Key population size, HIV prevalence, and ART coverage in sub-Saharan Africa: systematic collation and synthesis of survey data ============================================================================================================================== * Oliver Stevens * Keith Sabin * Sonia Arias Garcia * Rebecca Anderson * Kalai Willis * Abu Abdul-Quader * Anne McIntyre * Elizabeth Fearon * Emilie Grard * Alice Stewart-Brown * Frances Cowan * Louisa Degenhardt * Jinkou Zhao * Avi Hakim * Katherine Rucinski * Isabel Sathane * Makini Boothe * Lydia Atuhuire * Peter Nyasulu * Lucy Platt * Brian Rice * Wolfgang Hladik * Stefan Baral * Mary Mahy * Jeffrey W. Eaton ## Abstract **Background** HIV programmes in sub-Saharan Africa (SSA) require information about HIV among key populations to ensure equitable and equal access to HIV prevention and treatment. Surveillance has been conducted among female sex workers (FSW), men who have sex with men (MSM), people who inject drugs (PWID), and transgender populations, but is not systematically included in national HIV estimates. We consolidated existing KP surveys to create national-level estimates of key population size, HIV prevalence, and ART coverage for mainland SSA. **Methods** Key population size estimates (KPSE), HIV prevalence, and ART coverage data from 38 countries from 2010-2021 were collated from existing databases, deduplicated, and verified against primary sources. We used Bayesian mixed-effects regression to spatially smooth KPSE, and regressed subnational key population HIV prevalence and ART coverage against age/sex/year/province-matched total population estimates. **Findings** We extracted 1449 unique KPSE datapoints, 1181 HIV prevalence datapoints, and 242 ART coverage datapoints. Countries had data for a median of five of the twelve population/outcome stratifications. Across countries, a median of 1.44% of urban women were FSW (interquartile range [IQR] 0.83-1.89%); 0.60% of urban men were MSM; and 0.16% of urban adults injected drugs (IQR 0.14-0.24%). HIV prevalence in all key populations was higher than matched total population prevalence. ART coverage was correlated with, but lower than, total population ART coverage. Across SSA, key populations were estimated as 1.1% (95%CI 0.7-1.9%) of the population but 5.1% (95%CI 3.2-10.3%) of all PLHIV aged 15-49 years. **Interpretation** Key populations in sub-Saharan experience disproportionate HIV burden and somewhat lower ART coverage, underscoring need for focused prevention and treatment services. However, large heterogeneity and incomplete data availability limit precise estimates for programming and monitoring trends. Future efforts should focus on integrating and strengthening key population surveys and routine data within national HIV strategic information systems. **Funding** UNAIDS **Evidence before this study** Key populations (KPs), including female sex workers (FSW), men who have sex with men (MSM), people who inject drugs (PWID), and transgender women (TGW), are disproportionately vulnerable to HIV in sub-Saharan Africa (SSA). Delivering appropriate HIV prevention and treatment programming for these populations, and monitoring attainment of an equitable HIV response, requires robust information on key population size, HIV prevalence, the treatment cascade, and new HIV infections. For this reason, key population surveys including population size estimation and bio-behavioural survey are a standard component of a comprehensive national HIV surveillance portfolio, Several organisations have initiatives to consolidate HIV data about key populations for global monitoring, programme planning, and research purposes. These include the Joint United Nations Programme on HIV/AIDS (UNAIDS) Key Population Atlas, UNAIDS Global AIDS Monitoring submissions, the US Centers for Disease Control and Prevention (CDC), the Global Fund Against TB, AIDS, and Malaria (GFTAM), and the Johns Hopkins University Global.HIV initiative. Existing initiatives include much overlap of data, but vary in scope, inclusion and exclusion criteria, data elements recorded, and linkage to and validation against primary source data sets and reports. Omission or incomplete recording of key methodological details inhibits appraisal and formal evidence synthesis, and therefore utility of data for strategic planning. Many other research studies have systematically reviewed, analysed, and extrapolated key population survey data in sub-Saharan Africa in single countries or across multiple countries. These studies have tended to focus on specific outcomes or population groups of interest and primarily reviewed peer-reviewed literature. **Added value of this study** We consolidated and deduplicated data from 2010-2021 from existing key population surveillance databases maintained by the UNAIDS Key Population Atlas, UNAIDS Global AIDS Monitoring (GAM), the US CDC, GFTAM. We used the Johns Hopkins University Global.HIV repository of surveillance reports and additional web-based searches, and engagement with in-country HIV strategic information teams to source primary sources in the peer-reviewed and grey literature, and validated each observation of KP population size, HIV prevalence, or ART coverage against primary surveillance reports. Using regression, we characterised the relationship between key population and total population HIV indicators and extrapolated key population size estimates (KPSE), HIV prevalence, and ART coverage data to national-level estimates for all countries in mainland SSA. This exercise was the most comprehensive effort to date to consolidate KP HIV data in SSA. We analysed 90 KPSE, 159 HIV prevalence, and 72 ART coverage studies. We estimated that across SSA countries, a median of 1.44% of urban women were FSW; 0.60% of urban men have sex with men; and 0.16% of urban adults injected drugs. Though FSW, MSM, and PWID combined were estimated as only 1.1% of the population they comprised 5.1% of all people living with HIV aged 15-49 years. KP ART coverage increased consistently with total population ART coverage, but lagged behind at high population coverage levels. We identified large gaps in data availability. Of the four KPs and three indicators studied, only Kenya and Mozambique had data for all twelve indicators. Data were particularly sparse for PWID in Southern Africa. **Implications of all the available evidence** Consolidated data show that key populations, including FSW, MSM, PWID, and TGW, experience disproportionate vulnerability to HIV and lower ART coverage across all settings in sub-Saharan Africa. This evidence along with synthesised extrapolated estimates provide a foundation for planning appropriate key population focused services for HIV prevention and treatment in all settings, including those with no or limited data. However, large data availability gaps, large heterogeneity, and inconsistency of existing data, and associated wide uncertainty ranges in resulting estimates limits the ability of existing data to guide granular programmatic planning and target setting for KP services or robustly monitor trends. New strategies and more consistent surveillance implementation are required credibly to monitor equitable and equal access to HIV prevention and treatment programmes outlined in the Global AIDS Strategy 2021-2026 in order to end HIV/AIDS as a public health threat by 2030. ## Introduction Key populations, including female sex workers (FSW), men who have sex with men (MSM), people who inject drugs (PWID), and transgender women (TGW), are disproportionately vulnerable to HIV in sub-Saharan Africa (SSA).1,2 The Global AIDS Strategy 2021-2026 calls for equitable and equal access to HIV prevention and treatment programmes through the removal of structural barriers to reduce HIV incidence and end HIV/AIDS as a public health threat by 2030.3 Delivering appropriate HIV prevention and treatment programming for these populations, and monitoring attainment of an equitable HIV response, requires robust information on key population size, HIV prevalence, the treatment cascade, and new HIV infections. Key population surveys rely on respondent-driven or network sampling methods4,5 because KPs are hard to reach in household survey sampling frames due to small population size, high mobility and transient living arrangements, non-disclosure of risk behaviour, and societal marginalisation.6–8 Surveys are commonly restricted to or disproportionately conducted in urban areas and studies are conducted infrequently. Consequently, data are sparse and less representative of a consistent target population than standard general population sampling approaches such as national household surveys or sentinel surveillance. Systematic reviews and meta-analyses of population size, HIV prevalence, or ART coverage encompassing SSA have been conducted for FSW,9,10 MSM,11 and PWID.12,13 Studies on transgender populations in SSA have been published,14–17 but data were insufficient to conduct region-level meta-analysis of HIV prevalence. Several organisations have undertaken efforts to consolidate key population surveillance data, including the Joint United Nations Program on HIV/AIDS (UNAIDS) Key Population Atlas, the US Centers for Disease Control and Prevention (CDC), the Global Fund to Fight AIDS, TB, and Malaria (GF), and the Johns Hopkins University Global.HIV initiative.18–21 These efforts aim to monitor the state of the epidemic, evaluate programmes, and make recommendations on key population data reporting and quality thresholds. We consolidated and harmonised key population size estimates, HIV prevalence, and ART coverage data from existing databases and described data availability across countries and over time for each key population group. Using regression, we characterised the relationship between key population and total population HIV indicators and extrapolated key population size, HIV prevalence, and ART coverage data to national-level estimates for all mainland countries in SSA. ## Methods ### Data sources and extraction We consolidated data on key population size estimates (KPSE), HIV prevalence, and ART coverage from four existing databases: data submitted for UNAIDS Global AIDS Monitoring,18 UNAIDS Key Population Atlas,22 the Global Fund surveillance database, and a spreadsheet maintained by the CDC’s Division of Global HIV & TB Key Population Surveillance Team. For each observation, we extracted data elements about the study methodology, study location (country, subnational location), study population sex and age group, central estimate, sample size, standard error, and primary source or reference (e.g. survey report). Descriptions of the databases and data extraction elements are in Supplementary Table S1 and S2. Primary source documents were compiled from archives accompanying each database, the Global.HIV document repository,20 and internet searches, following which we contacted UNAIDS Strategic Information advisors and other HIV programme contacts in each country to seek missing reports and other sources. Four authors (OS, RA, EG, AS-B) reviewed primary source documents to deduplicate data recorded in multiple databases, validate against primary sources, and extract missing data elements. During primary source review, we added observations in cases where additional relevant data were reported in primary sources or additional sources were ascertained that were not included in the initial databases. Observations were excluded if: data were duplicated across or within databases; information about key population definition, year, or surveillance area were missing; the surveillance area was non-specific (e.g. ‘urban areas’ or ‘5 provinces’); estimates were modelled or extrapolated; or data could not be confirmed by primary source review. ### Data processing KPSE reported as counts were converted to population proportion to facilitate comparison of population sizes across settings. Geographic boundaries from the Global Rural/Urban Mapping Project (GRUMP) and UNAIDS Naomi subnational district populations were used to derive total population denominators for subnational locations in which key population surveys were conducted.23–25 Each KPSE was matched to the total population denominator by age, sex, year, and area. String matching was used to assign a GRUMP or Naomi area name to each survey area, thereby matching to a total population denominator. Age information was missing from nearly all key population data entries; missing data was assigned ages 15-49 years. Unless a gender was specified, all sex workers were assumed to be female, PWID assumed to be both genders, and transgender people assumed to be transwomen (TGW). ### Classifying KPSE methods KPSE methods were classified into seven categories: two- and three-source capture-recapture (2S/3S-CRC); PLACE/programmatic mapping;26 object, service, and event multipliers; and successive-sampling population size estimates (SS-PSE). Within a given survey, multiple size estimation methods were commonly conducted which were combined into a final consensus KPSE. Where possible, separate estimates for each method were extracted. For cases where only a final estimate of multiple methods was reported, two further categories were defined: “Multiple methods - empirical” or “Multiple methods - mixture”. The former contained estimates derived from multiple of the seven methods above, while the latter were derived from both empirical and non-empirical methods (enumeration, wisdom of the crowds, key informant interviews, and the Delphi method). KPSEs derived by solely non-empirical methods were excluded from regression analyses. ### Transforming HIV prevalence and ART coverage data To facilitate cross-country comparison, key population HIV prevalence and ART coverage data were compared to estimates for HIV prevalence or ART coverage among the total population estimates for a given age, sex, year, and first administrative level (henceforth ‘province’). Province-level estimates of age/sex-specific HIV prevalence and ART coverage were extracted from UNAIDS Naomi subnational estimates for 202023,24 and projected back in time 2000-2021 parallel to the Spectrum27 15-49 national HIV prevalence and ART coverage trajectories. KP survey ART coverage data were either self-reported or laboratory confirmed via either antiretroviral metabolite biomarker or viral load testing. Viral load suppression (VLS) observations were converted to estimates of ART coverage assuming 90% VLS among those on ART. ### Analysis To synthesise KPSE observations, we used Bayesian mixed-effects linear regression to model logit-transformed KP proportion among the urban total population aged 15-49 separately for each KP. The model included fixed and random effects for study method, a spatially-correlated country effect allowing population proportions to be correlated between neighbouring countries, and a study-level random effect allowing for correlation in KP proportions observed at multiple subnational locations in the same study. The fixed effect for study method had three levels: (1) a pooled reference category of capture-recapture, SS-PSE, and multiplier methods for which KP population proportions were assumed unbiased; (2) PLACE/mapped estimates; and (3) other methods. A random effect for each method permitted method-specific deviation from the fixed effect estimates. National KPSE counts were extrapolated using estimated population proportions for urban areas, the proportion of the total population living in urban areas,28 and the relative abundance of KPs in urban and rural areas. Expert consensus was that KP proportions were higher in urban areas than rural areas, but no empirical population size data were available to inform an estimate. To allow for uncertainty in the ratio of rural-to-urban population proportion, we assumed a *Beta*(5,3) distribution which had mean ratio of 0.6 and 80% of the mass between 0.4 and 0.8. For HIV prevalence, we modelled the relationship between logit-transformed KP HIV prevalence and logit total population HIV prevalence (age 15-49 years) separately for FSW and PWID, and modelled together for MSM and TGW.29 Missing denominators for HIV prevalence observations were imputed using the 25th-centile of observed denominators (imputed denominators: FSW n=192; MSM n=108; PWID n=41). The number of KPs living with HIV (KPLHIV) was modelled with a beta-binomial distribution to allow overdispersion. The model included fixed effects for logit population prevalence interacted with region (ESA/WCA) to allow a different relationship between KP and total population prevalence across regions, a fixed effect for whether HIV status was self-reported or via HIV diagnostic tests, and country-level and study-level random effects. An additional fixed effect for KP was included in the MSM-TGW model. Country-level random effects were spatially correlated. Studies that omitted whether HIV prevalence was self-reported or diagnostically confirmed were assumed to be self-reported. Logit-transformed KP ART coverage was modelled as a function of logit total population ART coverage in the same year, age group, and province, analogously to HIV prevalence. For each key population, missing denominators were imputed using the 25th-centile of known denominators (imputed denominators: FSW n=44; MSM n=23; PWID n=28). The observed number on ART was modelled with a beta-binomial distribution with fixed effect terms for logit population ART coverage interacted with region, spatially correlated country-level random effects and study-level random effects. We conducted sensitivity analyses which included a fixed effect for diagnostically confirmed or self-reported ART usage, with diagnostically confirmed as the reference category. For all three models (KPSE, prevalence, ART coverage), in sensitivity analysis we used ages 15-29 as the matched total population denominator for MSM and TGW regression analyses reflecting the young median age captured in MSM and TGW surveys.30,31 We combined national KPSE, HIV prevalence, and ART coverage estimates to calculate the number of KPLHIV and on ART. 95% uncertainty ranges were generated by combining 1000 posterior samples for each outcome. KPLHIV estimates were compared with 2021 national UNAIDS estimates32 for age 15-49 years to calculate the proportion of PLHIV who were members of each KP. Data were extracted, deduplicated, and validated in Microsoft Excel. Statistical analyses were conducted in R version 4.1.2 using the R-INLA package (v21.12.21.2). Further methods information is available in Supplementary Text S2. ### Role of the funding source Authors KS, SA-G, MB, and MM are employees of UNAIDS and contributed to the conceptualisation of the study, interpretation of results, and editing the manuscript. All authors had access to all study data. OS and JWE accept responsibility to submit for publication. ## Results ### Key population survey data availability 5014 KPSE conducted between 2010 to 2021 were compiled (Table 1; Supplementary Table S1). Following data cleaning, primary source review, and area matching, 1449 observations were extracted from 90 studies (Figure 1). Data were most available for FSW (n=657, data from 30/38 countries), followed by MSM (n=445, 27/38 countries), PWID (n=274, 16/38 countries), and TG (n=73, 9/38 countries). No TG KPSE were disaggregated by gender. View this table: [Table 1:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/T1) Table 1: Availability of population size, HIV prevalence, and ART coverage data by key population and region during period 2010-2021. ![Figure 1:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F1.medium.gif) [Figure 1:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F1) Figure 1: Flow diagrammes describing identified data for (A) key population size estimates, (B) HIV prevalence data, and (C) key population ART coverage. Each *n* represents number of population/method/location observations, with potentially observations for multiple subnational locations from the same study in same country. ART: Antiretroviral coverage; FSW: Female sex worker; MSM: Men who have sex with men; PWID: People who inject drugs; TG: Transgender people; KP: Key Population; GAM: Global AIDS Monitoring; CDC: US Centers for Disease Control and Prevention; PSE: Population size estimate. KPSE studies have become more common since 2010 (Figure 2). In 2016-2018, 23 FSW, 23 MSM, 12 PWID, and 6 TG PSE studies were recorded, increasing from 15, 12, 6, and 0, respectively, in 2010-12. The proportion of KPSE derived from empirical methods increased from 71% (22/31) in 2010-2012 to 90% in 2019-2021 (53/59) (Figure 2). Seventy percent (310/445) of MSM population proportion estimates were below 1% (68% [200/294] in ESA and 73% [110/151] in WCA; Supplementary Figure S1). ![Figure 2:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F2.medium.gif) [Figure 2:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F2) Figure 2: Distribution of size estimation methods implemented over time by key population. The number of size estimation studies is shown above the bars. Bars are colour coded by method empiricism from highest (3S-CRC) to lowest (non-empirical). Female sex workers (FSW); Men who have sex with men (MSM); People who inject drugs (PWID); Transgender people (TG); Three-source capture-recapture (3S-CRC); Two-source capture-recapture (2S-CRC); Successive-sampling population size estimate (SS-PSE). Data consolidation identified 3944 KP HIV prevalence estimates (Table 1; Supplementary Table S1), from which 1181 were extracted from 159 studies after deduplication, source review, and area matching. Denominators were reported for 88% (1039/1181) of observations. Most data were available for FSW (n=568, 33/38 countries), followed by MSM (n=359, 30/38 countries), PWID (n=159, 17/38 countries), TGW (n=87, 20/38 countries), and TGM (n=8, 4/38 countries). Most studies reported diagnostically confirmed HIV prevalence estimates (89%; 142/159). For ART coverage, data consolidation identified 607 initial observations (Table 1; Supplementary Table S1). Following data processing, 242 observations for KP ART coverage were extracted from 72 studies. Denominators were available for 90% (216/242). Data were most available for FSW (n=129, 20/38 countries), followed by MSM (n=78, 17/38), PWID (n=19, 6/38 countries), and TGW (n=16, 10/38 countries). Around half of studies reported laboratory-confirmed treatment status (46%; 33/72). ### Key population size estimates Across SSA countries, a median of 1.44% of urban women aged 15-49 years were FSW (interquartile range [IQR] 0.83-1.89%), 0.60% of men had sex with men (IQR 0.53-0.90%), and 0.16% of adults injected drugs (IQR 0.14-0.24%; Figure 3, Supplementary Table S4). Data were insufficient to estimate transgender population proportions. In sensitivity analysis using age 15-29 years as denominator for MSM KPSEs, the proportion increased to 1.00% (IQR 0.82-1.40%) of men aged 15-29 (Supplementary Figure S2). ![Figure 3:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F3.medium.gif) [Figure 3:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F3) Figure 3: Model estimated urban key population size estimate (KPSE) proportions for female sex workers (FSW), men who have sex with men (MSM), and people who inject drugs (PWID) as a proportion of sex-matched 15-49 adult total population. (A) Posterior median estimate for each country. Note: colour range is different for each KP chloropleth. (B) Posterior median estimate and 95% credible intervals for each country. Points are observed subnational KPSE proportion data points that used empirical methods. Countries where modelled estimates were informed by local surveillance data are shown in blue, and countries informed only by spatial smoothing from neighbouring countries in yellow. Vertical axis shown on log-scale. Urban PSE proportions for FSW were similar in ESA and WCA (ESA: 1.46%, IQR 1.13-1.92%; WCA 1.43%, IQR 0.83-1.64%). MSM proportions were lower in WCA than ESA (ESA: 0.86%, IQR 0.66-1.01%; WCA: 0.55%, IQR 0.49-0.59%). Conversely, PWID proportions were lower in ESA (0.14%, IQR 0.13-0.17%) than in WCA (0.20%, IQR 0.14-0.29%), but PWID KPSE data in ESA were sparse. Only two southern African countries had PWID KPSE data, and most data points in ESA were from district-level mapping studies in Kenya and Uganda. Relative to empirical methods, estimates derived from PLACE and mapping studies for PWID were lower (OR 0.35, 95%CI 0.13-0.88), and marginally lower for MSM (OR 0.65, 95%CI 0.36-1.33), but not for FSW (OR 1.12, 95%CI 0.48-2.60) (Table 2). View this table: [Table 2:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/T2) Table 2: Odds ratios for population size proportions by method relative to population size estimates pooled across all empirical methods (Two and three source CRC (2S- and 3S-CRC), successive sampling population size estimation (SS-PSE), and multiplier methods). Each key population estimated in separate regression models, but results presented together for ease of comparability. ### HIV prevalence among key populations Most KP HIV prevalence observations were higher than area-matched total population HIV prevalence (FSW 95% (492/518); MSM 84% (259/310); PWID 81% 105/130); TGW 84% (73/87)). Key population HIV prevalence in ESA was more strongly correlated with matched total population HIV prevalence than in WCA (ESA: FSW R2=0.40, MSM R2=0.44, PWID R2=0.09, TGW R2=0.48; WCA: FSW R2=0.34, MSM R2=0.12, PWID R2=0.31, TGW R2=0.31). Relative HIV prevalence differentials between KP and total population were larger when population prevalence was lower (Figure 4). For example, in ESA, 1% total population prevalence corresponded to a 20% predicted prevalence among FSW (95% prediction range 7-45%), 8% among MSM (3-21%), 3% among PWID (1-7%), and 10% among TGW (3-27%). Fifteen percent population HIV prevalence corresponded to FSW HIV prevalence of 41% (18-69%), MSM prevalence of 19% (8-40%), PWID prevalence of 36% (26-50%), and TGW prevalence of 24% (10-47%). In WCA, relative patterns were similar, but KP HIV prevalence varied less with total population HIV prevalence than in ESA. Supplementary Table S5 reports country-specific estimates. ![Figure 4:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F4.medium.gif) [Figure 4:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F4) Figure 4: Key population and total population HIV prevalence on the logit scale (upper), natural scale (middle), and expressed as a ratio of KP and total population prevalence (lower) for female sex workers (FSW), men who have sex with men (MSM), people who inject drugs (PWID), and transgender people (TG). Grey points indicate observed KP prevalence plotted against province-level total population prevalence. Black points represent country estimates and 95% uncertainty ranges. Dotted line represents line of equality. ART coverage among FSW and MSM ART coverage was strongly correlated with population ART coverage (logit-scale R2=0.67 and 0.73, respectively), but less correlated among PWID (R2=0.33; Figure 5). In models not adjusted for ART measurement method (lab-confirmed versus self-reported), at 40% population ART coverage, predicted FSW ART coverage was 8% higher (95% prediction range 0-18%), MSM ART coverage was 7% lower (−3-15%), and in PWID 4% lower (−13-18%). At 80% population coverage, predicted FSW ART coverage was 10% lower (3-18%), in MSM 19% lower (7-32%), and in PWID 21% lower (6-37%). Supplementary Table S6 reports country-specific estimates. ![Figure 5:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F5.medium.gif) [Figure 5:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F5) Figure 5: Estimated key population ART coverage as a function of total population ART coverage on the estimated logit scale (upper) and natural scale (lower). Black points represent national estimates. In sensitivity analysis including a fixed effect for ART measurement method, self-reported ART was significantly higher than lab-confirmed ART in FSW and MSM, but not in PWID (FSW: OR 1.68, 95%CI 1.00-2.86; MSM: OR 2.16, 1.07-4.39; PWID OR 2.22, 0.63-7.55; Supplementary Table S3). Assuming lab-confirmed ART coverage was unbiased reduced predicted key population ART coverage (Supplementary Figure S3). When assuming surveys represented MSM aged 15-29 years, the gap between total male population and MSM ART coverage was smaller. For example, 80% total population ART coverage corresponded to 66% MSM ART coverage when using 15-29 year reference age compared to 61% when using age 15-49 years (Supplementary Figure S6). ### Disproportionate HIV burden among key populations in SSA Across SSA, FSW, MSM and PWID combined were 1.5% (95%CI 1.1-2.0%) of the population aged 15-49 years, but 6.3% (4.8-8.5%) of PLHIV. In ESA this was 4.5% (3.3-6.4%) of PLHIV compared to 1.2% (0.9-1.8%) of the population and, in WCA, 14.2% (9.9-21.3%) of PLHIV compared to 1.6% (1.2-2.5%) of the population. FSW were 4.1% (2.8-5.9%) of all PLHIV in SSA (790,000 PLHIV; 540,000-1,100,000), MSM were 1.5% (0.9-2.7%) or 300,000 PLHIV (180,000-520,000), and PWID were 0.6% (0.3-1.3%) or 120,000 PLHIV (60,000-250,000) (Figure 6). ![Figure 6:](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2022/07/29/2022.07.27.22278071/F6.medium.gif) [Figure 6:](http://medrxiv.org/content/early/2022/07/29/2022.07.27.22278071/F6) Figure 6: (A) HIV positive KPs as a proportion of all PLHIV aged 15-49 years. Countries are ordered geographically. (B) Estimated number of KP living with HIV (KPLIV) in Eastern and Southern Africa and Western and Central Africa. ## Discussion Consolidated key population survey data reinforce that KPs are disproportionately vulnerable to HIV in sub-Saharan Africa. We estimated there are 3.7m FSW, 1.9m MSM, and 1.4m PWID in SSA who require comprehensive HIV prevention or treatment services. HIV prevalence was 18% among FSW, 13% among MSM, 6% among PWID, and 19% among TGW, 3.6, 4.8, 1.5, and 3.8 times higher than population prevalence, respectively, with larger relative differences when population prevalence was lower. Together, FSW, MSM, and PWID constituted 1.5% of the adult population in SSA but 6.3% of PLHIV. ART coverage among KPs increased consistently with total population coverage, but at high levels of ART coverage, coverage among KPs lagged somewhat behind. However, there was wide uncertainty about estimates at continental level and within countries. This exercise highlighted major data gaps and challenges interpreting existing data. Across the twelve combinations of four KPs and three indicators analysed here, we sourced a median of five combinations in each country. Only Kenya and Mozambique had data for all twelve, and six countries lacked any key population data. These gaps impede evaluation of the Global AIDS Strategy 2021-26 on progress towards equitable and equal access to HIV services. In settings with lacking or limited data, extrapolated estimates should be considered a foundation with which to target future surveillance, stimulate in-country data use and review processes, and estimate HIV epidemic indicators. Most countries had some surveillance for FSW; future efforts should especially focus on strengthening surveillance for MSM, PWID, and TG populations. Previous key population reviews and meta-analyses10,13,14,33 relied on systematic searches of peer-reviewed literature. Our initial sources were existing databases, primarily populated by in-country surveillance teams or implementing organisations. This resulted in inclusion of a larger grey literature body, such as survey reports, which is poorly indexed by peer-reviewed databases.29 However, many observations (12% of PSE data, 26% of HIV prevalence data, 16% of ART coverage data), could not be sourced and were excluded from this analysis, which in some countries excluded the only available data. Strong data use for general population HIV estimates in SSA has promoted high-quality comparable surveillance methods, documentation, and dissemination. The absence of such systems for key population data has led to weaker sourcing, ownership, dissemination, and retention of these data which limits more complex epidemiological analyses. This next step for this project is to produce data workbooks with HIV surveillance teams in each country to review and provide context for existing data; identify missing sources and add new data; and create a foundation for KP surveillance data and strategic information derived from them to become a core component of the UNAIDS-supported HIV estimates process. That FSW, MSM, and PWID constitute 6.3% of SSA PLHIV aged 15-49 does not directly relate to the distribution of new infections across population groups or future incidence patterns. UNAIDS estimates that key populations comprise 25% of new infections in 2021 in SSA.32 Moreover, mathematical modelling has highlighted that preventing and treating HIV among key populations has larger impact on reducing population incidence due to network effects of averting further onward transmission.34,35 Indicators and modelling that capture these impacts are a better basis for intervention prioritisation than solely burden estimates.36,37 Population size data measured in the same population at different times or the with multiple methods were highly heterogenous. This resulted in very large uncertainty ranges around modelled estimates for population size and PLHIV spanning five-fold differences or greater, even in countries with recent size estimates using modern empirical methods. Large uncertainty ranges were consistent with findings of others who have statistically summarised and extrapolated PSE from multiple sources.38–41 Virtually all KPSE data were collected among selected urban populations, and we lacked data from any setting to empirically inform plausible rural-urban population proportions. Such heterogeneity for KPs size may limit the utility of these data for establishing meaningful programmatic targets for those in need of HIV prevention, testing, and treatment services, or to monitor temporal trends. More sophisticated methods have been developed to pool multiple estimates of population size and extrapolate to district-level.41–44 Our median PWID population proportion of 0.16% across SSA countries was lower than estimates from Degenhardt et al. of 0.28%, primarily reflecting exclusion of SSA island states with concentrated HIV epidemics and higher PWID proportions from this analysis.13 Meta-analysis by Stannah et al.11 of studies conducted up to 2016 estimated that MSM ART coverage in SSA is 24%, lower than estimates from this analysis which may represent programmatic scale-up amongst MSM in recent years. This analysis was consistent with Hessou et al.33 in identifying heterogeneous HIV prevalence ratios between the total population and MSM, particularly in WCA where HIV burden in neighbouring countries can vary considerably. In our analysis, 70% percent of MSM KPSE observations and 34 of 38 modelled urban MSM estimates from this analysis were below 1%, which UNAIDS and WHO recommended in 2016 as a minimum MSM population proportion.19 The minimum threshold was specified to ensure that MSM were not overlooked due to lack of data. The difference may arise because empirical PSE methods focus on enumerating visible MSM with higher HIV vulnerability, rather the number men who would have sex with men if able, which may lie around 1% or greater. Our analysis included many empirical PSE studies conducted since the 2016 guidance, and the threshold should be reconsidered with the wider range of data now available, aligned with risk-based approaches to quantifying sex work and injecting drug use. Considering the intrinsic challenges robustly surveying hard-to-reach populations and the ultimate purpose of surveillance to guide programmatic response, incorporating key population programme data within key population estimates should be a priority to support the production of longitudinal estimates and subnational estimates.44 Triangulating routine HIV testing and treatment data with biobehavioural survey data will be required to address programmatic double counting, coverage of key population-specific services, and accounting for linkage to care in the estimation of ART coverage and viral load suppression.7,45,46 Future surveillance will also need to move beyond strictly defined key population definitions to reflect individuals in overlapping risk environments including cis-MSM and transwomen who sell sex, and FSW and MSM who inject drugs.47–49 Several limitations should be considered when interpretating these estimates. Firstly, surveys used a range of key population definitions and inclusion criteria which limited comparability. Second, defining appropriate total population denominators to calculate KPSE proportions for surveys without explicitly geographic sampling frames is challenging, and the urban extents derived from the Global Rural/Urban Mapping Project may not accurately reflect the catchment population for a given survey, particularly given the mobility of key populations.50 This likely exacerbates the observed heterogeneity in KPSE proportions (Figure 3B), and is a major limitation for extrapolating KPSE proportions for programmatic planning. Going forward, inclusion of population catchments with KP surveillance reports will improve future data use. Third, sample size or standard errors were omitted from most population size data entries and all size estimates were equally weighted within the model. In some cases, sample sizes or standard error estimates could be recovered with further review of primary sources. Fourth, data were insufficient and too heterogeneous to estimate time trends in population proportions, HIV prevalence, or ART coverage. Application of standardised KP survey methodology in consistent locations will improve monitoring of trends. Finally, regression models for KP ART coverage that included measurement method estimated lower odds of ART usage when it was laboratory confirmed versus self-reported. However, this estimate was inconsistent with studies comparing self-report and lab-confirmed ART status in the same study population, which consistently find that self-report understated true ART usage among KP samples51–54 and general population samples.55–57 Therefore, we did not incorporate systematic difference by method into modelled ART coverage estimates, but this adds uncertainty to our assessment of ART coverage. In conclusion, key populations across sub-Saharan Africa experience disproportionate HIV burden and have somewhat lower antiretroviral treatment coverage. Despite increasing focus on key populations in sub-Saharan Africa in AIDS strategies, estimates that may be insufficiently precise to plan and monitor programmes and limited ability to evaluate trends due to large heterogeneity in surveillance data. Consolidated KP data and synthesised estimates provide a basis for key population programming in all countries, including those with limited locally available data. However, future effort should focus on integrating and strengthening key population surveillance and routine data within national HIV surveillance systems to plan and monitor more effective HIV responses. ## Supporting information Supplementary File 1 - Figures and Tables [[supplements/278071_file03.pdf]](pending:yes) Supplementary File 2 - Data [[supplements/278071_file04.xlsx]](pending:yes) ## Data Availability All survey data produced in the present work are contained in the supplementary materials All remaining data are available upon reasonable request to the authors ## Authors’ contributions OS, KS, MM and JWE conceptualised the study. KS, SAG, JZ, AA-Q, KW, LD curated key population survey databases. OS, RA, EG, AS-B reviewed primary source documents, deduplicated data, and extracted data. OS analysed the data and wrote the first draft of the manuscript. All authors contributed to interpretation of results and edited the manuscript for intellectual content. All authors read and approved the final version of the manuscript for submission. ## Acknowledgments This research was supported by UNAIDS and the MRC Centre for Global Infectious Disease Analysis (reference MR/R015600/1), jointly funded by the UK Medical Research Council (MRC) and the UK Foreign, Commonwealth & Development Office (FCDO), under the MRC/FCDO Concordat agreement and is also part of the EDCTP2 programme supported by the European Union. * Received July 27, 2022. * Revision received July 27, 2022. * Accepted July 29, 2022. * © 2022, Posted by Cold Spring Harbor Laboratory This pre-print is available under a Creative Commons License (Attribution-NonCommercial-NoDerivs 4.0 International), CC BY-NC-ND 4.0, as described at [http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/) ## References 1. 1.UNAIDS. The Gap Report. (2014). 2. 2.Garnett, G. P. Reductions in HIV incidence are likely to increase the importance of key population programmes for HIV control in sub-Saharan Africa. J. Int. AIDS Soc. 24, e25727 (2021). 3. 3.United Nations. End Inequalities. End AIDS. Global AIDS Strategy 2021-2026 | UNAIDS. [https://www.unaids.org/en/resources/documents/2021/2021-2026-global-AIDS-strategy](https://www.unaids.org/en/resources/documents/2021/2021-2026-global-AIDS-strategy). 4. 4.Baneshi, M. R., Rastegari, A. & Haghdoost, A. A. Review of Size Estimation Methods. in Advances in Experimental Medicine and Biology vol. 1333 1–15 (Springer, 2021). 5. 5.Abdul-Quader, A. S., Baughman, A. L. & Hladik, W. Estimating the size of key populations: Current status and future possibilities. Current Opinion in HIV and AIDS (2014) doi:10.1097/COH.0000000000000041. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/COH.0000000000000041&link_type=DOI) 6. 6.Jin, H., Restar, A. & Beyrer, C. Overview of the epidemiological conditions of HIV among key populations in Africa. Journal of the International AIDS Society vol. 24 (2021). 7. 7.Hakim, A. J. et al. Gaps and opportunities: measuring the key population cascade through surveys and services to guide the HIV response. J. Int. AIDS Soc. 21, e25119 (2018). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jia2.25119&link_type=DOI) 8. 8.Sabin, K. et al. Availability and quality of size estimations of female sex workers, men who have sex with men, people who inject drugs and transgender women in low- and middle-income countries. PLoS One 11, (2016). 9. 9.Atuhaire, L., Adetokunboh, O., Shumba, C. & Nyasulu, P. S. Effect of community-based interventions targeting female sex workers along the HIV care cascade in sub-Saharan Africa: a systematic review and meta-analysis. Syst. Rev. (2021) doi:10.1186/s13643-021-01688-4. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1186/s13643-021-01688-4&link_type=DOI) 10. 10.Baral, S. et al. Burden of HIV among female sex workers in low-income and middle-income countries: A systematic review and meta-analysis. Lancet Infect. Dis. (2012) doi:10.1016/S1473-3099(12)70066-X. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1473-3099(12)70066-X&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=22424777&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000306045300027&link_type=ISI) 11. 11.Stannah, J. et al. HIV testing and engagement with the HIV treatment cascade among men who have sex with men in Africa: a systematic review and meta-analysis. Lancet HIV (2019) doi:10.1016/S2352-3018(19)30239-5. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S2352-3018(19)30239-5&link_type=DOI) 12. 12.Mathers, B. M. et al. Global epidemiology of injecting drug use and HIV among people who inject drugs: a systematic review. Lancet (2008) doi:10.1016/S0140-6736(08)61311-2. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(08)61311-2&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18817968&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000261077900023&link_type=ISI) 13. 13.Degenhardt, L. et al. Global prevalence of injecting drug use and sociodemographic characteristics and prevalence of HIV, HBV, and HCV in people who inject drugs: a multistage systematic review. Lancet Glob. Heal. 5, e1192–e1207 (2017). 14. 14.Stutterheim, S. E., Van Dijk, M., Wang, H. & Jonas, K. J. The worldwide burden of HIV in transgender individuals: An updated systematic review and meta-analysis. PLoS ONE vol. 16 (2021). 15. 15.King, R. et al. HIV and transgender women in Kampala, Uganda–Double Jeopardy. Cult. Heal. Sex. 21, 727–740 (2019). 16. 16.Stevens, O. Estimating key population size, HIV prevalence, and ART coverage for sub-Saharan Africa at the national level. (2022). 17. 17.Mujugira, A. et al. “You are not a man”: a multi-method study of trans stigma and risk of HIV and sexually transmitted infections among trans men in Uganda. J. Int. AIDS Soc. 24, (2021). 18. 18.UNAIDS. Global AIDS Monitoring 2022. [https://www.unaids.org/en/global-aids-monitoring](https://www.unaids.org/en/global-aids-monitoring) (2022). 19. 19.UNAIDS & WHO. Recommended population size estimates for men who have sex with men. (2020). 20. 20.Rao, A. et al. HIV-related data among key populations to inform evidence-based responses: protocol of a systematic review. Syst. Rev. (2018). 21. 21.UNAIDS, FHI 360, WHO, CDC & PEPFAR. Biobehavioural Survey Guidelines For Populations At Risk For HIV. (2017). 22. 22.UNAIDS. Key Populations Atlas. [https://kpatlas.unaids.org/dashboard](https://kpatlas.unaids.org/dashboard). 23. 23.UNAIDS. HIV sub-national estimates viewer. [https://naomi-spectrum.unaids.org/](https://naomi-spectrum.unaids.org/) (2021). 24. 24.Eaton, J. W. et al. Naomi: a new modelling tool for estimating HIV epidemic indicators at the district level in sub-Saharan Africa. J. Int. AIDS Soc. (2021) doi:10.1002/jia2.25788. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jia2.25788&link_type=DOI) 25. 25.Global Rural-Urban Mapping Project (GRUMP). Urban boundaries. [https://sedac.ciesin.columbia.edu/data/collection/grump-v1](https://sedac.ciesin.columbia.edu/data/collection/grump-v1). 26. 26.Weir, S. Priorities for Local AIDS Control Efforts (PLACE). [https://www.measureevaluation.org/resources/publications/ms-19-173/at_download/document](https://www.measureevaluation.org/resources/publications/ms-19-173/at_download/document) (2019). 27. 27.Eaton, J. W. et al. The Estimation and Projection Package Age-Sex Model and the r-hybrid model: New tools for estimating HIV incidence trends in sub-Saharan Africa. AIDS 33, S235–S244 (2019). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1097/QAD.0000000000002437&link_type=DOI) 28. 28.Nations, U., of Economic, D., Affairs, S. & Division, P. World Urbanization Prospects The 2018 Revision. (2018). 29. 29.Stevens, O. HIV prevalence in transgender populations and cisgender men who have sex with men in sub-Saharan Africa 2010-2021: a meta-analysis. (2022). 30. 30.Johnson, L. F., Mulongeni, P., Marr, A. & Lane, T. Age bias in survey sampling and implications for estimating HIV prevalence in men who have sex with men: Insights from mathematical modelling. Epidemiol. Infect. 146, 1036–1042 (2018). 31. 31.Stannah, J. et al. HIV testing and engagement with the HIV treatment cascade among men who have sex with men in Africa: a systematic review and meta-analysis. Lancet HIV 6, e769–e787 (2019). 32. 32.UNAIDS. Global AIDS Update. [https://www.unaids.org/en/resources/documents/2021/2021-global-aids-update](https://www.unaids.org/en/resources/documents/2021/2021-global-aids-update) (2021). 33. 33.Hessou, P. H. S. et al. Comparison of the prevalence rates of HIV infection between men who have sex with men (MSM) and men in the general population in sub-Saharan Africa: A systematic review and meta-analysis. BMC Public Health 19, 1634 (2019). 34. 34.Stone, J. et al. Estimating the contribution of key populations towards HIV transmission in South Africa. J. Int. AIDS Soc. 24, e25650 (2021). 35. 35.Mishra, S., Pickles, M., Blanchard, J. F., Moses, S. & Boily, M. C. Distinguishing sources of HIV transmission from the distribution of newly acquired HIV infections: Why is it important for HIV prevention planning? Sex. Transm. Infect. 90, 19–25 (2014). [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoic2V4dHJhbnMiO3M6NToicmVzaWQiO3M6NzoiOTAvMS8xOSI7czo0OiJhdG9tIjtzOjUwOiIvbWVkcnhpdi9lYXJseS8yMDIyLzA3LzI5LzIwMjIuMDcuMjcuMjIyNzgwNzEuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 36. 36.Mishra, S. et al. Validation of the modes of transmission model as a tool to prioritize HIV prevention targets: A comparative modelling analysis. PLoS One 9, (2014). 37. 37.Shubber, Z., Mishra, S., Vesga, J. F. & Boily, M. C. The HIV modes of transmission model: A systematic review of its findings and adherence to guidelines. Journal of the International AIDS Society vol. 17 (2014). 38. 38.Edwards, J. K. et al. Estimating Sizes of Key Populations at the National Level: Considerations for Study Design and Analysis. Epidemiology 29, 795–803 (2018). 39. 39.Niu, X. M. et al. Using factor analyses to estimate the number of female sex workers across Malawi from multiple regional sources. Ann. Epidemiol. 55, 34–40 (2021). 40. 40.Fearon, E. et al. Estimating the Population Size of Female Sex Workers in Zimbabwe: Comparison of Estimates Obtained Using Different Methods in Twenty Sites and Development of a National-Level Estimate. J. Acquir. Immune Defic. Syndr. 85, 30–38 (2020). 41. 41.Datta, A. et al. Bayesian Estimation of MSM Population Size in Côte d’Ivoire. Stat. Public Policy 6, 1–13 (2019). 42. 42.Wesson, P. D., Mirzazadeh, A. & McFarland, W. A Bayesian approach to synthesize estimates of the size of hidden populations: The anchored multiplier. Int. J. Epidemiol. 47, 1636–1644 (2018). 43. 43.Fellows, I. Consensus Estimate Calculator. [https://epiapps.com/shiny/app\_direct/shinyproxy\_combine\_estimates/](https://epiapps.com/shiny/app_direct/shinyproxy_combine_estimates/). 44. 44.USAID. Namibia Small Area Estimation: Final Report Small Area Estimation of Key Population Sizes in Namibia: Final Report. (2021). 45. 45.Scheibe, A. P. Still left behind: Using programmatic data to assess harm reduction service coverage and HIV treatment cascades for people who inject drugs in five South African cities. in IAS (2019). 46. 46.Scheibe, A. et al. Modelling the UNAIDS 90-90-90 treatment cascade for gay, bisexual and other men who have sex with men in South Africa: using the findings of a data triangulation process to map a way forward. AIDS Behav. 22, 853–859 (2018). 47. 47.Syvertsen, J. L., Agot, K., Ohaga, S. & Bazzi, A. R. You can’t do this job when you are sober: Heroin use among female sex workers and the need for comprehensive drug treatment programming in Kenya. Drug Alcohol Depend. 194, 495–499 (2019). 48. 48.Poteat, T. et al. HIV prevalence and behavioral and psychosocial factors among transgender women and cisgender men who have sex with men in 8 African countries: A cross-sectional analysis. PLoS Med. (2017) doi:10.1371/journal.pmed.1002422. [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1371/journal.pmed.1002422&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) 49. 49.Baral, S. D. et al. Worldwide burden of HIV in transgender women: A systematic review and meta-analysis. Lancet Infect. Dis. 13, 214–222 (2013). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S1473-3099(12)70315-8&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23260128&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000318386100022&link_type=ISI) 50. 50.Davey, C. et al. Mobility and sex work: why, where, when? A typology of female-sex-worker mobility in Zimbabwe. Soc. Sci. Med. 220, 322–330 (2019). 51. 51.Hakim, A. J. et al. An urgent need for HIV testing among men who have sex with men and transgender women in Bamako, Mali: Low awareness of HIV infection and viral suppression among those living with HIV. PLoS One 13, 1–14 (2018). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1371/journal.pone.0191962&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=29370308&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) 52. 52.Smith, A. D. et al. HIV burden and correlates of infection among transfeminine people and cisgender men who have sex with men in Nairobi, Kenya: an observational study. Lancet HIV 8, e274–e283 (2021). 53. 53.Fogel, J. M. et al. Accuracy of Self-Reported HIV Status Among African Men and Transgender Women Who Have Sex with Men Who were Screened for Participation in a Research Study: HPTN 075. AIDS Behav. 23, 289–294 (2019). [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2022%2F07%2F29%2F2022.07.27.22278071.atom) 54. 54.Cowan, F. M. et al. The HIV care cascade among female sex workers in Zimbabwe: Results of a population-based survey from the sisters antiretroviral therapy programme for prevention of HIV, an integrated response (SAPPH-IRe) trial. J. Acquir. Immune Defic. Syndr. 74, 375–382 (2017). 55. 55.Grabowski, M. K. et al. The validity of self-reported antiretroviral use in persons living with HIV: A population-based study. AIDS 32, 363–369 (2018). 56. 56.Kim, A. A. et al. Undisclosed HIV infection and antiretroviral therapy use in the Kenya AIDS indicator survey 2012: Relevance to national targets for HIV diagnosis and treatment. in AIDS vol. 30 2685–2695 (Lippincott Williams and Wilkins, 2016). 57. 57.Huerga, H. et al. A comparison of self-report and antiretroviral detection to inform estimates of antiretroviral therapy coverage, viral load suppression and HIV incidence in Kwazulu-Natal, South Africa. BMC Infect. Dis. 17, (2017).