Despite continued efforts to control HIV epidemics, 1.7 million new HIV infections occurred in 2019, with the greatest disease burden found in sub-Saharan Africa (SSA) [1]. In 2014, the Joint United Nations Programme on HIV/AIDS (UNAIDS) announced its objective to end AIDS by 2030 by considerably increasing diagnosis, treatment, and viral suppression among people living with HIV [2]. To achieve the HIV incidence reduction targets, interventions must prioritize key populations, which include sex workers, men who have sex with men, people who inject drugs, transgender people, and incarcerated people [3]. Key populations have unmet HIV prevention needs and contribute disproportionately to HIV transmission dynamics. Worldwide, over 60% of new adult HIV infections in 2019 were in individuals from key populations and their partners [1]. Even in high HIV prevalence settings, focusing HIV prevention approaches on key populations is important for limiting transmission [4].
Globally, sex workers experience a high HIV burden. Worldwide, an estimated 12% of female sex workers were living with HIV in 2011, reaching 37% in SSA [5]. The increased HIV acquisition risk among female sex workers is exacerbated by structural factors—including criminalization of sex work, stigma, and physical and sexual violence—which undermine sex worker engagement in HIV risk reduction behaviors and prevention [6–8]. Modeling studies suggest that the population attributable fraction of new HIV infections due to sex work ranges from less than 5% to 95%, depending on context [9–14]. This population-level impact is the result of chains of transmission linking sex workers and their clients to partners not involved in sex work [15].
Despite the central position of men who pay for sex in sexual networks, there has been comparatively little attention devoted to systematically reviewing representative epidemiological data on these men and on interventions focused on this population. Clients of sex workers are not designated, nor recognized, as a key population by UNAIDS, in part because of their lack of perceived structural vulnerabilities [1]. However, neglecting this population places the responsibility to prevent HIV transmission solely on sex workers. Developing appropriate interventions for clients of sex workers can be challenging and requires a granular understanding of the population sizes, sexual behaviors, HIV epidemiology, and uptake of HIV prevention interventions of this group. As with other key populations, clients of sex workers are hard to reach, and there can be wide variations in the definition of sex work [16]. Time–location surveys that collect information on clients of sex workers are often limited by their high non-response rates and lack of representativeness [17–19]. In contrast, nationally representative population-based surveys that collect information on paid sex may provide a promising alternative for characterizing men who pay for sex [20,21]. However, these surveys rely on self-reports, which are susceptible to underreporting of stigmatized behaviors such as paid sex [22].
The goal of this study is to improve our understanding of the complex HIV transmission dynamics arising from sex work. To achieve this, we first synthesize national population-based surveys conducted in SSA from 2000 to 2020 that collected information on paid sex ever. Second, we use meta-analyses to estimate population sizes, lifetime number of sexual partners, condom use, and HIV prevalence, testing, and treatment outcomes among men who do, and do not, pay for sex in SSA.
METHODS
Data sources and selection criteria
We searched for nationally representative population-based surveys conducted in SSA over the time period 2000–2020 with available microdata on ever paying for sex (Table A S1 Text). Specifically, we considered Demographic and Health Surveys (DHS), AIDS Indicator Surveys (AISs) (https://dhsprogram.com/methodology/survey-types/ais.cfm), Population-based HIV Impact Assessment (PHIA) (https://phia-data.icap.columbia.edu/), Multiple Indicator Cluster Surveys (MICS) (https://mics.unicef.org/surveys), and other country-specific population-based surveys (e.g., Kenya AIDS Indicator Survey [KAIS] and South Africa National HIV Prevalence, Incidence, Behaviour and Communication Survey [SABSSM]; Table A in S1 Text). We included all available surveys and did not exclude based on survey language.
Variables of interest and definitions
We extracted data on paid sex (ever and past 12 months), lifetime number of sexual partners, condom use during last paid sex, HIV serostatus, HIV testing history (ever and past 12 months), antiretroviral (ARV) use (as determined by ARV biomarker data), and viral load suppression (VLS) among sexually active men. For most surveys, men were identified as having ever paid for sex if (1) they reported that any of their last 3 sex partners was a sex worker or (2) they reported either ever paying for sex or doing so in the past 12 months. Men who had never had sex were excluded.
Data analysis
Using respondent-level data from each survey, we calculated relevant estimands, along with their 95% confidence intervals (95% CIs), for men aged 15–54 years, accounting for complex survey designs (i.e., survey weights, stratification, and clustering). We did not pool estimates if denominators were smaller than 10. We pooled outcomes using inverse-variance-weighted random-effects meta-analysis with the empirical Bayes estimator for heterogeneity. We used I2 statistics to assess heterogeneity across estimates [23]. We calculated the following estimands: pooled proportions of men who paid for sex ever and in the past 12 months; pooled proportions of men who used a condom during their last paid sex; pooled proportions of men who ever tested for HIV; HIV prevalence among men who paid for sex; prevalence ratios (PRs) of HIV, HIV testing history (ever and past 12 months), ever HIV testing among people living with HIV, ARV use, and VLS among men who had paid for sex and those who had not; and mean and ratio of means (log-transformed) of lifetime number of sexual partners for men who had paid for sex and those who had not. Meta-analyses were performed on logit-transformed proportions and log-transformed PRs. Calculations were stratified by regions and by time periods (2000–2009 and 2010–2020). When calculating lifetime number of sexual partners and the PRs for HIV and HIV testing, we standardized results by age and urban/rural residence type.
We performed univariable meta-regression to assess whether the proportion of men who paid for sex, condom use at last paid sex, HIV testing, HIV prevalence, and PRs of HIV and HIV testing (ever and in the past 12 months) varied by survey year and whether the proportion of men who paid for sex varied by age and urban/rural residence type. Meta-regression was performed using logit-transformed proportions and log-transformed PRs, and we assessed time trends by using our models to estimate outcomes in 2010 and 2020. These analyses were not pre-registered. R software was used (4.0.0), and the DHS/AIS data were extracted using the rdhs package [24]. Survey data were analyzed with the survey package [25], and meta-analyses were performed using the metafor package [26]. This meta-analysis was reported in accordance with MOOSE guidelines [27].
Discover more from MEZIESBLOG
Subscribe to get the latest posts sent to your email.