of sex was highly correlated (p<0.00001). The protective effects of condoms were particularly strong (95% for during receptive anal sex) when analysed as ‘always’ versus ‘not always’, but also when considered as categories (always, often, sometimes, never).
Discussion
MSM represent a neglected but emerging key population within the African HIV epidemic. Here we present extremely high HIV prevalence (40%) and HIV incidence rate (11 per 100 person-years) in this population; these rates are 7-fold and 16-fold higher than what has been observed in the general Kenyan population (5.6% and 0.7 per 100 person-years, respectively22). HIV incidence in MSM SW was five times higher than FSW attending the same clinic during a similar period (McKinnon et al, submitted). These rates are consistent with other African MSM studies and slightly higher but generally comparable to those previously reported in Kenyan MSM (8.6–9.1 per 100 person-years).14 ,16 A better understanding of these high rates of HIV transmission is an important research priority.
Consistent condom use during receptive and insertive anal sex was low in the cohort (approximately one-third), and frequent condom use was strongly protective. Disconcertingly, only one-quarter of participants were able to correctly demonstrate how to use a condom and less than 60% reported using a condom during last sex. When compared with FSW who attended the same clinic during this period, MSM SW report lower numbers of casual and regular partners but used condoms much less frequently. Despite fewer partners, frequency of sex with men was a predictor of HIV risk, suggesting this might be important for risk reduction counselling. These findings underscore the need for scaled up basic prevention programmes among this group, including research around determinants and barriers of condom use, while other effective biomedical prevention options come online in this region.
Approximately three-quarter of MSM SW participants reported combinations of receptive anal, insertive anal and insertive vaginal sex in the past 3 months. This included about one-fifth who practiced all three, and a large proportion that practiced both receptive and insertive anal sex. These could have implications for HIV risk; at baseline, exclusive MSM (approximately 60% of the cohort) were more likely to be HIV-positive despite being younger and having fewer sex partners within the last week. Prospectively, receptive anal sex (particularly if unprotected) was associated with increased rates of HIV acquisition in univariate analysis. Urethral discharge was associated with lower rates of acquisition; one explanation for this observation is that urethral STIs are a marker for more frequent insertive anal (and less frequent receptive anal); indeed, we found urethral discharge and receptive anal sex to be inversely associated (data not shown). The diversity of sexual behaviour including frequent sex with women and low condom use and anecdotal reports of MSM SW and FSW sharing clients suggest this represents a true bridging population for HIV transmission. This bridging could include MSM who have sex with women, but also MSM who have sex with men who in turn have female partners. Finally, while this report focused on the epidemiology of HIV, social determinants of HIV risk among these men is an important additional area in need of research.
Successful interventions in this population will require better knowledge of those at risk. Consistent with other MSM SW cohorts,12 ,13 ,16 prevalent HIV in this cohort was associated with increased age, history of STIs and lower education. The cohort was generally young and highly educated, consistent with data from other regions.23 ,24 Approximately 80% of MSM SW received education at secondary school or higher compared with 40% of FSW at the same clinic. In addition to education, it will be critical to also better understand the dynamics of sex work in MSM, including the proportion of sex that is paid and differences between MSM who pay and receive payment for sex. Follow-up studies including sexual network analysis will be important to confirm and expand these observations to better understand HIV transmission in this population.
Our study has some limitations, including that the sample size and duration of follow-up were relatively limited for our prospective analyses. Limited follow-up among MSM cohorts has been reported previously, especially among exclusive MSM groups, highlighting the importance of tailoring outreach programmes that encourage sustained participant contact. Other potential limitations include reliance on self-reported data, which can be subject to numerous biases; restricting the analyses to baseline data might limit one of these moderately (social desirability bias). Most participants were recruited through sex work hot spot mapping and self-identified as sex workers, suggesting that these results cannot be generalised to the greater MSM community within Nairobi, many of whom are not likely to be involved in sex work. Additional efforts to engage this community will be necessary to gain a better understanding of the representativeness of this MSM SW sample, including a better understanding of part-time and transient, informal sex work. Properly diagnosed STI data were not widely collected in this study, revealing a major programme gap in this population. Lastly, the use of stepwise regression could lead to exaggerated p values and estimates in our multivariate analyses. We have addressed this in part by inclusion of several sensitivity analyses.
Overall, the diverse sexual behaviour and partner types reported by this group of MSM SW imply that this represents a true bridging population to the general population. The proportion of men in this cohort reporting sex with both men and women was higher than has previously been reported in other cohorts.12 Given the extremely high HIV incidence and prevalence, the relative high frequency of risk-taking behaviour and the possibility that HIV-positive men would be unlikely to return for HIV testing make it clear that targeted HIV/STI preventions in this population—including antiretroviral provision, pre-exposure prophylaxis and others—are urgently needed.
Key messages
-
HIV incidence and prevalence were extremely high in Nairobi-based men who have sex with men (MSM) sex workers (MSM SW), suggesting an urgent need for HIV prevention in this key population.
-
Consistent condom use was relatively infrequent, but strongly protective against HIV acquisition in prospective follow-up (in particular for receptive anal sex).
-
Many participants practice anal receptive and insertive sex, and sex with women, suggesting MSM SW represent a true bridging population in the Kenyan HIV epidemic.
Acknowledgments
We would like to thank the staff at the Kenya AIDS Control Project for their support in the operation of this clinical programme and the participants to dedicating their time to this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
LRM and GG are the co-first authors.
-
Handling editor Jackie A Cassell
-
Contributors LRM, LG, RK and JK designed the study. GG, JM, NI, and CWK collected the data and managed clinic activities. LRM, JAJ, JM, SYS and NJDN analysed the data. LRM, GG, JAJ, PI, SYS, NJDN, RK, RL and JK wrote the manuscript. All authors approved the final submitted version of the manuscript.
-
Funding US PEPFAR, Canadian Institute of Health Research. Funding was provided by US-PEPFAR (CDC-PEPFAR U62/CCU024510-01) to LG and JK and CIHR MOP-89983 to RK. RK is supported by an OHTN Career Scientist Award. LRM is supported by a CIHR fellowship and the International Infectious Diseases and Global Health Training Program.
-
Competing interests None.
-
Ethics approval Kenyatta National Hospital ERC and University of Manitoba.
-
Provenance and peer review Not commissioned; externally peer reviewed.