Male Sex Workers Who Sell Sex to Men Also Engage in Anal Intercourse with Women: Evidence from Mombasa, Kenya

Objective To investigate self-report of heterosexual anal intercourse among male sex workers who sell sex to men, and to identify the socio-demographic characteristics associated with practice of the behavior. Design Two cross-sectional surveys of male sex workers who sell sex to men in Mombasa, Kenya. Methods Male sex workers selling sex to men were invited to participate in surveys undertaken in 2006 and 2008. A structured questionnaire administered by trained interviewers was used to collect information on socio-demographic characteristics, sexual behaviors, HIV and STI knowledge, and health service usage. Data were analyzed through descriptive and inferential statistics. Bivariate logistic regression, after controlling for year of survey, was used to identify socio-demographic characteristics associated with heterosexual anal intercourse. Results From a sample of 867 male sex workers, 297 men had sex with a woman during the previous 30 days – of whom 45% did so with a female client and 86% with a non-paying female partner. Within these groups, 66% and 43% of male sex workers had anal intercourse with a female client and non-paying partner respectively. Factors associated with reporting recent heterosexual anal intercourse in bivariate logistic regression after controlling for year of survey participation were being Muslim, ever or currently married, living with wife only, living with a female partner only, living with more than one sexual partner, self-identifying as basha/king/bisexual, having one’s own children, and lower education. Conclusions We found unexpectedly high levels of self-reported anal sex with women by male sex workers, including selling sex to female clients as well as with their own partners. Further investigation among women in Mombasa is needed to understand heterosexual anal sex practices, and how HIV programming may respond.


Introduction
It is now well acknowledged that anal intercourse (AI) is an efficient mode of HIV transmission [1][2][3][4][5][6]. Despite this, the discourse around AI and HIV has largely focused on men who have sex with men, with relatively little evidence available in relation to heterosexual populations. Recent studies have estimated the probability of HIV-1 transmission per unprotected receptive male-to-male AI act at 1.4% [7] and 1.7% per maleto-female AI act [8], compared to a much lower probability of transmission in the range of 0.08-0.3% for receptive vaginal intercourse [8]. These findings suggest that the role of AI in shaping heterosexual HIV epidemics should not be underestimated.
In sub-Saharan Africa, prevention and control of heterosexual HIV transmission has solely revolved around penile -vaginal intercourse despite the evidence of the practice of heterosexual AI [9]. Within various populations sampled in studies undertaken in South Africa, percentages reporting heterosexual AI over different time periods ranged from 43% among female sex workers [10], 42% among truck drivers [11], 10% and 14% among patients of sexually transmitted infection (STI) clinics [12], and 3.6%-15% within the general population [13,14]. A study on female college students in Togo found that 12% had ever engaged in AI [15], whilst 35% of males and females aged 18-27 years sampled in a large, national Zimbabwean study reported heterosexual AI over the past two months [16]. These findings, coupled with the estimated increased risk of HIV transmission from AI, highlight the need to better understand heterosexual AI in sub-Saharan Africa [17].
In Kenya, little research has been undertaken on heterosexual AI. In two studies, 15% [18] and 41% [19] of female commercial sex workers sampled reported lifetime AI, whilst another study had 4.3% of female sex workers reporting AI during the previous month [20]. Here, we aim to add to the current body of evidence on heterosexual AI in Kenya, and in sub-Saharan Africa overall, from surveys undertaken among male sex workers who sell sex to men in Mombasa, Kenya. These male sex workers are among a subgroup of men who have sex with men with many vulnerabilities and a high prevalence of HIV and other STIs [21][22][23]. A qualitative study exploring the social and behavioral determinants of sexual risks among men who sell sex to men in Mombasa found that male-to-male sex is driven by socio-economic needs, same-sex sexual attraction, or a combination of both [24]. Due to factors ranging from sexual pleasure to maintaining trust in a stable relationship, the majority of men interviewed also reported low levels of consistent condom use with clients as well as regular partners. In addition, stigma and discrimination towards same sex relationships meant that many of the men selling sex to men also maintained heterosexual relationships.
Whilst high levels of unprotected AI with men have been documented among male sex workers in Kenya [25], little is known about their heterosexual behaviors in general and even less about AI with women. There is a need to gain a better understanding of the latter given that sexual networks of men who have sex with men in Kenya also include women from the general population [26].
The objectives of this study were to determine the prevalence of heterosexual AI among male sex workers in Mombasa, and to identify socio-demographic characteristics associated with practice of the behavior.

Methods
We analyzed data from two surveys undertaken among male sex workers who sell sex to men in Mombasa. These surveys were carried out as part of a larger study evaluating the impact of peer outreach on HIV knowledge and behaviors among male sex workers, the detailed methodology of which is described elsewhere [21].
In brief, the surveys were carried out from October to December 2006 and February to April 2008. Participants were defined as men aged 16 years of older who had recently sold and were currently willing to sell sex to other men in exchange for money and/or goods. These male sex workers were recruited by trained peer mobilizers at known community sites which had been mapped in a previous study [22]: bars, nightclubs and discos, beach areas and beach bars, public streets, public parks, and other private brothels, businesses and estates. Sampling was based on a probability proportional to size approach from a time-venue sampling frame [25]. Peer mobilizers identified the male sex workers based on an initial visual characterization, followed by a casual conversation to confirm that the subject had recent history of and was currently willing to sell sex to men.
Male sex workers who agreed to participate were interviewed by trained interviewers at secure and centrally located venues. Information on participants' socio-demographic characteristics, self-described sexual identity, sexual behaviors with men and women, HIV prevention knowledge and practices, reported STI symptoms, and exposure to health services was collected based on a structured behavioral questionnaire using handheld computers.

Ethics Statement
Ethical approval was obtained from the Kenyatta National Hospital Ethics and Review Committee and the Population Council's Institutional Review Board.

Statistical Analysis
For the purposes of this study, data on study participants' socioeconomic characteristics and sexual behaviors with women were analyzed. Statistical analysis was carried out using the Intercooled Stata 12.0 statistical package (StataCorp, College Station, Texas). To begin with, the socio-demographic characteristics and heterosexual behaviors of male sex workers were analyzed using descriptive statistics. As described previously [25], here, sexual identity was grouped into two categories: ''Basha, king or bisexual'' who are more likely than other identities to report being insertive partners during anal sex, and ''shoga, queen, gay, homosexual or others'' who are more likely to report being receptive anal sex partners.
Even though the 2006 and 2008 surveys were independent, some male sex workers may have participated in both surveys as personal identifiers were not used. Thus, with the goal of combining both surveys for analysis, we first assessed sociodemographic characteristics and heterosexual behaviors for each survey separately, and compared differences in results using chisquared tests for categorical variables and Wilcoxon rank-sum test for medians in order to determine whether data from both surveys could be combined in analysis. Following this, the same descriptive analysis was carried out on the entire sample (surveys combined).
In a second analysis which included all participants from both surveys, socio-demographic characteristics associated with recent (during the previous 30 days) vaginal or oral sex, and with recent AI were identified in bivariate logistic regression. The group 'recent vaginal or oral sex' was defined as those who reported either or both of those sexual behaviors. This definition did not exclude men who may have also recently practiced AI in addition to vaginal or oral sex. As men who participated in the 2006 survey may have also participated in the 2008 survey, bivariate logistic regression was carried out after controlling for year of survey participation.

Results
Separate analyses of data from both surveys showed that several socio-demographic characteristics of men surveyed in 2006 were significantly different (p,0.05) from those surveyed in 2008, but that reported sexual behaviors with women were similar (Table 1). This suggested that the data from both samples could be combined. A total of 425 male sex workers were surveyed in 2006, and 442 in 2008.
The median age of participants was 25 years (IQR = 22-28), and nearly half (45%) had completed primary education (Table 1). Just over half (53%) were Muslim and self-identified themselves as basha, king, or bisexual (57%). Close to 75% of participants reported ever having had sex with a woman, and a much smaller percentage, 20%, were ever married to a woman. Nineteen percent of the entire sample were living with a male sexual partner only, and 7% with a female sexual partner only. Close to 2% were living with more than one sexual partner, or with a sexual partner and their wife. One quarter of male sex workers stated that they had children of their own.
A third of male sex workers (34%, n = 297) within the combined sample recently had sex with a woman. More specifically, within this group of men, 45% had recent sex with a female client and 86% with a non-paying female partner. Thirty-one percent had sex with both a female client and a non-paying partner. Vaginal intercourse was widely practiced, with over 80% of men in both groups (female client or non-paying partner) reporting this behavior. Significant percentages of men also reported having heterosexual AI: 66% (88/134) of those who recently had sex with a female client, and 43% (108/252) of those who recently had sex with a non-paying female partner. In the former group, 60% of men used a condom during AI, whilst in the latter group the figure was 54%. Reported condom use was higher for vaginal intercourse, at 73% for men who had sex with female clients and 60% for those who had sex with non-paying female partners. Table 2 shows the socio-demographic characteristics associated with report of recent AI in male sex workers who sell sex to men, after controlling for survey year in bivariate logistic regression. Characteristics associated with AI were being currently married (OR, 2.72; 95% CI, 1.24-5.98), having ever been married (OR, 3.01; 95% CI, 1.99-4.54), and having children of their own (OR, 3; 95% CI, 2.02-4.46). In addition, men who reported recent AI

Discussion
This study is one of the first to examine heterosexual AI behaviors among male sex workers who sell sex to men in Kenya. Our results show that whilst vaginal intercourse is the more widely practiced heterosexual behavior among male sex workers who sell sex to men in Mombasa, a substantial number of these men are also engaging in AI with women. The proportion of men in this sample reporting heterosexual AI was 65% among men who sold sex to a female client and 43% among men who had sex with a non-paying female partner. We also found that condom use in AI was lower than in vaginal intercourse, which may be due to a general perception that anal sex is safer than vaginal sex as was reported in qualitative discussions with the male sex workers [24]. These results are in line with those from another diary study undertaken among a small cohort of male sex workers in Mombasa, which found that 54% of 215 sexual encounters with women comprised AI. Of these anal sex acts, only 54% were protected -a figure which was lower than that obtained for vaginal sex at 62% [27].
In addition, this study found that, male sex workers engaging in heterosexual AI were more likely to be ever or currently married, living with their wife only, a female sexual partner only, or several sexual partners (male and female) as opposed to with a male sexual partner only, having children of their own, and to self-identify as basha/king/bisexual. These findings suggest that male sex workers who practice bisexual behaviors show closer interactions with women in their daily lives. As mentioned previously, male sex workers in Mombasa who self-identify as basha, king, or bisexual are more likely to be an insertive partner during anal sex with men than other identities [25]. Not surprisingly this analysis also found that bashas/kings/bisexuals were also more likely to report vaginal or anal sex with female partners. These male sex workers likely play a more masculine sexual role with both male and female partners, but more qualitative inquiry is needed to understand these complex social, sexual, and commercial interactions and their public health implications. Additional factors associated with heterosexual AI were being Muslim rather than Protestant, and decreased likelihood of having completed secondary or higher education compared to primary education. There is no evidence to explain why Muslim respondents were more likely to engage in heterosexual AI, although key informants anecdotally report that some young Muslim women in Coastal Kenya may be using AI as a ''virginity preservation'' strategy to avoid vaginal intercourse. More qualitative enquiry into this possible explanation is needed. Given that men who have sex with men are at higher risk of HIV infection than general populations [28], these male sex workers may be acting as a bridge for HIV transmission between higher-risk and lower-risk heterosexual female populations. Paying female clients may also be populations at higher risk of HIV than non-paying female partners. More detailed study of sexual networking among male sex workers, their male and female clients, and non-paying partners are needed to understand the dynamics of HIV transmission among these groups. The surveys were based on self-reported data of sexual behaviors that are stigmatized, and thus subject to underreporting. Similarly, recall bias may have also resulted in an overestimation or underestimation of the frequencies reported here. Another limitation of this study was the inability to compare sociodemographic characteristics of male sex workers who only had anal sex with those who only had vaginal or oral sex, due to small sample sizes. Lastly, whilst we attempted to address repeat participation in the surveys by controlling for year of the survey in logistic regression analyses, it must still be noted that combining both surveys may have also resulted in overestimation or underestimation of frequencies.
To conclude, we found that male sex workers who sell sex to men in Mombasa are also engaging in sex with women, including AI. The reported low condom use in AI, coupled with the high rate of HIV transmission associated with the behavior, suggest that a potential route of HIV transmission is being overlooked. Thus, there may be an under-assessment of heterosexual AI behaviors among different populations, requiring more research to inform HIV prevention programming. In particular, data on the characteristics of women engaging in or paying male sex workers for AI was not collected in this study, and thus there is a need to gain a better understanding of these women.