Progress toward UNAIDS 90-90-90 targets: A respondent-driven survey among female sex workers in Kampala, Uganda

Background We investigated progress towards UNAIDS 90-90-90 targets among female sex workers in Kampala, Uganda, who bear a disproportionate burden of HIV. Methods Between April and December 2012, 1,487 female sex workers, defined as women, 15–49 years, residing in greater Kampala, and selling sex for money in the last 6 months, were recruited using respondent-driven sampling. Venous blood was collected for HIV and viral load testing [viral load suppression (VLS) defined as <1,000 copies/mL]. We collected data using audio computer-assisted self-interviews and calculated weighted population-level estimates. Results The median age was 27 years (interquartile range: 23 to 32). HIV seroprevalence was 31.4% (95% confidence interval [CI]: 29.0, 33.7%). Among all female sex workers who tested HIV-positive in the survey (population-level targets), 45.5% (95% CI: 40.1, 51.0) had knowledge of their serostatus (population-level target: 90%), 37.8% (95% CI: 32.2, 42.8) self-reported to be on ART (population-level target: 81%), and 35.2% (95% CI: 20.7, 30.4) were virally suppressed (population-level target: 73%). Conclusions HIV prevalence among Kampala female sex workers is high, whereas serostatus knowledge and VLS are far below UNAIDS targets. Kampala female sex workers are in need of intensified and targeted HIV prevention and control efforts.


Introduction
In 2014, the Joint United Nations Programme on HIV/AIDS (UNAIDS) launched three treatment targets to achieve epidemic control by 2020; that is, 90% of all people living with HIV will know their HIV status, 90% of those diagnosed with HIV infection will receive antiretroviral therapy (ART), and 90% of all people receiving ART will have viral suppression (90-90-90 targets) [1].
Key populations such as female sex workers are considered a priority population in efforts to achieve 90-90-90 targets [1]. Female sex workers bear a disproportionately large burden of global HIV infections and are estimated to be 13.5 times more likely to become infected than the general population of women in low and middle income countries [2]. Multiple and concurrent sexual partners, inconsistent condom use, and a high prevalence of sexually transmitted infections (STI) alongside structural determinants expose female sex workers and, subsequently, their clients and partners to a greater risk of HIV acquisition [2]. In sub-Saharan Africa, an estimated 18% of the total HIV burden among women 15 years or older has been attributed to female sex work [3] and the estimated HIV prevalence among female sex workers is 36.9% [2].
Uganda has a mature, generalized HIV epidemic; however, sex work is likely an important driver of sexual transmission [4]. According to the Uganda AIDS Commission, sex work accounted for an estimated 11% of new infections in 2009 [4]. A 2009 study found an HIV prevalence of 33% among Kampala female sex workers [5] compared to 9.5% among the general population of Kampala females of reproductive age [6]. Despite that, access to HIV prevention and treatment services for female sex workers is limited [2,5]. The high prevalence of HIV in Uganda suggests that there are significant barriers for female sex workers to obtain essential health services [7,8]. In Uganda, sex work remains criminalized, increasing stigma and marginalization [9]. Factors such as poverty, discrimination, gender inequality, severe physical violence, and criminalization of sex work increase female sex worker's risk for infection and deter these women from learning their HIV status or accessing prevention and treatment services [7,10].
Scale-up of prevention and treatment programs could reduce HIV transmission among female sex workers, however, the effectiveness of these interventions are dependent on the extent to which Ugandan female sex workers engage in prevention and treatment services. Being tested and knowing one's HIV status is associated with a reduction in HIV risk behaviors, prevents onward transmission, and can lead to mobilization of support networks [11], increasing the ability to make informed decisions and, in turn, utilize necessary health services [12]. There is a dearth of population-level data examining the burden of HIV among female sex workers and the uptake of HIV services (from testing and diagnosis to virological suppression), mainly due to challenges in representative population-based sampling [13]. To address this gap in knowledge, we utilized data from the 2012 Crane Survey, a key population focused survey in Kampala, Uganda. We report here on the estimated prevalence of HIV among female sex workers, the population-level progress toward 90-90-90 targets, and factors associated with serostatus knowledge, ART use, and viral load suppression (VLS).

Study design
Female sex workers were recruited using respondent driven sampling (RDS), a modified form of chain-referral sampling with a mathematical system for weighting, suitable for hard-to-sample populations [14,15]. Recruitment occurred from April to December 2012. Women were eligible if they were 15 years or older at the time of recruitment, living in greater Kampala, and sold sex to men in the preceding 6 months. Survey staff identified and enrolled four female sex workers as "seeds" (initial participants) to initiate recruitment of other female sex workers in their social networks through the use of coupons. Seeds were well-connected within their networks, well regarded by their peers, sympathetic to the survey's goals and diverse with regard to education, socioeconomic status, age, and place of residence. These "waves" of recruitment were repeated until the desired sample size was achieved (maximum number of waves achieved: 25). All variables of interest reached convergence. Participating women were provided ($4.00 USD) for their time and travel to the interview site, in addition to an incentive for each successful peer recruitment ($1.25 USD per recruit). Initially, each participant was provided three coupons; this was later reduced to two coupons, then to one, and finally to zero, to bring sampling to a controlled end.

Data collection
Eligible and consenting women were interviewed using audio computer assisted self-interview (ACASI) in either Luganda or English. The interview's key domains included demographics, lifetime sexual characteristics, sexual behaviors in the last three months, sexual violence, and HIV testing and treatment history. Other data measures included alcohol use and drug use (including injection drug use).

HIV testing
All HIV testing was voluntary. Ugandan ministry of guidelines for HIV counseling, testing and referral services were followed [16]. Pre-test counseling included an explanation of HIV infection and transmission, the meaning of HIV test results, risks associated with sexual behaviors, as well as means for HIV prevention. Post-test counseling messages were tailored to recruits' HIV result and risk profiles and included goals, means, and strategies for behavioral risk reduction, maintenance of risk reduction, and explanation of risk reduction methods (e.g., condom use, drug use). Counseling also included an assessment of psychosocial needs, a discussion of living with HIV-infection, treatment and care, and issues related to discrimination. All HIV-positive FSW were provided a referral and study staff offered to accompany the individual.

Laboratory procedures
At their initial visit, female sex workers provided a venous blood sample for HIV and syphilis testing and, if HIV-positive, for CD4+ T cell count and viral load. HIV serologic testing was conducted at the Uganda Virus Research Institute (UVRI) laboratory in Entebbe using Viro-nostika1 Uniform II plus O2, 3 rd generation (bioMeriéux, Marcy l'Etoile, France) and Murex HIV Ab, 3 rd generation (Abbott Laboratories, Abbott Park, Illinois, U.S.A.) in parallel. STAT--PAK (Inverness Medical, Princeton, New Jersey, U.S.A.) was used as a tiebreaker for discordant results. Respondents with concordent reactive results or STAT-PAK reactive results were classified as HIV seropositive. HIV seropositive specimens were further tested for viral load using COBAS TaqMan1 assay. An undetectable viral load was defined as <50 copies/ml and viral load suppression was defined as <1000 copies/ml, chosen based on World Health Organization guidelines [17]. Plasma was also tested for Treponema pallidum (TP) infection, using the anti-syphilis IgG ELISA (Biotec Laboratories, Suffolk, UK) and, if reactive, the rapid plasma reagin (RPR) Syfacard-R Test (Murex Biotech, Dartford, UK). Respondents with RPRreactive test results were classified as having active TP infection.

Statistical methods
In total, 1,497 female sex workers were included in the analysis, including 2 of the 4 seeds. Two seeds were unproductive and did not recruit other female sex workers. RDS Analyst (RDS-A) v5.7 was used to generate unweighted and weighted descriptive statistics [18]. We utilized Gile's sequential sampling estimator [19], with 100,000 resamples for bootstrapping to generate weighted population estimates and 95% confidence intervals (CI) for demographic and behavioral characteristics.
We defined serostatus knowledge as an HIV seropositive female sex workers who reported a previous positive test result. Female sex workers were classified as on ART if she self-reported ART use among those who reported a previous positive test result. VLS was defined as <1,000 copies/ml among those female sex workers self-reporting ART use.
Population-level estimates of each unconditional 90 target were calculated using all HIV seropositive FSW as the denominator, i.e., the proportion of all HIV seropositive female sex workers who reported a previous positive test, reported ART use, or were virally suppressed.
Weighted logistic regression was used to examine bivariate associations between serostatus knowledge, ART use, or viral suppression and demographic characteristics, sexual behaviors, substance use, drug use, HIV knowledge, and STI history. All regression analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC).

Sensitivity analyses
We conducted a sensitivity analysis for population-level estimates of 90-90-90 targets by assuming that any female sex workers with viral suppression was taking ART at the time of the study, and further that all female sex workers on ART knew their serostatus. We also varied the definition of viral load suppression. We calculated population level estimates for 90-90-90 targets with viral load suppression defined as <50 copies/ml and <200 copies/ml.
The study protocol and consent procedures were approved by the human subjects protection boards at Makerere University School of Public Health, the Uganda National Council of Science and Technology (UNCST), and Centers for Disease Control and Prevention (CDC). Informed consent was obtained verbally and separately for interview, blood draw and testing, as well as storage of specimens for future testing. No personal identifiers were collected; biomarker results were returned to recruits at their scheduled return visit, three weeks following the initial visit. Recruits testing positive for syphilis were offered treatment; those testing HIVpositive were referred to care and ART per national guidelines. Female sex workers between the ages of 15-17 years were treated as emancipated minors and provided informed consent themselves.

Sampling and demographics
Of 4,018 coupons issued, 1,915 (47.7%) were redeemed. Of the 1,915 redeemed, 1,501 (78.4%) women were eligible to participate and 1,497 completed the visit and were included in the analyses. Women were ineligible if they had not sold sex in the 6 months prior and were under the age of 15. Unweighted demographic characteristics and weighted population estimates are reported in Table

Population-level 90-90-90 estimates
Among seropositive female sex workers, serostatus knowledge was 26.8%, 18% reported being on ART, and 35.2% were virally suppressed (Fig 1). Under the assumption that virally suppressed (<1,000 copies/ml) female sex workers were on ART and thus knew their serostatus, knowledge increased to 45.5% and the proportion of those on ART increased to 37.8%.

Correlates of being outside the 90-90-90 targets
Female sex workers who started sex work at the age of 25 or older had 0.57 (95% CI: 0.33, 0.99) times the odds of not knowing their serostatus (Table 3) and 0.47 (95% CI: 0.20, 1.08) times the odds of not being on ART compared to female sex workers who started sex work under the age of 25. Compared to protestant female sex workers, catholic female sex workers had 2.26 (95% CI: 1.14, 4.48) times the odds of unsuppressed viremia. Female sex workers who had been working 3 to 5 years had 0.53 (95% CI: 0.29, 0.96) times the odds of having unsuppressed

Discussion
HIV prevalence among Kampala female sex workers (31.4%) is more than three times the HIV prevalence of the general female population of reproductive age in Kampala (9.5%) [6]. The high prevalence of unsuppressed viremia (14.7%) among seropositive female sex workers is concerning given the association with decreased survival and increased onward transmission. Even after informing our population-level 1 st and 2 nd 90's by respondents' VLS status, we still found that only a minority of HIV positive female sex workers were aware of their HIV status (45.5%), only 37.8% were on ART, and just 35.2% were virally suppressed. These estimates are  [5,20]. Sex work in Uganda remains criminalized; and enforcement of prohibitive sex-work policies are associated with a limited ability to negotiate safer sex practices and increased risk of HIV infection [21,22]. Among the target population, 71.9% had been previously tested. Nevertheless, 28.1% of female sex workers indicated they had never been tested for HIV and 32.4% had not been tested in the last 12 months. These low estimates are concerning given that current recommendations suggest sex workers test every 6 to 12 months [17].
Among seropositive female sex workers, self-reported serostatus knowledge was low and likely to be an underestimate as demonstrated in previous studies [23]. We attempted to correct for this using viral load biomarker data. Under the assumption that female sex workers with VLS were aware of their HIV status and on ART, serostatus knowledge increased to 45.5%. Generally, serostatus knowledge among sex workers in sub-Saharan Africa has been notably low. In a 2011 Zimbabwe study, only half of HIV-positive female sex workers were aware of their status and less than 25% of negative female sex workers reported testing in the previous 6 months [24]. In three cities in Mozambique, serostatus knowledge among female sex workers ranged from 10.4 to 51.9% [25]. As with the general population, female sex workers face similar barriers to HIV testing including distance to the testing site, inconvenient schedules, and  1 Serostatus knowledge (yellow) includes all HIV positive women who self-reported a previous positive test result; blue column also includes FSW with viral suppression. 2 On ART (yellow) includes all HIV positive women who self-reported taking ARTs; blue column also includes FSW with viral suppression. 3 Viral suppression was laboratory-based and defined as <1000 copies/ml. 4 Denominator for viral suppression was n = 341 due to missing data; the denominator for all other characteristics was n = 485. 5 Error bars represent 95% confidence intervals. lack of awareness of services [26,27]. In addition, fear of authorities, concerns about confidentiality, and that possibility that clients or others may learn their status and occupation, impede female sex workers from accessing services [28,29]. Furthermore, female sex workers are more likely to experience gender-based violence as reported previously [30]. Almost half of female sex workers in our study population experienced rape in their lifetime and about onethird experienced three or more rape occurrences in the 6 months prior to the survey [30]. Increasing female sex workers empowerment has been associated with positive outcomes, including a reduction in HIV prevalence and increased utilization of health services [31,32]. Developing more innovative approaches to delivering HIV testing and counseling services could increase female sex workersengagement.
There was marked discordance between actual serostatus and perceived serostatus. While 48.7% of seropositive women (among those previously tested for HIV) reported a previous positive test result, only 7.8% of seropositive women believed themselves to be positive and 79.3% indicated they did not know their status. We speculate that women on ART with VLS were more likely to believe they were no longer HIV-positive. Whether women are embarrassed or frightened to acknowledge their serostatus, or believe that ART is a 'cure,' additional education and counseling is urgently needed. blue column also includes FSW with viral suppression. 2 On ART (yellow) includes all HIV positive women who self-reported taking ARTs; blue, green, purple columns also includes FSW with viral suppression by definition. 3 Denominator for viral suppression was n = 341 due to missing data; the denominator for all other characteristics was n = 485. 4 Error bars represent 95% confidence intervals.
https://doi.org/10.1371/journal.pone.0201352.g002 We speculate that the self-reported estimate for ART use is an underestimate. Female sex workers were only asked if they were on ART if they had previously reported a positive HIV test result. Like serostatus knowledge, we corrected our ART estimate by including women with VLS. Our estimate of 37.8% is similar to past studies in surrounding countries. In Mozambique, only 30-40% of eligible female sex workers self-reported ART use [25], and similarly in Zimbabwe, 26-38% of female sex workers self-reported ART use [24]. These estimates suggest the need to strengthen linkage to care among female sex workers.
VLS was low (35.2%) compared to other studies among female sex workers. In a 2014 Malawi study, almost half of HIV-positive female sex workers were virally suppressed [33]. Similar results were found in Zimbabwe, where 49.5% of female sex workers had viral loads of <1000 copies/ml [34]. We defined VLS as <1000 copies/ml, however other studies have used different thresholds, therefore we conducted a sensitivity analysis varying the definition of VLS [35,36]. In Uganda, low service utilization as evidenced by the limited number of women on ART appears to be a major contributing factor in this population. The differences in time on ART and varied levels of adherence may also play a role [34]. For example, among female sex workers in Burkina Faso, high levels of viral suppression after 6 to 36 months on ART (79-82%) were correlated with high levels of ART adherence after 6 to 36 months (83-100%) [37]. Additionally, our corrected estimates could represent overestimates if the population includes a higher proportion of "elite controllers" who maintain undetectable viral loads without therapy, as evidenced in other studies conducted in Kenya and Uganda [38,39].
There are number of limitations to our study. Female sex workers who received a coupon had to travel to the survey site to participate. Eligible female sex workers were then asked to provide self-reported data, which could be subject to recall bias. Social desirability bias could impact our results, as under-reporting of factors related to HIV, such as the number of clients, known serostatus, and self-reporting of ART are possible. We aimed to minimize these biases through the confidential nature of ACASI. Viral load data was not available for 29% of seropositive female sex workers and the questionnaire did not ask about the date of diagnosis, nor did it include time to treatment initiation or time since ART initiation, making qualification of VLS difficult. Additionally, HIV risk dynamics among female sex workers can vary substantially by the type and structure of the sex work and can change over time [40,41]. Our survey questions included place of primary work, although it was not possible to select multiple locations and decipher changing environments over time. The illegality of sex work in Uganda likely influences the location of sex acts [42] and also may have discouraged participation in the survey. Moreover, RDS is dependent on the social connections of the recruitment participants and we could have missed women who were not socially connected. Additionally, only two of our seeds were productive. Our laboratory analyses were conducted in 2012, however guidance for HIV serological testing has since been updated [43]. Although our VL-adjusted estimates for serostatus knowledge and ART uptake represent a major strength in this analysis, it is possible that some women on ART may have failed to suppress viral load or may have been on ART not long enough to achieve VLS, and, therefore, may not have been classified as knowing their serostatus and being on ART.
Despite these limitations, the high prevalence of HIV and low proportion of HIV-positive female sex workers with VLS is concerning. Reduction of barriers to service utilization among both the general population and female sex workers must be addressed. Increasing testing uptake, more effective linkage to and uptake of ART, along with condom and pre-exposure prophylaxis (PrEP) provision by increasing service awareness and flexibility will be useful. Efforts that facilitate female sex workers' ability to negotiate safer sex-work environments and criminalize abuse are urgently needed. Incorporation of community-and rights-based approaches with tailored prevention packages that included engagement of peer networks may reduce stigma and discrimination. Ongoing surveillance of HIV incidence and VLS in this population should occur to monitor trends in order to inform HIV control and prevention programs. A combination of intensified interventions promoting prevention and treatment are necessary if we are to achieve the ambitious 90-90-90 targets in Uganda.