Introduction

Tanzania has a generalized HIV epidemic with an estimated HIV prevalence of 5.1% and close to 1.5 million people living with HIV (PLHIV) [1, 2]. The epidemic is highly varied across geographical, socio-demographic sub-groups, and key populations. HIV prevalence is twice as high in urban as in rural areas [1] and communities located along major transport corridors and in borders towns have been found to have significantly higher HIV burdens than the rest of the country. Iringa, a major hub in the transportation sector has an HIV prevalence of 16% and Mbeya, a region on the border with Zambia and Malawi, has an HIV prevalence of 9% [1, 3]. Further, key populations, including female sex workers (FSWs), experience a higher HIV prevalence and incidence than the general population and contribute to steady HIV transmission to the general population through sexual contact [1, 2]. FSWs are often not in a position to negotiate safer sex or control the environment in which they live and work, putting them at greater risk of HIV infection.

Worldwide, FSWs bear a disproportionately larger burden of HIV and are a population at high risk of acquiring and transmitting HIV infection [4,5,6]. The HIV prevalence among FSWs in Sub-Saharan Africa was estimated at 36.9%, with a 14 times greater odds of HIV infection compared to women in the general population [6, 7]. Despite evidence suggesting that FSWs were at higher risk of HIV infection while being faced with challenges in accessing HIV care, in 2012, when this Integrated Bio-Behavioral Surveillance Survey (IBBS) was conceptualized, there were few data on these indicators among FSWs in Tanzania. HIV prevalence data were only available for Dar es Salaam, and were insufficient to inform policies and programming [8]. Moreover, with the potential merits of expanding antiretroviral therapy (ART) in populations at higher risk of HIV infection and transmission, such as FSWs [9, 10], it is crucial to understand the HIV burden among them, their risks, vulnerability and access to care. In response to this need, this first national-level IBBS was conducted in 7 regions in Tanzania (Dar es Salaam, Iringa, Mbeya, Mwanza, Shinyanga, Tabora and Mara) in 2013.

This paper presents key findings of the data from the 2013 IBBS, aiming to estimate the population-level prevalence of HIV, syphilis, HSV-2 and factors influencing HIV infection among FSWs in Tanzania. These data are timely since the epidemic in Tanzania is still largely unknown and this was the first HIV and STI study among FSWs conducted at the national level. It is crucial that national health authorities understand the current situation so that appropriate actions can be taken. Additionally, for the first time, we were able to estimate new HIV infection rates from IBBS data. This is of particular importance as there are no estimates of HIV incidence among FSWs in Tanzania and scant data for sub-Saharan Africa [11]. Furthermore, this study enriches the pool of available data on HIV and STI among FSWs in sub-Saharan Africa and globally, and has invaluable implications for future meta-analyses, targeted intervention programs, government and funding agencies, and program evaluations in the country and globally.

Methods

Study Sites

The study sites were selected in careful consultation with the National AIDS Control Program (NACP), Ministry of Health, Community Development, Gender, Elderly and Children of the Republic of Tanzania, aiming to get a national estimate of HIV among FSWs. Seven largest regions representing diverse national geographical and cultural differences were chosen: Dar es Salaam, Mbeya, Iringa, Shinyanga, Tabora, Mara, and Mwanza.

Study Design

FSWs were recruited for a cross-sectional survey through respondent-driven sampling (RDS), an adaptation of snowball or chain-referral sampling where peers recruit their peers into the study. RDS has been validated for over a decade and is considered the best method for reaching hidden and stigmatized populations such as FSWs [12, 13]. To initiate the recruitment of participants, “seed” participants were purposively selected based on a diverse set of characteristics, including age, education, type of sex work, and HIV status. Seeds recruited 3 of their peer FSWs who met the recruitment criteria and each successive peer recruited 3 peers until the desired sample size was reached. Recruiters and their recruits were linked by unique identification numbers on the recruitment coupons, and participant’s network sizes were recorded. These two pieces of information were used to estimate population-based prevalence of key indicators. Detailed information on RDS can be found elsewhere [12,13,14].

Sample Size Estimation

The sample size was calculated using the following formula recommended by Salganik for use in an RDS prevalence study [15]:

$$n = deff.\frac{{P_{A} \left( {1 - P_{A} } \right)}}{{\left( {se\left( {\hat{P}_{A} } \right)} \right)^{2} }}$$

where n = sample size, deff = design effect and P = assumed prevalence. Using the available HIV prevalence data among FSWs in Dar es Salaam (30%), a design effect of 2, and standard error no greater than 0.005, we anticipated a sample size of at least 200 per region. For the three larger regions: Dar es Salaam, Mwanza, and Shinyanga, we purposely aimed to recruit a sample size of 300 per region. Recruitment waves were monitored during data collection to ensure long RDS recruitment chains [16].

Study Population

Eligible FSWs were defined as women aged 15 years and older who reported selling sex for money at least once in the past 3 months and living in one of the 7 regions for at least 3 months prior to the survey. A total of 1914 FSWs living in Dar es Salaam (N = 346), Mbeya (N = 244), Iringa (N = 220), Shinyanga (N = 320), Tabora (N = 229), Mara (N = 205), and Mwanza (N = 350) were recruited between March and September 2013.

Data Collection Procedures

Participants underwent a face to face interview that lasted about 30 to 45 min, after which, they provided a blood sample specimen for HIV and STI testing. All interviews, testing, counseling and return of results were carried out at the study sites. Each region had one data collection site, with the exception of Dar es Salaam where two study sites were set up. Study site selection was based on formative research on specific criteria: convenience, accessibility, confidentiality and safety for participants [17].

Blood samples were tested with HIV rapid tests using the serial algorithm for the detection of HIV antibodies in accordance with Tanzania’s National HIV Testing Guidelines. The whole-blood specimen was tested on site using Determine HIV1/2 as a first test. If the first test was non-reactive, results was recorded as “HIV Negative”. If the first test was reactive, a second test was conducted using Uni-Gold HIV ½ and if reactive, the results were recorded as “HIV Positive”. Discordant results were treated as inconclusive. Syphilis testing followed national guidelines. First, rapid RPR test was conducted. Second, positive RPR tests were sent to the Central Laboratory for Treponema Pallidum Hemagglutination (TPHA) testing as confirmation [18]. Those testing positive on both RPR and TPHA tests were recorded as positive for Syphilis. For HSV2, Biokit HSV-2 rapid test for antibody detection was used on site [19] (Biokit, US).

Data Analysis

Descriptive analysis was conducted to provide the overall characteristics of the study population. Adjusted HIV, syphilis and HSV-2 prevalence with 95% confidence intervals were estimated using RDSAT Software Version 7 [12]. In addition, unadjusted HIV, syphilis and HSV-2 prevalence were calculated using Stata Version 13 (StataCorp., College Station, TX). RDSAT software adjusts for recruitment patterns and the relative sizes of participant’s peer networks, and its prevalence estimate is considered population-based [13, 14]. Individualized weights based on the outcome variable (testing positive for HIV) were then generated in RDSAT and exported to STATA Software for bivariate and multivariate logistic regression analyses [16, 20]. In addition, to improve statistical power, regression analysis was performed on pooled data from all regions, adjusting for regional difference and RDSAT generated outcome weights. Additional details on RDS-type data analysis technique can be found elsewhere [14].

In multivariate analysis, the selection of independent variables was initially determined through literature and theoretical concepts, including socio-demographic characteristics, number of years selling sex, health seeking behavior (previous HIV testing), condom use, experience of violence, and other STIs (syphilis and HSV-2). Next, only variables that were significant at p ≤ 0.15 in bivariate analysis were included in the final multivariate regression analysis. Unadjusted and adjusted odds ratios and 95% confidence intervals were reported. Because syphilis and HSV-2 are highly correlated, two separate multivariate models were constructed: Model 1 includes syphilis as an independent variable, while Model 2 includes HSV-2 as an independent variable.

We estimated HIV incidence by comparing biological HIV result with self-reported HIV status, assuming that infection occurred mid-year. The survey included several questions on: (1) HIV testing in the past 12 months; and (2) testing for HIV (past 12 months) and knowing HIV status (positive or negative), that allows us to estimate the number of new HIV infections. Details of the calculation can be seen in Fig. 1.

Fig. 1
figure 1

HIV incidence estimate

Full size image

Ethical Considerations

The study was approved by the National Institute for Medical Research (NIMR, Dar es Salaam, Tanzania) and PSI Research Ethics Boards (Washington DC, USA). Special precautions were taken to maximize the safety and confidentiality of participants. The survey did not collect participant’s personal information and participants were reminded before being asked to report their HIV status that they could refuse to answer the question. Participants were offered free male and female condoms, lubricants, and risk-reduction educational materials. Participants who tested positive for syphilis and came back for the second visit were informed of their result and referred to STI clinics for free treatment. No adverse events were observed during the study.

An incentive-based strategy was used, in keeping with RDS methodology. Each participant received 10,000 Tanzanian shillings (about 5 USD) for completing the survey and HIV/STI testing, and for each eligible FSW peer a participant brought to the study, a participant would receive an additional 5000 Tanzanian shillings.

Results

Characteristics of the Study Population and Key HIV-Related Indicators

Table 1 describes key characteristics of the study population. The median age of the FSWs was 26 years (Interquartile range 21–31), 63% had never been married, 60% had completed primary education and 20% had some secondary education. About 74% of the FSW participants reported selling sex as their main sources of income.

Table 1 Descriptive characteristics of the study population (N = 1914)
Full size table

The median age at which FSWs first sold sex was 19 years, and on average, FSWs had been selling sex for about 7 years. Nearly 70% of FSWs sold sex at bars, pubs, hotels or guesthouses, 11% on streets, 13% through phone, the Internet or an agent, and 8% sold sex at home. The average number of paying clients in the past 1 month was 33. About 50% of FSWs also reported having a steady or casual sex partner, with whom the participant had a non-monetary relationship.

Only 46% of the FSW participants reported being exposed to any HIV and STI prevention messaging within the past year. About 30% of the FSWs reported using a condom during sex with their last paying client, and 32% with steady or casual sex partners. The main reason for not using a condom was client objection. About 65% of participants reported ever being tested for HIV, and 43% reported testing for HIV within the past 12 months. Being sexually and physically abused in the past 12 months was reported by over 25 and 34%, respectively.

Prevalence of HIV, Syphilis and HSV-2

Table 2 shows the unadjusted and adjusted prevalence of HIV, syphilis and HSV-2. The unadjusted and adjusted HIV prevalence are quite comparable across sites for HIV and HSV-2, indicating that the RDS sampling method used in this study had a low homophily (similar to homophily found during the monitoring of the data collection), and that the network size of the HIV-negative and positive recruiters was quite similar. Adjusted HIV prevalence ranged from 14% in Tabora to 38% in Shinyanga. Of the 7 regions, Mbeya had an HIV prevalence of 29%, while 3 others (Dar es Salaam, Iringa, and Shinyanga) had an HIV prevalence of over 30%. The average HIV prevalence among FSWs in all 7 regions was 28%. The average syphilis prevalence in all 7 regions was 8% with significant higher burden found in Iringa (11%), Mbeya (13%), and Shinyanga (12%). Nearly 60% of the study population was infected with HSV-2, with extremely higher HSV-2 infection found in Shinyanga (75%).

Table 2 Adjusted and unadjusted HIV, Syphilis, and HSV-2 prevalence by regions
Full size table

Estimated HIV Incidence

Of the 398 FSWs who reported having undergone an HIV test in the 12 months prior to the IBBS and who knew that their test result was negative, 50 tested positive for HIV during this IBBS. This yields an incidence estimate of 13 per 100 person-years, assuming that the infection occurred mid-year (Fig. 1).

Factors Associated with HIV Infection

Table 3 describes factors associated with HIV infection (HIV prevalence measure). Bivariate analysis found that age, education, income from sex work, type of sex work, number of years selling sex, condom use with paying client, HIV testing in the 12 months prior to survey, syphilis and HSV-2 were associated with HIV infection. In multivariate logistic regression Model 1 (including syphilis) controlling for study sites and individualized RDSAT weights, we found that higher education [Completed primary: AOR 0.7; 95% CI 0.4–0.96); some secondary: AOR 0.4; 95% CI 0.3–0.7)] and having an HIV test 12 months prior to the IBBS (AOR 0.6; 95% CI 0.4–0.8) were independently associated with lower odds of testing positive for HIV. Syphilis infection (AOR 1.8; 95% CI 1.1–2.8), age (AOR 1.09; 95% CI 1.06–1.11) and number of years selling sex (AOR 1.04; 95% CI 1.01–1.08) were associated with a higher odds of testing positive for HIV. Multivariate logistic regression Model 2 shows the same pattern as Model 1, except that number of years selling sex becomes borderline significant. Odds of HSV-2 were higher among HIV-positive FSWs (AOR 3.3; 95% CI 2.3–4.6), a greater magnitude of association than that between syphilis infection and testing positive for HIV.

Table 3 Factors associated with HIV infections
Full size table

Discussion

This first national-level IBBS survey confirmed high HIV prevalence among FSWs in all major regions of Tanzania. Nearly one-third of FSWs were HIV positive. The prevalence in this population was found to be nearly 4.5 times that among women aged 15–49 in the general population in Tanzania (6%) [1]. This finding suggests that the HIV epidemic among FSWs in Tanzania continues to expand. Further, the substantial regional heterogeneity reflects differing epidemics and risk factors of FSWs living with HIV across the country. The findings also indicate significant differences in HIV burden with age, number of years selling sex, and whether they have had an HIV test in the previous 12 months. Interventions must take into account sub-group disparities such as those based on age and access to HIV testing. The study also found an alarming prevalence of active syphilis and HSV-2 and that these two STIs were independently associated with HIV infection. Regardless of high prevalence of HIV and syphilis, only one-third of the study population perceived themselves as being at high risk of HIV infection, less than half had tested for HIV within the past 12 months, one-third had tested for other STIs in the past 6 months, and less than one-third reported condom-use the last time they had sex with a paying client. The high HIV and STI prevalence and risk patterns, particularly low condom use, seem aligned with the high estimate of HIV incidence of nearly 13%. These findings indicate the urgent need in HIV prevention, care and treatment to reduce HIV infection rates among FSWs in Tanzania. It appears that after decades of prevention programs led by many initiatives, HIV transmission continues to increase rapidly, condom use is suboptimal, and a high proportion of FSWs are unaware of their HIV status.

It is clearly established in this study that STIs were associated with HIV infection. This finding is consistent with previous research that confirms the facilitating role of STIs in HIV transmission [21,22,23]. However, we cannot establish a causal relationship between HIV and other STI infections in this study because we were not able to document the timing of the infection. Therefore, interpretation of this finding must be made cautiously. Syphilis prevalence is unacceptably high, calling for urgent efforts in STI treatment among FSWs in Tanzania. Periodic presumptive treatment (treatment for a presumed infection in a person at high risk of infection, often given at repeated intervals) of curable STIs has been effective in reducing STIs and has been recommended for use among FSW by the World Health Organization [21, 22]. The government should consider adoption and implementation of STI periodic presumptive treatment strategies for all FSWs.

The high HIV prevalence and incidence establish the need for a combination HIV prevention approach that addresses behavioral, structural, and biological drivers of HIV infection [24, 25]. Earlier ART initiation [Treat-All] for HIV-positive individuals can improve clinical outcomes and reduce transmission of HIV to their HIV-negative sexual partners, including clients [26,27,28,29,30,31]. It is recommended that Treat-All and pre-exposure prophylaxis or PrEP [a daily anti-retroviral pill taken by HIV negative persons to prevent HIV infection prior to exposure] should be part of a combination prevention package that includes behavior change, condom promotion, and should be implemented in conjunction with strategies that create an enabling environment and increase access to FSW-friendly services. It is estimated that voluntary access to PrEP for FSWs could further reduce HIV incidence among this population and their clients by 40% [25]. As the introduction of PrEP is fairly new to FSWs, community engagement and implementation science research will be needed [32, 33].

The findings that only over a half of FSWs had ever tested for HIV and one-third used condom at last sex with paying and non-paying clients are concerning. The levels of HIV testing and condom use in this study are much lower than levels of HIV testing and condom use found in the regions [34,35,36,37]. This is worrisome because there have been significant intervention efforts targeting FSWs in Tanzania over the past decade. In particular, there are health promotion programs aiming to increase condom use and uptake of HIV testing in Tanzania using multiple approaches, including peer education and community outreach, creating drop-in-centers to serve as safe space for health education and HIV and STI testing and treatment, mobile testing, and provider-initiated testing. Condoms have been distributed free of charge to FSWs for more than a decade through major social marketing initiatives supported by several global health organizations. These programs should undergo rigorous evaluation to document lessons learned and to inform future programming. It is noteworthy that many FSWs in this study perceived low risk of being infected with HIV that could have led to the low testing rate. Lack of awareness of HIV status would potentially escalate HIV transmission to the general population, especially when condom use is suboptimal and only half of FSWs living with HIV are estimated to be on ART [38]. This calls for strategies to scale up rapid HIV testing, implement community-based testing and self-testing, and improve FSWs’ perceptions of HIV risk, increase the visibility of HIV testing sites, and destigmatize HIV testing [10, 25, 39]. These strategies will be critical in achieving the WHO/UNAIDS 90-90-90 global goal (90% of people with HIV diagnosed, 90% of them on ART, and 90% of them virally suppressed) by 2020 [40].

Our data also show that violence against FSWs is highly prevalent. Violence undermines HIV prevention efforts and increases the vulnerability of FSWs to HIV transmission [41, 42]. In particular, sexual and physical violence or the threat of it could result in FSWs giving higher priority to their safety than to preventing HIV infection or transmission. Interventions including sex-worker education on rights, and community mobilization to respond to violence and policy advocacy to promote human rights of sex workers are critical [10, 43]. Community-empowerment interventions have shown to be effective and can serve as platforms for health education, demand-creation, income-generating activities, and for starting ART and other HIV prevention technologies [44, 45]. Currently, sex work is illegal in Tanzania and police harassment of FSWs has been documented and is considered a significant barrier to HIV programs for FSWs and other key populations [42]. Research suggests that decriminalization of sex work would help reduce 33 to 46% of new HIV infections over the next decade [46]. Therefore, it is important to recognize sex work as an occupation, that many people will continue to sell sex, and that reduction of HIV transmissions associated with sex work should be considered a key component in the achievement of universal HIV prevention, care and treatment services.

Limitations

First, self-reporting of sensitive indicators such as condom use, HIV testing, and HIV status, are susceptible to social desirability and recall biases. Second, we estimated HIV incidence based on data of a cross-sectional survey, comparing self-reporting HIV status with a biological HIV test. However, we believe that bias in the self-reporting of HIV status in this survey is minimal. This is because we stressed the importance of confidentiality and anonymity for participants during the interview; all interviewers are trained on maintaining confidentiality and anonymity, as well as how to ask the questions pertaining to individual’s HIV status disclosure. In fact, only 0.5% (9 FSWs) refused to respond to the questions asking about HIV status and data showed only three discrepancies. While asking participants to self-report HIV status has raised concerns, we believe it can be done effectively and ethically and that the benefits outweigh potential risks. The data allow us to better understand the dynamic of the HIV epidemics that can lead to better responses. Third, we did not have reliable regional FSW population size estimates to further adjust for in the multivariate, pooled analysis. However, we adjusted for region to account for regional variation, and for recruitment pattern and network size using RDS weights.

Conclusions

FSWs in Tanzania bear an unacceptably high burden of HIV and STIs compared to the general population, and their HIV risks and vulnerability vary significantly by types of sex work, region, and age. More alarmingly, the epidemics continue to expand, with high HIV incidence, regardless of many HIV prevention initiatives and significant financial assistance from donors. Effectiveness and cost-effectiveness evaluation of ongoing interventions for FSWs in Tanzania are important to ensure that resources are allocated where they are most needed and that investments produce optimal results. The high HIV prevalence and incidence coupled with suboptimal condom use indicate the need to follow the WHO’s Treat-All Guidelines [26] and start ART to all FSWs living with HIV regardless of their CD4 count. Innovative efforts to increase testing uptake and to identify HIV infection, including self-testing and community-based testing, should be introduced or scaled up [39]. In addition, ARV-based prevention technologies such as oral PrEP and microbicides should be piloted and evaluated. Finally, decriminalizing sex work and promoting human rights of FSWs and ensuring their safety must be made part of all HIV prevention efforts in Tanzania.