HIV epidemiology among female sex workers and their clients in the Middle East and North Africa: systematic review, meta-analyses, and meta-regressions

Background HIV epidemiology among female sex workers (FSWs) and their clients in the Middle East and North Africa (MENA) region is poorly understood. We addressed this gap through a comprehensive epidemiological assessment. Methods A systematic review of population size estimation and HIV prevalence studies was conducted and reported following PRISMA guidelines. Risk of bias (ROB) assessments were conducted for all included studies using various quality domains, as informed by Cochrane Collaboration guidelines. The pooled mean HIV prevalence was estimated using random-effects meta-analyses. Sources of heterogeneity and temporal trends were identified through meta-regressions. Results We identified 270 size estimation studies in FSWs and 42 in clients, and 485 HIV prevalence studies in 287,719 FSWs and 69 in 29,531 clients/proxy populations. Most studies had low ROB in multiple quality domains. The median proportion of reproductive-age women reporting current/recent sex work was 0.6% (range = 0.2–2.4%) and of men reporting currently/recently buying sex was 5.7% (range = 0.3–13.8%). HIV prevalence ranged from 0 to 70% in FSWs (median = 0.1%) and 0–34.6% in clients (median = 0.4%). The regional pooled mean HIV prevalence was 1.4% (95% CI = 1.1–1.8%) in FSWs and 0.4% (95% CI = 0.1–0.7%) in clients. Country-specific pooled prevalence was < 1% in most countries, 1–5% in North Africa and Somalia, 17.3% in South Sudan, and 17.9% in Djibouti. Meta-regressions identified strong subregional variations in prevalence. Compared to Eastern MENA, the adjusted odds ratios (AORs) ranged from 0.2 (95% CI = 0.1–0.4) in the Fertile Crescent to 45.4 (95% CI = 24.7–83.7) in the Horn of Africa. There was strong evidence for increasing prevalence post-2003; the odds increased by 15% per year (AOR = 1.15, 95% CI = 1.09–1.21). There was also a large variability in sexual and injecting risk behaviors among FSWs within and across countries. Levels of HIV testing among FSWs were generally low. The median fraction of FSWs that tested for HIV in the past 12 months was 12.1% (range = 0.9–38.0%). Conclusions HIV epidemics among FSWs are emerging in MENA, and some have reached stable endemic levels, although still some countries have limited epidemic dynamics. The epidemic has been growing for over a decade, with strong regionalization and heterogeneity. HIV testing levels were far below the service coverage target of “UNAIDS 2016–2021 Strategy.” Electronic supplementary material The online version of this article (10.1186/s12916-019-1349-y) contains supplementary material, which is available to authorized users.


Background
The Middle East and North Africa (MENA) is one of only two regions where HIV incidence and AIDS-related mortality are rising [1]. Between 2000 and 2015, the increase in the number of new infections was estimated at over a third, while that of AIDS-related deaths, at over threefold [1][2][3]. MENA has been described as "a real hole in terms of HIV/AIDS epidemiological data" [4], with unknown status and scale of epidemics in multiple countries [5][6][7].
Despite recent progress in HIV research and surveillance in MENA [8], including the conduct of integrated biobehavioral surveillance surveys (IBBSS) [5,9], many of these data are, at best, published in country-level reports, or never analyzed. Since 2007, the "MENA HIV/AIDS Epidemiology Synthesis Project" has maintained an active regional HIV database [6]. The first systematic syntheses of HIV data documented concentrated and emerging epidemics among men who have sex with men (MSM) [10] and people who inject drugs (PWID) [11]. The majority of these epidemics emerged within the last two decades [10,11].
Although the size of commercial heterosexual sex networks is expected to be much larger than the risk networks of MSM and PWID [6,7], estimates for the population proportion of female sex workers (FSWs), volume of clients they serve, and geographic and temporal trends in infection remain to be established. This evidence gap was highlighted in the latest gap report by the Joint United Nations Programme on HIV/AIDS (UNAIDS) [3], indicating "a lack of data on the burden of HIV among sex workers in the region" and stressing that "the epidemic among them is poorly understood" though "HIV in every country is expected to disproportionately affect sex workers" [3].
This study characterizes HIV epidemiology among FSWs and their clients in MENA by (1) systematically reviewing and synthesizing all available published and unpublished records documenting population size estimates, population proportions, HIV incidence, and HIV prevalence (including in proxy populations of clients such as male sexually transmitted infection (STI) clinic attendees); (2) estimating, for each population, the pooled mean HIV prevalence per country and regionally; (3) identifying the regional-level associations with prevalence, sources of heterogeneity, and temporal trends; and (4) synthesizing the key measures of sexual and injecting risk behaviors.

Search strategy and selection criteria
Evidence for population size estimate, population proportion, HIV incidence, and HIV prevalence in FSWs and clients was systematically reviewed as per Cochrane's Collaboration guidelines [12]. Findings were reported following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [13] (checklist in Additional file 1: Table S1). MENA definition here includes 23 countries extending from Pakistan to Morocco (Additional file 1: Figure S1), based on the convention in HIV research [6,7,10,11] and on World Health Organization (WHO), UNAIDS, and World Bank definitions [6]. MENA was also classified by subregion comprising Eastern MENA (Afghanistan, Iran, Pakistan), the Fertile Crescent (Egypt, Iraq, Jordan, Lebanon, Palestine, Syria), the Gulf (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, United Arab Emirates, Yemen), the Horn of Africa (Djibouti, Somalia, recently independent South Sudan), and North Africa (Algeria, Libya, Morocco, Sudan, Tunisia).
Systematic searches were performed, up to July 29, 2018, on ten international-, regional-, and country-level databases; abstract archives of International AIDS Society conferences [14]; and Synthesis Project database which includes country-level and international organizations' reports and routine data reporting [6,7] (Additional file 1: Box S1). No language or year restrictions were used.
Titles and abstracts of unique citations were screened for relevance, and full texts of relevant/potentially relevant citations were retrieved for further screening. Any document/report including outcomes of interest based on primary data was eligible for inclusion. Case reports, case series, editorials, commentaries, and studies in populations (such as "vulnerable women") where overlap with FSWs is implied but engagement in sex work is not explicitly indicated were excluded. Reference lists of reviews and all relevant documents were hand searched for eligible reports.
In this article, the term study refers to a specific outcome measure (population size estimate, incidence, or prevalence) in a specific population. Therefore, one report could contribute multiple studies, and one study could be published in different reports. Duplicate study results were included only once using the more detailed report.

Data extraction and synthesis
Data extraction was performed by HC and double extraction by MH, with discrepancies settled by consensus or by contacting authors. Data were extracted from full texts by native speakers (extraction list in Additional file 1: Box S2).
Population size estimates and population proportions were grouped based on being of national coverage or for specific subnational settings, and distinguishing between current FSWs/clients and history of sex work/ex-client. For FSWs, population proportion is defined as the proportion of all reproductive-age women that are engaged in sex work, that is the exchange of sex for money (sex work as a profession) [15,16], and for clients, as the proportion of men buying sex from FSWs using money. Studies with mixed or non-representative samples (samples biased towards oversampling FSWs with no estimate adjustment) were excluded.
Due to the paucity of studies directly looking at HIV prevalence in clients of FSW, HIV prevalence studies in male STI clinic attendees, or mixed-sex samples of predominantly men (> 60%), were used as a proxy for HIV prevalence in clients of FSWs [17,18].
HIV incidence studies were identified and reported. Additional contextual information was extracted from FSW studies included in the review. These include age, age at sexual debut, age at sex work initiation, sex work duration, marital status, and HIV/AIDS knowledge and perception of risk, as well as behavioral measures of condom use, injecting drug use, sexual partnerships, and HIV testing.
Data were summarized using medians and ranges.

Quality assessment
Risk of bias (ROB) assessments for population size estimates/population proportions and for HIV prevalence were conducted as informed by Cochrane Collaboration guidelines [12] (criteria in Additional file 1: Table S2).
Briefly, size estimation studies were classified as having "low" versus "high" ROB on each of the three domains assessing the (1) validity of sex work definition/engagement in paid sex (clear/valid definition; otherwise), (2) rigor of estimation methodology (likely-to-yield representative estimate; otherwise), and (3) response rate (≥ 60%; < 60%).
Prevalence studies were similarly classified on each of the four domains assessing the (1) validity of sex work definition/engagement in paid sex (clear/valid definition; otherwise), (2) rigor of sampling methodology (probability-based; non-probability-based), (3) response rate (≥ 60% or ≥ 60% of target sample size reached for studies using respondentdriven or time-location sampling; < 60%), and (4) type of HIV ascertainment (biological assays; self-report).
Studies with missing information for a specific domain were classified as having "unclear" ROB for that domain. Measures only extracted from routine databases were considered of unknown quality, as original reports were not available for assessing ROB, and were not included in the quality assessment. The impact of quality domains on observed prevalence was examined in meta-regression (described below).

Meta-analyses
Pooled mean HIV prevalence in FSWs and client populations were estimated using random-effects meta-analyses, by country and for the whole region. Variances were stabilized using Freeman-Tukey-type arcsine square-root transformation [22,23]. Weighting was performed using the inverse-variance method [23,24]. Pooling was performed using Dersimonian-Laird random-effects models to allow for sampling variation and true heterogeneity [25,26]. Overall prevalence measures were replaced by their stratified measures where applicable.
Heterogeneity was assessed using Cochran's Q statistic to confirm the existence of heterogeneity, I 2 to estimate the magnitude of between-study variation, and prediction intervals to estimate the 95% interval of distribution of true effect sizes [26,27].

Meta-regression analyses
Random-effects meta-regression analyses were conducted to identify the regional-level associations with HIV prevalence in FSWs, sources of between-study heterogeneity, and temporal trend. Independent variables considered a priori were country/subregion, FSW population type, sample size, median year of data collection, sampling methodology, response rate, validity of sex work definition, and HIV ascertainment (details in Additional file 1: Table S3). The same factors (as applicable) were considered for clients' meta-regression analyses.
To avoid the exclusion of studies with zero prevalence, an increment of 0.1 was added to the number of events in all studies to calculate the log-transformed odds, that is prevalence/(1 − prevalence), and corresponding variance [29]. Factors showing strong evidence for an association with the odds (p value ≤ 0.10) in univariable analysis were included in the multivariable analysis.

Results
Search results and scope of evidence Figure 1 shows the study selection process. Population size estimates and population proportions of FSWs and clients Table 1 and Additional file 1: Table S4 show the population size estimate and population proportion studies for FSWs and clients at the national and subnational levels, respectively. At the national level, the median number of current/recent FSWs (engaged in sex work in the past year) was 58,934 (range = 2218 in Djibouti to 167,501 in Pakistan), and the median population proportion (out of reproductive-age women aged 15-49 years) was 0.6% (range across studies = 0.2% in Egypt to 2.4% in Iran). The median population proportion of current/recent clients (buying sex from FSWs in the past year) based on diverse samples of general population men was 5.7% (range across studies = 0.3% in Sudan to 13.8% in Lebanon).
With high heterogeneity in estimation methodology, time frame, and scope between and within countries, it was deemed not meaningful to generate country-specific or regional-pooled estimates for the size/population proportions.

HIV incidence overview
There were six incidence studies among FSWs (three from each of Somalia and Djibouti; data not shown). Three studies reported zero seroconversions [51,52].  1 Flow chart of the study selection process in the systematic review following PRISMA guidelines [13]  One study from Somalia reported a cumulative incidence of 2.6% after 6 months of follow-up [51]. The other two from Djibouti-among predominantly Ethiopian FSWs (91%)-reported a cumulative incidence of 3.4% [51] and 11.6% [51] after 3 and 9 months of followup, respectively. All incidence studies were conducted before the year 2000 and were limited in scale and scope.

Quality assessment
Additional file 1: Tables S6-S9 show the summarized and study-specific quality assessments for the size estimation and HIV prevalence studies in FSWs and clients. Almost all size estimation studies used clear/valid sex work definitions, and > 70% used rigorous size estimation methodologies. Similarly, > 70% of prevalence studies in FSWs used clear/valid sex work definitions and probability-based sampling for participants' recruitment. Meanwhile, > 85% of prevalence studies in clients used convenience sampling.
Overall, studies were of reasonable quality. The majority of size estimation studies in FSWs and clients had low ROB on ≥ 2 quality domains (94.4% and 82.1%, respectively), and none had high ROB on ≥ 2 domains. Similarly, 85.0% of prevalence studies in FSWs and 39.4% of studies in clients had low ROB on ≥ 2 domains (studies among STI clinic attendees mostly used convenience sampling, and few reported on contact with FSWs), while 0.7% and 6.1% had high ROB on ≥ 2 domains, respectively.

Pooled mean HIV prevalence
The pooled mean HIV prevalence for the MENA region was 1.4% (95% confidence interval (CI) = 1.1-1.8%) in FSWs and 0.4% (95% CI = 0.1-0.7%) in clients (Table 5). A difference was observed between the median prevalence and the pooled mean prevalence due to the high clustering of prevalence measures close to zero.
There was evidence for the heterogeneity in effect size (prevalence) in meta-analyses. p value for Cochran's Q statistic was mostly < 0.0001, prediction intervals were wide, and I 2 was often > 50% indicating that most between-study variability is due to the true differences in prevalence across studies rather than chance.
Associations with prevalence, sources of between-study heterogeneity, and temporal trend Univariable meta-regressions for FSWs demonstrated strong evidence for an association with odds for subregion, population type, sample size, year of data collection, and response rate (Table 6). Meanwhile, there was poor evidence for an association with sampling methodology, validity of sex work definition, and HIV ascertainment, which were hence dismissed from inclusion in the multivariable model. Most variability in odds was explained by subregion (adjusted R 2 = 39.8%).
Compared with studies with data collection pre-1993, studies conducted after 2003 showed strong evidence for higher odds (AOR = 2.0, 95% CI = 1.2-3.3). Notably, the trend of increasing odds was evident only after controlling for the strong confounding effect of the subregion. The trend for each subregion was also overall increasing, though the strength of evidence varied across subregions              Abbreviations: CI confidence interval, FSWs female sex workers *These medians and ranges are calculated on the stratified HIV prevalence measures **Missing sample sizes for measures (or their strata) were imputed using median sample size calculated from studies with available information. Analyses excluding these studies had no impact on study findings † Q-the Cochran's Q statistic is a measure assessing the existence of heterogeneity in effect size (here, HIV prevalence) across studies ‡ I 2 -a measure assessing the magnitude of between-study variation that is due to the differences in effect size (here, HIV prevalence) across studies rather than chance £ Prediction interval-a measure estimating the 95% interval of the distribution of true effect sizes (here, HIV prevalence) Based on results of meta-analyses for FSWs, countries were classified as having low-level HIV epidemic (prevalence < 1%), intermediate-intensity HIV epidemic (prevalence 1-5%), and concentrated HIV epidemic (prevalence > 5%) ¥ Point estimate as only one study was available € Before 2011, South Sudan was part of Sudan, and thus, earlier measures from Sudan were based on studies that may have included participants from both Sudan and South Sudan  added to a number of events in all studies when generating log odds, and Eastern MENA was thus used also as a statistically better reference. While this choice of increment was arbitrary, other increments yielded the same findings, though some of the effect sizes changed in scale **Year grouping was driven by independent evidence identifying the emergence of HIV epidemics among both men who have sex with men [10] and people who inject drugs [11] in multiple MENA countries around 2003. Missing values for year of data collection (only six stratified measures) were imputed using data for year of publication adjusted by the median difference between year of publication and median year of data collection (for studies with complete information) † A large fraction of studies did not separate the different forms of female sex workers, and thus it was not possible to analyze these as separate categories ‡ Measures extracted only from routine databases with no reports describing the study methodology were not included in the ROB assessment Meta-regression analyses for clients demonstrated similar results to those of FSWs, but with wider CIs considering the smaller number of prevalence studies (Additional file 1: Table S10). There was evidence that subregion was associated with HIV odds in clients, but no evidence that sample size or year of data collection explained the between-study heterogeneity.

Sex work context and sexual and injecting risk behaviors
For the detailed sex work context and behavioral measures, we provide here (for brevity) only a high-level summary of key measures.

Reported condom use
There was high heterogeneity in reported condom use among FSWs by sexual partnership type and across and within countries (Additional file 1: Table S11). Condom use at last sex with clients ranged from 1.2 to 94.8% (median = 44.0%). Consistent condom use with clients ranged from 0 to 95.2% (median = 26.3%) among all FSWs and from 38.2 to 45.3% (median = 42.3%) among FSWs reporting condom use with clients.
The median fraction of FSWs who reported having a condom at the time of study interview was 12.5% (range = 0-66.1%).

Clients and partners
Studies varied immensely in types of measures reporting data on clients and partners. Some reported a mean number of regular/non-regular clients, but over various time frames. Others reported different distributions for the number of clients (and by client type), also over various time frames. Summarizing the evidence was therefore challenging, given the large type of measure variability.
This being said, the mean number of clients in the past month ranged from 4.4 to 114.0, with a median of 34.0 clients. Median fraction of FSWs reporting (during the past month) < 5 clients, 5-9 clients, and 10+ clients was 28.5%, 28.1%, and 19.1%, respectively. FSWs were equally likely to report regular and one-time clients during the past month (medians = 80.0% and 81.0%, ranges = 54.3-92.4% and 59.2-97.5%, respectively).
FSWs reported a distribution of sex acts in the past week, with a median of 41.2% reporting 1-2 acts, 32.0% reporting 3-4 acts, and 12.9% reporting 5+ acts. Anal sex with clients in the past month was reported by a median of 8.0% (range = 2.3-100%).
Median fraction of FSWs that are married/cohabiting was 45.3% (range = 0-99.6%), while that of FSWs reporting non-paying partners was 48.5% (range = 6.8-86.2%). The mean number of non-paying partners in the past month ranged between 1 and 3, with about two thirds reporting only one partner.
Only few studies investigated group sex: 7.7% [90] of FSWs reported ever engaging in group sex, 6.2% [68] and 12.9% [68] reported group sex in the past month, and 10.0% [58] in the past week.

Injecting risk behavior, sex with PWID, and substance use
There was a large variability in injecting risk behavior and substance use among FSWs, but the highest levels of injecting drug use were reported in Iran and Pakistan (Additional file 1: Table S12). Median of current/recent injecting drug use was 2.1% (range = 0-26.6%), but the majority of studies were from Pakistan. Studies in Iran reported a history of injecting drug use in the range of 6.1-18.0% (median of 13.6%) among all FSWs and range of 16.4-25.5% (median of 22.3%) among only ever/active drug users. A history of injecting drug use was reported by < 1% (median) of all FSWs (range = 0%-11.8%) in the rest of MENA countries.
Fraction of FSWs reporting current/recent sex with PWID ranged from 0.5 to 13.6% within Afghanistan and 0-54.9% within Pakistan, with medians of 5.2% and 5.6%, respectively. Sex with PWID was reported at 23.6% [93] among FSWs in Iran.

Knowledge of HIV/AIDS and perception of risk
Knowledge of HIV/AIDS was generally high among FSWs across MENA (Additional file 1: Table S13). Vast majority of FSWs ever heard of HIV (median = 81.9%, range = 25.4-100%) and were aware of sexual (median = 72.0%, range = 50.8-94.9%) and injecting (median = 88.7%, range = 11.5-99.6%) modes of transmission, but to a lesser extent of condoms as a prevention method (median = 51.6%, range = 14.1-89.8%)-condoms were more perceived as a contraception method. Levels of knowledge, however, varied often substantially within the same country.

HIV testing
HIV testing among FSWs varied across countries, but was generally low, with a median fraction of 17.6% (range = 4.0-99.4%) ever tested for HIV (Additional file 1: Table S15). Only a median of 12.1% (range = 0.9-38.0%) of all FSWs tested for HIV in the past 12 months, and nearly two thirds of those who ever tested did so in the past 12 months (median = 59.2%, range = 33.3-82.0%).

Discussion
Through an extensive, systematic, and comprehensive assessment of HIV epidemiology among FSWs and clients, including data presented in the scientific literature for the first time, we found that HIV epidemics among FSWs have already emerged in MENA, and some appear to have reached their peak. Based on a synthesis and triangulation of evidence from studies on a total of 300,000 FSWs and 30,000 clients, a strong regionalization of epidemics has been identified. In Djibouti and South Sudan, the HIV epidemic is concentrated with a prevalence of~20% in FSWs. In Algeria, Libya, Morocco, Somalia, and Sudan, the epidemic is of intermediate-intensity (prevalence 1-5%). Strikingly, in the remaining countries with available data, the prevalence is < 1%, and most often zero.
A key finding is that HIV prevalence in FSWs has been (overall) growing steadily since 2003. This is the same time in which independent evidence has identified the emergence of major epidemics among both PWID [11] and MSM [10] in MENA. It is probable that the epidemics among these key populations have been bridged to FSWs. An example is Pakistan, where the prevalence among FSWs was < 1% in almost all cities in three consecutive IBBSS rounds between 2005 and 2012 [38,40,69]. However, prevalence ranging from 1.5 to 8.8% was documented in half of the cities in the latest round in 2016-2017 [42]. These emerging epidemics among FSWs were preceded by large and growing epidemics first among PWID [11] and then among MSM [10,11].
Some of the FSW epidemics, particularly those in Djibouti and South Sudan, emerged much earlier, most likely by late 1980s [6], mainly affected by geographic proximity and stronger population links to sub-Saharan Africa (SSA) [6]. Djibouti is a port country and the major trade route for Ethiopia and a station for large international military bases [6,151]. The majority of FSWs operating in Djibouti are Ethiopians catering to the Ethiopian truck drivers transporting shipments from the Djibouti port [84][85][86]. South Sudan is socioculturally part of SSA, with a major fraction of FSWs coming from Uganda, Congo, and Kenya [79]. In these MENA countries, HIV in commercial heterosexual sex networks (CHSNs) is well-established and epidemics are concentrated-though at levels lower than the hyperendemic epidemics observed in SSA [152].
Unlike the epidemics among PWID and MSM [10,11], the FSW epidemics have been overall growing rather slowly, with the prevalence being mostly < 5%. Strikingly, a considerable fraction of countries still do not appear to have much HIV transmission in CHSNs, with consistently very low prevalence, quite often even at zero level-46.8% of studies in FSWs reported zero prevalence, and 7 out of 18 countries had a pooled mean prevalence of zero or nearly zero. One explanation for the observed low HIV prevalence could be that HIV has not yet been effectively introduced into CHSNs-it took decades for HIV to be effectively introduced into PWID [11] and MSM [10] networks. Another possible factor pertains to the structure of CHSNs, characterized apparently by low connectivity [6,153,154], which reduces the risk of HIV being introduced, or efficiently/sustainably transmitted. Unlike PWID and MSM, FSWs are also exposed to HIV mainly through their clients, who have a lower risk of exposure to HIV than themselves, thus possibly contributing to slower epidemic growth [6].
Other factors may also contribute to explaining the observed low HIV prevalence. The synthesized evidence suggests a lower risk environment for FSWs in MENA, compared to other regions. The reported number of clients is rather low at a median of 34 per month, at the lower end of global range [155][156][157][158]. Close to half of commercial sex acts are protected through condom use, with no difference between regular and one-time clients, despite noted variability across and within countries. HIV/AIDS knowledge also varies, but is generally substantial, with the majority of FSWs being aware of sexual and injecting modes of transmission, and over half are aware of condoms as a prevention method. Injecting drug use and sex with PWID is low in most countries, except for countries in Eastern MENA, notably Afghanistan, Iran, and Pakistan. Serological markers for hepatitis C virus (a marker of injecting risk) [159][160][161] are also low in FSWs, assessed at a median of 1.1% (range = 0-9.9%, not shown), with the highest measures reported in Iran [61,162]. These relatively lower levels of risk behavior than other regions [163][164][165] stand in contrast to what has been observed in PWID and MSM in MENA [10,11].
Importantly, with the efficacy of 60% in randomized clinical trials [166][167][168][169], male circumcision, which is essentially at universal coverage across MENA [170], may have also slowed, or even substantially reduced HIV transmission in CHSNs leading to the observed low HIV prevalence [171]. Incidentally, the two most affected countries-South Sudan and Djibouti-are nearly the only two major settings where male circumcision is at low coverage in MENA, either nationally, as is the case for South Sudan [170], or among clients of FSWs, as is the case for Ethiopian truckers and international military personnel stationed in Djibouti [151,170]. Though HIV prevalence will probably continue to increase among FSWs and clients, the high levels of male circumcision coupled with lower levels of risk behavior may prevent significant epidemics, as seen elsewhere [172][173][174], from materializing in CHSNs in multiple MENA countries.
HIV prevalence in FSWs in few countries, particularly in Eastern MENA, may not necessarily reflect heterosexual as much as iatrogenic exposures through injecting drug use. Specifically, in Iran and Pakistan, countries with large HIV epidemics among PWID [11], a considerable fraction of FSWs report current/recent/history (14% in Iran and 2% in Pakistan) of injecting drug use. High prevalence of sex work is also reported in women engaging in injecting drug use [93,175,176]. Current/recent/history of sex with PWID is also common (24% in Iran and 6% in Pakistan). The overlap between these key populations suggests a potential for HIV to be bridged from PWID networks to CHSNs, as seem to have occurred in Pakistan recently [42,177,178].
Population proportion of current/recent FSWs ranged from 0.2 to 2.4% across studies with a median of 0.6%, while that of current/recent clients ranged from 0.3 to 13.8% with a median of 5.7%, both on the lower end of global range [179,180]. Though these population proportions may seem small, the size of CHSNs is much larger than that of PWID and MSM [10,11,181]. This suggests that CHSNs could be a main driver of HIV incidence in many countries despite the low HIV prevalence in FSWs. An example is Morocco where the mode of transmission analyses estimated that over half of HIV incidence is driven by CHSNs, despite an HIV prevalence of only~2% in FSWs [182][183][184]. The role of CHSNs is even more significant in countries with concentrated epidemics. In Djibouti, for example, the large HIV epidemic among FSWs was mirrored shortly after by a rapid rise in prevalence among clients (as proxied by male STI clinic attendees; Table 4), leading eventually to a prevalence > 1% in pregnant women [6].
HIV response to the epidemic in CHSNs in MENA continues to be weak and limited in scope and scale [185]. Criminality [151,185] and stigma [186][187][188] associated with sex work persist as barriers to surveillance and targeted programming [189][190][191], leading even to the resistance to acknowledge the existence of sex work [192]. These challenges are compounded by the diverse typologies and increased mobility of FSWs [41,70,151]. Across MENA, only 18% of FSWs reported ever testing for HIV, and fewer (12%) reported testing in the past 12 months, far below the 90% service coverage target of "UNAIDS 2016-2021 Strategy" [193]. Programs, including healthcare provision, where they exist, are nearly always implemented by non-governmental organizations (NGOs), who often lack the resources or legal coverage to deliver comprehensive prevention interventions [6,185].
There are, however, notable exceptions. Morocco has established an evidence-informed national strategy and rapidly scaled up provision of comprehensive services for at-risk populations, including outreach peer education programs as well as testing and case management services [183,185]. Voluntary counseling and testing centers were established nationwide, with FSWs estimated to constitute about a quarter of attendees in 2007 [183,194]. Findings of the 2011-2012 IBBSS indicated that over a third of FSWs ever tested for HIV, the vast majority of whom were aware of their status [67]. Condom use at last sex also increased from 37% in 2003 to a median of 50% in 2011 (Additional file 1: Table S11). Morocco's success has been grounded on a strong multisectorial response where NGOs, in partnership with the government, play a leading role in implementing interventions [185]. In Iran, the large expansion of harm reduction services, including the first women-operated services in MENA [11], is a promising step for targeting FSWs most at risk. This study is limited by gaps in evidence. Epidemic status among FSWs remains unknown in six countries, as no data were identified. Others (Bahrain and Libya) also had limited data to warrant a meaningful characterization of the epidemic. The high heterogeneity of epidemics within countries suggests that caution is needed when interpreting data without a representative national coverage. For instance, while concentrated epidemics among FSWs are documented in southern Morocco [67,195] and southern Algeria [113,[196][197][198], these do not appear to be representative of FSWs at the national level [42,67,74,78,81,82,113,[195][196][197][198][199]. Hidden epidemics or outbreaks may also exist in specific geographies within the country, but not necessarily elsewhere. Data varied over time with high quality and volume of evidence available mostly post-2000, thanks to the expansion and funding of IBBSS studies. While the pooled prevalence estimates were meant to provide a summary of the relative standing of MENA countries in the HIV epidemic, the large between-study heterogeneity suggests that caution is warranted when interpreting these estimates. Studies in clients of FSWs/proxy populations remain limited with wide variability in evidence availability across MENA.
A considerable fraction of studies used convenience sampling, although meta-regression indicated no difference in the prevalence by sampling methodology. This may be explained by FSWs being more "visible" [151,200] compared to PWID [11] and MSM [10]. A sizable fraction of studies was from routine data reporting with no sufficient documentation of study methodology. However, most of these country-level program data were presumably based on rigorous case definitions following WHO guidelines [6]. There is also a possibility that a fraction of studies may have enrolled women without a strict and valid definition for sex work, yet metaregression findings showed no effect for the validity of sex work definition on HIV prevalence. There was also no evidence that other study-specific quality domains, including HIV ascertainment method and response rate, had an effect on prevalence. A considerable fraction of studies reported zero prevalence, thus an increment of 0.1 was added to a number of events to be able to conduct the meta-regressions. While this choice of increment was arbitrary, other increments yielded the same findings, though some of the effect sizes changed in scale. There was evidence for a small-study effect in meta-regression suggesting potential publication bias towards studies reporting higher prevalence.

Conclusions
HIV epidemics among FSWs are emerging in MENA, with some already established. The epidemic has been growing steadily in recent years, with strong regionalization and heterogeneity. A contributing factor to epidemic growth appears to be the epidemics that emerged among PWID [11] and MSM [10] nearly two decades ago. Strikingly, a large fraction of countries still do not appear to have any significant epidemic dynamics in CHSNs. These findings demonstrate the need for expanding surveillance systems, including the conduct of repeated IBBSS studies with national coverage to monitor HIV prevalence trends and to detect the emergence of epidemics. There is also a pressing need for mapping and size estimation studies to delineate the diverse typologies of sex work and to ensure evidence-informed response with adequate coverage of interventions.
Achieving "UNAIDS 2016-2021 Strategy" [193] service coverage targets entails reaching out to the increasingly dispersed FSW population [41,70,151]. Building on Morocco's success, this would be best achieved through NGOs leading the provision of comprehensive interventions, with governmental support, even if discrete. Extending harm reduction services to women PWID is also critical to curb HIV burden in FSWs most at risk, specifically in Eastern MENA. The window of opportunity for detecting epidemics at their nascence, and for controlling incidence in CHSNs, should not be missed.

Additional file
Additional file 1: Supplementary information including further details and additional results for the systematic review and meta-analytics of HIV infection in female sex and their clients workers in the Middle East and North Africa. Tables S1-S15. Figure S1. Box S1-S2.