1 Sero-epidemiological surveys of HIV exposure among men who have sex with men (MSM) in sub-Saharan Africa only emerged in the mid-2000s, having been non-existent for over two decades of the AIDS epidemic. The first study to measure HIV prevalence among MSM on the continent was conducted in Senegal in 2004 and published in 2005 (1). With an HIV prevalence of 22% among MSM, compared to the low prevalence observed in the general population (less than 1%), this study highlighted the high exposure of MSM to HIV, which was confirmed by the numerous surveys developed since in sub-Saharan Africa, in parallel with the growing international attention given to MSM in resource-limited countries (2): the Global Forum pre-conference on MSM and HIV alongside the larger 2006 World AIDS Conference; the creation of the Africagay network in French-speaking Africa in 2007; and thematic sessions dedicated explicitly to MSM for the first time at the 2008 African AIDS Conference (ICASA), etc.
2 The MSM acronym (men who have sex with men) emerged at the beginning of the 1990s (3) in works on HIV. It conveys the idea that it is behaviours, not identities, which put individuals at risk of HIV exposure, an essential distinction since some scientists and medical experts had initially presented gay identity as a risk factor, which was a source of stigma (4). The MSM category has gradually become more prevalent in epidemiological surveys, enabling greater inclusiveness by including men who do not necessarily consider themselves “gay” but have sex with men. It includes both men who have sex with men exclusively (MSME) and men who have sex with both men and women (MSMW).
3 The 2004 Senegalese survey also provided important information about bisexual behaviour: 94% of respondents reported having had sex with a woman at least once in their lifetime (1). In the article, the authors highlighted this information as being a risk factor for the epidemic to spread to the general population. In 2005, the Senegalese authorities used this argument to advocate for the implementation of a prevention program for MSM (5): targeting MSM would protect the general population, as bisexuals could constitute a “bridge” between groups.
4 This large proportion of men with bisexual practices has become a recurring feature of surveys on MSM in Africa and one of the most debated aspects, in a context where the idea that MSMW are more susceptible to HIV acquisition than MSME was prevalent.
5 This paper presents an overview of epidemiological knowledge about HIV acquisition among different categories of homo-bisexuals in epidemiological surveys conducted in sub-Saharan Africa since 2005, including a comparison of available data on HIV prevalence and incidence along different dimensions of measuring sexual orientation.
Materials and methods
6 Since 2005, we have maintained an ongoing scientific monitoring on publications concerning sexual and gender minorities in sub-Saharan Africa, based on searches of major bibliographic portals, references identified through reading or at international conferences, and reports or even unpublished analyses of which we have become aware through our research networks. This is not a systematic review, as the identification and collection of articles have been carried out continuously over a period of 15 years. The methodology used is similar to that of “narrative reviews” or “critical reviews” (6)
7 From this review of the scientific literature, which includes publications issued up to the end of 2020, we selected those presenting a quantified comparison of HIV prevalence or incidence among MSM, based on biological measurements (seroprevalence surveys, incidence cohorts, etc.). A review of the literature based on the same corpus and focusing on the construction of sexual orientation and gender categories in quantitative surveys conducted in sub-Saharan Africa has also been published (7) (see also https://doi.org/10.5281/zenodo.4569342 for a summary table of the results).
Measuring sexual orientation
8 In light of published data, we distinguished three dimensions of individuals’ sexual orientation: (i) sexual activity category, (ii) self-reported sexual orientation, and (iii) sexual attraction. The sexual activity category measures bisexuality in terms of the sex of sexual partners over a given period of time. A distinction is therefore made between men who have sex with men exclusively (MSME) and men who have sex with both men and women (MSMW), with the reference period indicated in brackets if relevant. Self-reported sexual orientation is captured in most surveys as a “How would you define yourself?” question with pre-coded response modalities, most often “homosexual” and/or “gay,” “bisexual,” “heterosexual”. Sexual attraction to one or the other sex is usually measured with a question such as “Are you attracted only to men; rather to men but also to women; to both men and women; rather to women but also to men; only to women?”.
9 For each publication selected, and following the dimensions of sexual orientation used by the authors, we obtained the reported HIV prevalence and incidence as well as the comparative ratio between bisexuals and exclusive gay men: odds ratio (OR) or prevalence ratio (PR) for prevalence and hazard ratio (HR) or incidence risk ratio (IRR) for incidences. Depending on the publication, the comparative ratios could be unadjusted, weighted according to RDS (respondent driven sampling, a sampling technique used by the majority of surveys) (8), or adjusted (multivariate analysis, the cofactors considered varying from one study to another). Some articles report comparative ratios without indicating the observed prevalence. For those studies which did not publish a comparative ratio, we calculated an unadjusted OR from the data in the article. Ratios were standardized to systematically compare bisexuals (including or not including heterosexuals) with exclusive gay men, depending on the dimension considered. Some studies compared bisexuals and exclusive gay men on several dimensions.
10 Given the significant heterogeneity in terms of methods and indicators, it seemed preferable not to perform a statistical meta-analysis.
11 We pre-identified and retrieved between 400 and 500 publications, then reviewed the full texts of 355 of these and identified 62 measures of association between bisexuality and HIV prevalence (corresponding to 41 publications – a single publication may present several different indicators of bisexuality – covering 38 different surveys) and 8 measures of association between bisexuality and HIV incidence (corresponding to 6 publications covering 5 different surveys).
Bisexuality and prevalence of HIV
13 HIV prevalence among bisexuals and exclusive gay men was documented for 48 (77%) of the 62 identified measures of association (Figure S2). In the majority of cases (43/48), the prevalence measured among bisexuals was lower than those measured among exclusive gay men (points positioned above the diagonal line). In 4 cases, the prevalence was higher; in one case, the same value was observed among bisexuals and exclusive gay men.
14 A similar result can be seen in Figure 1, which reports 62 measures of comparison between bisexuals and exclusive gay men, mainly odds ratios (for 58 of 62 measures) but also prevalence ratios (4/52). The ratio is greater than 1 (indicating that bisexuals are more infected than exclusive gay men) for only 9 (14.5%) measures, equal to a 1 for 1 (1.6%) measure and less than 1 for 52 (83.9%) measures. In half of the cases (32/62), the 95% confidence interval included the value 1, indicating no significant difference. Of the 30 measures with a significant effect, 26 had a ratio of less than 1.
15 Only 4 measures had a ratio significantly greater than 1. Two had a low confidence interval close to 1: in Ghana in 2017 (9) with an unadjusted odds ratio of 1.18 [95% Confidence Interval: 1.00-1.38] and in Benin in 2016 (10) with a weighted odds ratio of 1.14 [1.01-1.29]. Note that, in the absence of RDS weighting, the odds ratio falls below unity (OR: 0.77 [0.42-1.39]), and the prevalence observed in bisexuals was lower than that observed in exclusive gay men (11.7% vs 14.6%). The only study to report odds ratios greater than 2 was conducted in Tanzania in 2014 in two towns/cities (11,12) and reported adjusted odds ratios of 5.6 [3.9-12.8] in Dodoma and 8.0 [4.1-15.6] in Dar es Salaam.
Bisexuality and incidence of HIV
16 Five studies measured and compared HIV incidence between bisexuals and exclusive gay men (Table 1). Only one study (13), conducted in Benin between 2016 and 2018, reported higher incidence among bisexuals (7.7 per 100 person-years [5.1-11.7]) compared to exclusive gay men (4.9 [3.3-7.2]), a non-significant difference (unadjusted HR: 1.5 [0.9-2.7]).
17 The other studies observed a significantly lower incidence of HIV among bisexuals, regardless of the sexual orientation dimension considered. A South African study (14) conducted between 2015 and 2017 compared those identifying as bisexual, heterosexual, or other with those identifying as gay (adjusted HR: 0.09 [0.01-1.0]). In Kenya, a 2005 to 2011 study (15) compared MSMW (past 3 months) with MSME (adjusted HR: 0.27 [0.2-0.5], p<0.001). One article (16) reported follow-up data from this same cohort over 2005 to 2016 and found similar results (adjusted IRR: 0.50 [0.3-0.9], p=0.016).
Figure 1. Odds Ratio (OR) or Prevalence Ratio (PR) of the probability of being infected with HIV for a bisexual compared to an exclusive homosexual, in sub-Saharan Africa (2004-2018), according to different dimensions of sexual orientation (sexual activity, declared sexual orientation, sexual attraction)
Figure 1. Odds Ratio (OR) or Prevalence Ratio (PR) of the probability of being infected with HIV for a bisexual compared to an exclusive homosexual, in sub-Saharan Africa (2004-2018), according to different dimensions of sexual orientation (sexual activity, declared sexual orientation, sexual attraction)MSMW men who have sex with men and women, according to the time period considered; MSME: men who have sex with men exclusively; RDS: respondent driven sampling. Horizontal bars indicate 95% confidence intervals.
18 A Nigerian study (17) conducted between 2013 and 2018 reported a lower incidence by both sexual activity categories measured over the past 12 months (10.1 per 100 person-years vs 22.9) and by self-reported sexual orientation (bisexual vs gay, 12.5 vs 23.1). In this article, the unadjusted (unweighted) HRs are significantly (p<0.05) lower than unity (0.46 and 0.56, respectively). When RDS weighting is applied, the HRs remain below unity (0.82 and 0.83) but are no longer significant (p>0.05).
19 Lastly, a final study (18) conducted in Malindi, Kenya in 2016/17 reported similar incidences (no significant difference) comparing MSMW (3 months) and MSME (adjusted IRR: 1.25 [0.3-5.0], p=0.798). Unlike previous studies which did not distinguish gender, this one categorized participants defining themselves as female separately, given their significant number.
20 It is noteworthy that a Kenyan study of only MSM sex workers in Nairobi between 2009 and 2012 (19) also reported a lower incidence of HIV among those who reported having had at least one vaginal intercourse in the past three months (unadjusted HR: 0.76 [0.4-1.6], p=0.489).
Table 1. Incidence of HIV by self-reported sexual orientation, gender identity and sexual attraction in quantitative surveys of MSM in sub-Saharan Africa (2005-2018)
Table 1. Incidence of HIV by self-reported sexual orientation, gender identity and sexual attraction in quantitative surveys of MSM in sub-Saharan Africa (2005-2018)For multi-site surveys, results are presented in multiple lines. Self-reported gender identity and sexual orientation are reported using the terms used in the original publication (which may differ from the terms and/or language of the questionnaire). RDS: Respondent Driven Sampling. MSMW: men who have sex with men and women. MSME: men who have sex with men exclusively. For sexual activity, the duration cited in brackets indicates the time period considered. HR: hazard ratio. IRR: incidence risk ratio. For ease of reading, eligibility criteria related to being HIV-negative upon inclusion, consent, geographic residence and/or not participating in another study are not presented. Brackets indicate 95% confidence intervals.
Bisexual practices and risky behaviours
21 The high proportion of heterosexual practices among MSM in Africa is often thought to result from a fear of stigmatization and thus a need to camouflage themselves under the guise of a heterosexual lifestyle. Some consider that this increases the risk of spreading HIV to the rest of the population (implied to be “heterosexual”) through the intermediary of the MSM’s female partners, but also among the MSM themselves. Furthermore, there is sometimes a psychologizing view that bisexuals are less able to accept themselves as gay and, therefore, less able to take care of themselves and protect themselves against HIV. The results of the surveys available provide insight into the factors of exposure to risks from another angle.
22 Whether considering HIV prevalence or incidence, sub-Saharan bisexuals are generally less infected with HIV than exclusive gay men, regardless of the dimension considered to define bisexuality: sexual activity, reported sexual orientation, or sexual attraction, and although they do not overlap precisely, these three dimensions are highly correlated (7,20,21). These differences in prevalence and incidence do not necessarily imply an intrinsic susceptibility to HIV risk related to having bisexual or exclusive same-sex practices but may result from behavioural differences that impact the risk of HIV exposure.
23 From a prevention perspective, most studies found no significant difference in the frequency of condom use between bisexuals and exclusive gay men (5,22-25), and a few measured a higher likelihood of condom use among MSMW (26,27). In contrast, bisexuals appear to have sex with men less frequently and have fewer sexual partners (28,29). Therefore, the likelihood of having unprotected anal intercourse would be lower (30,31) or equal (28,32,33) among bisexuals compared with exclusive gay men. Bisexuals generally reported having had their first sexual intercourse with a man later in life (20,28,29) and had a lower total number of male sexual partners in their lifetime (28,34)
24 Most importantly, it appears that bisexuals consistently engage in less receptive anal sex and more insertive sex (19,20,23,28,29,33,35,36). This finding is consistent with a study (37) conducted in South Africa between 2015 and 2016 highlighting how bisexuals (whether self-defined as “bisexual”, “heterosexual” or “other”, or those who reported having had at least one female sexual partner in the past 12 months) had lower prevalence of anal chlamydia and gonorrhoea but higher prevalence of urethral chlamydia and gonorrhoea. Bisexuals are therefore less exposed to HIV as it has been shown that the likelihood of acquiring HIV during receptive anal sex is 15 to 20 times greater than during insertive anal sex (38).
Bisexuality and perception of risk
25 Bisexuals are often perceived, especially by MSM themselves, whether exclusive or not, as being at greater risk than exclusive gay men, based on the assumption that bisexuals are less likely to comply with preventive guidelines. This is reinforced by the idea reported in several surveys that sex with a woman is more (or as much) at risk of transmission as sex with a man.
26 In surveys that have measured this representation, most participants do not perceive sex between men as being “riskier” than sex with a woman. In Abidjan, Côte d’Ivoire, in 2011/12, 13% of participants identified men as being the riskier partners, 23% identified women, and 66% considered both men and women to be equally risky (39). In Kampala, Uganda, in 2012/13, these proportions were respectively 15.6%, 34.8% and 49.5% (40). In Lesotho in 2014, only 28.4% of respondents in the town of Maseru and 22.9% in Maputsoe knew that anal sex was riskier (41). In Burkina Faso in 2013, these proportions were 8% in Ouagadougou and 25% in Bobo-Dioulasso (42).
27 Similarly, several studies have shown that most respondents did not perceive receptive anal sex as being riskier than insertive anal sex in terms of HIV infection. In Nigeria in 2013/16, only 42.7% of respondents knew this (43). This compares to 46.3% in Mesuru and 59.7% in Maputsoe, Lesotho in 2014 (41); 15% in Ouagadougou and 16% in Bobo-Dioulasso, Burkina Faso, in 2013 (42).
Access to care and prevention services
28 Although bisexuals appear to be less exposed to HIV than exclusive gay men, they have less overall access to prevention and care services. Thus, concerning access to HIV testing, the reviewed articles show similar (31,44-47) or lower (20,26,48) testing rates. A Nigerian study documented a lower, but not significant, probability of early initiation of antiretroviral therapy among HIV-infected participants who identified themselves as bisexual (OR: 5,67 [0,30-1,42], p=0,148) (49).
29 These differences in access to services may reflect differential degrees of insertion in MSM sexual and social networks. Stahlman et al. combined data from three RDS surveys conducted in Lesotho, Swaziland, and Malawi and analyzed the results according to waves of RDS recruitment (50). Differences emerged between early recruitment waves (waves 0-3, indicative of the core of the recruitment network) and later recruitment waves (8-14, on the fringe of the network): more peripheral participants defined themselves as bisexual or heterosexual (73% vs 51%), less HIV-infected (15% vs 48% prevalence), but also less aware of their HIV status (33% vs 53%).
30 On all dimensions, bisexuals are less exposed to HIV than exclusive gay men, in part due to behavioural differences. Thus, it is not exclusively same-sex practices or identities that in themselves expose people to HIV, but rather they are associated with greater insertion in the network of sexual exchanges between men and with a higher cumulative exposure factor.
31 However, it is problematic to think in terms of these two categories alone. Bisexuality is multifaceted in sub-Saharan Africa: bisexuals are not a homogeneous group, and sexual identities are so diverse that simply contrasting exclusive gay men with bisexuals is reductive (7,28). The standardization of questionnaires and the lack of attention to the social environment of practices tend to decontextualize the reported data.
32 From a health perspective, it is crucial that prevention and treatment programs for MSM take their diversity into account and thus design differentiated services. Although responses specifically targeting this population, particularly those provided by community NGOs, have been – and remain – essential to improving access to health services, they are not necessarily welcomed by all MSM: some say that they benefit from a more attentive ear and feel more confident, while others prefer to use services intended for the general public and to avoid those dedicated to MSM, for reasons of discretion, fear of the gaze of others and fear of being recognized (51). Although those who do not identify themselves as gay or who are less involved in MSM socialization networks are more frequently found in the second group, the gay/bisexual distinction does not overlap exactly with preferences for health services. This suggests the need, beyond this one dimension, to adapt health services to the diversity of situations, expectations and health needs of MSM and their female partners in sub-Saharan Africa.
33 Table S1. HIV prevalence by self-reported sexual orientation, gender identity and sexual attraction in quantitative surveys of MSM in sub-Saharan Africa (2004-2018).
34 Figure S2. HIV prevalence measured among bisexual and exclusively gay MSM in sub-Saharan Africa (2004-2018), by self--reported sexual orientation, gender identity and sexual attraction
35 NB: we have deposited this appendix on Zenodo to make it accessible via a dedicated DOI, which is cited in the body of the manuscript.