https://orcid.org/0000-0002-0306-5171
Figures
Abstract
As antiretroviral treatment (ART) coverage for people living with HIV (PLHIV) increases, HIV programmes require up-to-date information about evolving HIV risk behaviour and transmission risk, including those with low-level viremia (LLV; >50 to ≤1000 copies/mL), to guide prevention priorities. We aimed to assess differences in sexual risk behaviours, distribution of viral load (VL) and proportion of transmission across PLHIV subgroups. We analysed data from Population-based HIV Impact Assessment surveys in 14 sub-Saharan African countries during 2015–2019. We estimated adjusted prevalence ratios (aPR) of self-reported HIV high-risk behaviour (multiple partners and condomless sex) across cascade stages via generalised estimation equations. We modelled the proportions of transmission from each subgroup using relative self-reported sexual risk, a Hill function for transmission rate by VL, and proportions within cascade stages from surveys and UNAIDS country estimates for 2010–2020. Compared to PLHIV with undetectable VL (≤50 copies/mL), undiagnosed PLHIV (aPR women: 1.28 [95% CI: 1.08–1.52]; men: 1.61 [1.33–1.95]) and men diagnosed but untreated (2.06 [1.52–2.78]) were more likely to self-report high-risk sex. High-risk behaviour was not significantly associated with LLV. Mean VL was similar among undiagnosed, diagnosed but untreated, and on ART but non-suppressed sub-groups. Across surveys, undiagnosed and diagnosed but untreated contributed most to transmission (40–91% and 1–41%, respectively), with less than 1% from those with LLV. Between 2010 and 2020, the proportion of transmission from individuals on ART but non-suppressed increased. In settings with high ART coverage, effective HIV testing, ART linkage, and retention remain priorities to reduce HIV transmission. Persons with LLV are an increasing share of PLHIV but their contribution to HIV transmission was small. Improving suppression among PLHIV on ART with VL ≥1000 copies/mL will become increasingly important.
Citation: Edun O, Okell L, Chun H, Bissek A-CZ, Ndongmo CB, Shang JD, et al. (2024) HIV risk behaviour, viraemia, and transmission across HIV cascade stages including low-level viremia: Analysis of 14 cross-sectional population-based HIV Impact Assessment surveys in sub-Saharan Africa. PLOS Glob Public Health 4(4): e0003030. https://doi.org/10.1371/journal.pgph.0003030
Editor: Nicola Abrescia, AORN Ospedali dei Colli, ITALY
Received: October 16, 2023; Accepted: February 25, 2024; Published: April 4, 2024
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: The Nigeria HIV/AIDS Indicator and Impact Survey (NAIIS) data and data for the remaining 13 PHIA surveys are publicly available upon request from the NAIIS Data Archive portal: https://www.nigerianstat.gov.ng/nada/ and the PHIA website, hosted by ICAP, Columbia University: https://phia.icap.columbia.edu/, respectively. The UNAIDS country estimates are publicly available from the UNAIDS website: https://aidsinfo.unaids.org/.
Funding: OE was funded by Wellcome Trust (reference 222376/Z/21/Z) 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. JWE was supported by UNAIDS, the Bill & Melinda Gates Foundation (INV-006733). NAIIS was supported by PEPFAR through the CDC under the Cooperative Agreement #U2GGH002108 to the University of Maryland, Baltimore, and by the Global Fund to Fight AIDS, Tuberculosis and Malaria through the National Agency for the Control of AIDS, Nigeria, under the contract #NGA-H-NACA to the University of Maryland, Baltimore. The PHIAs were supported by PEPFAR through the CDC under the terms of cooperative agreement #U2GGH001226 and U2GGH000994 to ICAP Columbia. Funders had no role in the study design, data analysis, decision to publish or preparation of the manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Jeffrey W. Imai-Eaton has received grants/contracts from NIH and WHO, consulting fees from BAO Systems, support for attending meetings from UNAIDS, SACEMA and the International AIDS Society and is a member of the editorial board for PLOS Global Public Health. Olanrewaju Edun has received consulting fees from University of Cape Town and WHO and support for attending meetings from UNAIDS. All other authors have declared that no competing interests exist.
Introduction
Over the past decade, countries most-affected by HIV in sub-Saharan Africa (SSA) have aggressively expanded HIV testing and antiretroviral treatment (ART) as part of combination HIV prevention. These efforts have increased ART coverage among adults with HIV from 25% in 2010 to 80% in 2021, and contributed to a 56% reduction in AIDS deaths and 42% decline in new HIV infections in the region [1]. Despite this progress, an estimated 730,000 adults acquired HIV in 2021, representing 56% of the 1.3 million global new adult HIV infections [1].
Understanding the relative contributions of people living with HIV (PLHIV) subgroups to ongoing HIV transmission informs prevention interventions. ‘Test-and-treat’ strategies have focused on reaching PLHIV who are undiagnosed and untreated to reduce HIV transmission [2]. However, as ART coverage reaches high levels, PLHIV on ART with detectable viremia, including low-level viremia (LLV), defined as viral load (VL) >50 but ≤1,000 copies/mL [3], represent an increasing share of PLHIV. Low-level viremia among PLHIV on treatment can be due to poor ART adherence, poor drug absorption, or drug resistance [4]. Compared to those with undetectable VL (≤50 copies/mL), PLHIV with LLV have increased risk of virological failure, acquiring drug resistance mutations, and onward HIV transmission [5–7].
Sexual HIV transmission risk correlates directly to the HIV VL in the HIV infected partner [8–10], with those with undetectable VL being non-infectious [11]. While observational data quantifying HIV infectiousness at low viraemia levels are limited, some have raised concern that PLHIV with LLV could become an increasingly important population with HIV transmission risk as this population grows and the undiagnosed and untreated population shrinks [5, 12]. This could particularly be the case if PLHIV with LLV engage in HIV risk behaviour due to a perception of being non-infectious. Studies have consistently found lower reported sexual risk among those aware of their HIV-positive status [13–16] and on ART [17, 18] compared to undiagnosed or untreated PLHIV. However, data on the association between HIV risk behaviour and virologic non-suppression (>1,000 copies/mL) or LLV are sparse, with disparate findings and limited data from SSA settings [19–23].
Since the 2000s, national household health surveys with HIV serological testing have provided evidence about the relationship between HIV status, HIV testing, and sexual risk [24]. The Population-based HIV Impact Assessment (PHIA) surveys, conducted in high HIV burden SSA countries since 2015, additionally measured VL and ART usage among PLHIV [25]. These surveys provide unique data about the determinants of HIV transmission—sexual risk behaviours and distribution of VL—across stages of the HIV testing and treatment cascade following implementation of universal ART eligibility. To guide understanding of potential HIV transmission risk following rapid ART scale-up and guide future HIV programme priorities, we (i) assessed the association between HIV sexual risk behaviour and subgroups of PLHIV defined by HIV awareness, treatment status, and virologic status; (ii) characterised the distribution of viremia and consequent risk of transmission among subgroups; and (iii) quantified the expected proportion of HIV transmission across subgroups over time.
Methods
Study design and participants
We analysed data from 14 PHIA surveys conducted during 2015–2019 in West/Central Africa: Cameroon (June 2017–February 2018), Côte d’Ivoire (August 2017–March 2018), Nigeria (July–December 2018); Eastern Africa: Ethiopia (October 2017–April 2018), Kenya (May 2018–March 2019), Malawi (November 2015–August 2016), Rwanda (October 2018–March 2019), Uganda (August 2016–March 2017), Tanzania (October 2016–June 2017), Zambia (March–August 2016); and Southern Africa: Eswatini (August 2016–March 2017), Lesotho (November 2016–May 2017), Namibia (June–December 2017), Zimbabwe (October 2015–August 2016) [26, 27]. For the purpose of this study, the PHIA survey data were accessed in November 2021, except the Kenya PHIA data which were accessed in September 2022. PHIAs were cross-sectional household surveys that collected standardised data across countries and assessed HIV-related health indicators. Details of survey design, sampling, and procedures are described elsewhere [28, 29]. Briefly, surveys used a stratified multistage sampling design, with households randomly selected within sampled enumeration areas. Adult household members were eligible for the survey if they slept in the household the night before the interview and were aged 15–64 years, except in Zambia and Lesotho (15–59 years), and in Tanzania, Eswatini, and Zimbabwe (≥15 years). Consenting participants completed a structured individual questionnaire, including sociodemographic information, marital and pregnancy status, and HIV testing, care, and treatment history. Participants’ also self-reported their last three sexual relationships in the past 12-months. Following interviews, HIV status was determined through home-based serological testing using country-specific HIV diagnostic algorithms with immediate return of results. HIV-positive blood samples underwent laboratory-based confirmatory testing using Geenius HIV-1/2 Confirmatory Assay (Bio-Rad, Hercules, CA), HIV RNA VL testing, and qualitative screening for selected antiretrovirals (ARVs) to determine ART usage. PHIA biomarker testing methodology has previously been described [28, 29].
All surveys received ethical approval from institutional review boards in the respective countries, the U.S. Centers for Disease Control and Prevention (CDC), and Columbia University or the University of Maryland at Baltimore (Nigeria survey). Written informed consent was obtained from adult participants; assent and parental permission was obtained for participants under the age of consent (typically 15–17 years). Secondary analysis reported here received ethical approval from the Imperial College Research Governance and Integrity Team (20IC6451).
Study measures
Table 1 describes the sexual risk outcomes analysed. The primary risk outcome, “HIV high-risk behaviour”, was defined as self-reporting ≥2 sexual partners and condomless sex with a non-spousal partner in the past 12 months. Alternative measures were condomless sex with a casual partner, transactional partnership, condomless last sex (with any partner), and self-reporting ≥2 sexual partners with condomless last sex (with any partner).
The main exposure of interest was the population subgroup defined by HIV infection, diagnosis, and viremia category among those on treatment (Table 1). Participants were excluded if their VL was censored at a lower limit of quantification (LLOQ) within the LLV category (e.g., <LLOQ: 839 copies/mL) or had undetectable VL (≤50 copies/mL) but ARVs were not detected and they self-reported not taking ART (S2 Fig).
Covariates for regression analysis were age, education level, wealth quintile, marital status, pregnancy status, and urban/rural residence. For the Ethiopia survey, which was restricted to an urban-only sampling frame, the urban area size (small: <50,000; large: ≥50,000 people) was included. All other surveys were nationally representative.
Data analyses
Analysis was restricted to sexually active adults with results for HIV status, ART usage, and VL, and who reported sexual activity. All analyses were sex-stratified. We summarised participant characteristics for each survey using medians, interquartile ranges (IQR), and survey weighted proportions using the PHIA blood test weights (PHIA weights). We calculated weighted proportions with HIV risk behaviour and proportions of PLHIV in the defined cascade subgroups for each survey, with associated design-based 95% confidence intervals. We also summarised characteristics of the pooled sample (unweighted) from all 14 surveys.
Second, to identify the association between HIV high-risk behaviour and HIV subgroup, we pooled data from the 14 surveys and used generalized estimating equations (GEE) with log-binomial distribution and exchangeable correlation structure. The GEE model incorporated survey cluster random effects and country-level fixed effects. Regression models were sex-stratified and adjusted for age, wealth quintile, education level, urban/rural residence, marital status, and pregnancy status (for women), identified from our conceptual framework (S1 Fig). To allow a non-linear relationship between age and HIV risk behaviour, age was modelled using a cubic spline with knots at ages 30 and 50 (S9 Fig). Survey weights were not used in the regression model because the model adjusted for factors determining weights. In sensitivity analysis, we repeated the regression incorporating the normalized survey weights.
Third, to characterise the distribution of viremia among PLHIV subgroups, we calculated the mean log10 VL among the PLHIV subgroups and plotted kernel density estimates of the log10-transformed VL distribution for each survey. We used a Hill function parameterised by Fraser et al. [30] to model the annual HIV transmission rate in a discordant partnership as a function of VL (S11 Fig). The Hill function was applied to survey-weighted individual-level observations in each PLHIV subgroup to estimate average transmission rates by sex, country, and cascade subgroup.
Finally, we combined the determinants of HIV force of infection—the relative rate of partner change (using the relative self-reported sexual risk as a proxy), per-partnership risk of transmission (transmission probability from the Hill function), and size of each PLHIV subgroup—to estimate the proportion of transmission from each subgroup using the risk equation:
where pdi represents the proportion of PLHIV in the i th exposure category, PRi is the prevalence ratio of HIV high-risk behaviour for the i th exposure category relative to the unexposed category, and β(Vi) is the average transmission rate as a function of VL. Given the differential distribution of covariates adjusted for in the pooled model across the survey countries, we obtained country-specific prevalence ratios(PRi) by using the pooled model to predict the probability of self-reporting HIV high-risk sex for each PLHIV subgroup, accounting for the survey-weighted distribution of covariates in each group.
To explore how proportions changed over time as HIV awareness, ART coverage, and viral suppression increased, we obtained UNAIDS estimates of the proportion of PLHIV in each subgroup from 2010 to 2020 [1]. Assuming the proportion of LLV among PLHIV with VL <1,000 copies/mL and ‘HIV high-risk behaviour’ prevalence ratios were constant over time at values estimated from the surveys, we calculated the proportion of transmission attributable to each PLHIV subgroup from 2010 to 2020. Statistical significance level assessed as p <0.05. Analyses were conducted in R version 4.1.2.
Sensitivity analyses conducted includes repeating the analyses excluding the Nigeria survey, which contributed 38% of all participants, to explore the influence of the survey. We also stratified the analyses by West/Central, Eastern, and Southern Africa regions. We repeated analyses using alternative outcome measures of relative HIV transmission-risk behaviour (Table 1), and hypothetically assumed: (i) the same HIV high-risk behaviour across all groups (i.e., aPR = 1) and (ii) higher HIV high-risk behaviour among individuals on ART relative those not on ART (i.e., aPR = 1 among those on ART and aPR = 0.5 among untreated).
Lastly, to explore sensitivity to the Hill function for the relationship between viremia and infectiousness, we repeated analyses using a linear function estimated by Wilson et al. [31]. This function, which has been shown to overestimate the risk of transmission among people on ART [32], is likely an upper-bound risk of transmission among PLHIV on ART with LLV.
Results
Among 378,166 participants (≥15 years) in the 14 surveys who reported ever having had sex, our analyses included 368,373 (97%) participants with complete data (214,305 [58%] women; 154,068 [42%] men (Table 2 and S1 Table); see S2 Fig for details of participants excluded).
In all surveys, a higher proportion of men reported HIV high-risk behaviour (2.1–17.0% across surveys), condomless casual partnerships (1.1–21.6%), and multiple partnerships with condomless last sex (3.0–22.1%) than women (0.8–5.7%, 0.6–11.1%, and 1.1–5.3%, respectively). More women reported transactional partnerships (2.9–16.0%) than men (1.7–5.7%), with similar proportions reporting condomless last sex with any partner (S3 Fig).
In Western/Central Africa, undiagnosed individuals were the largest group of PLHIV. In Eastern and Southern African surveys, individuals on ART with undetectable VL were the largest proportion. The proportion of PLHIV with LLV varied across surveys from 5.0% (Malawi, 2015–2016) to 12.1% (Uganda, 2016–2017) among women, and 4.6% (Malawi, 2015–2016) to 18.8% (Ethiopia, 2017–2018) among men.
Association between HIV high-risk behaviour and HIV infection, diagnosis, and treatment population subgroups
Compared to women on ART with undetectable VL, undiagnosed women were more likely to report HIV high-risk behaviour (adjusted prevalence ratio [aPR]: 1.28, 95% confidence interval [95% CI]: 1.08–1.52), condomless casual partnership (aPR: 1.42; 1.23–1.63), condomless last sex (aPR: 1.42, 1.38–1.47) and multiple sexual partnership with condomless last sex (aPR: 1.80, 1.51–2.14). Women who were diagnosed but untreated were more likely to report condomless casual partnership (aPR: 1.79; 1.41–2.26), condomless last sex (aPR: 1.28, 1.20–1.36) and multiple sexual partnerships with condomless last sex (aPR: 1.49, 1.11–2.02; Fig 1; S2–S6 Tables).
Results are stratified by sex and models were adjusted for age, education level, wealth quintile, marital status, urban/rural dwelling or urbanicity size and pregnancy status in women. Note: x-axis is log10 scaled. Full regression results are reported in S2–S6 Tables. Country-specific predicted prevalence ratios are reported in S4–S8 Figs, S7–S11 Tables.
Similarly, compared to men with undetectable VL, undiagnosed men were more likely to report HIV high-risk behaviour (aPR: 1.61; 1.33–1.95), condomless casual partnership (aPR: 1.43; 1.19–1.73), condomless last sex (aPR: 1.67; 1.59–1.75) and multiple sexual partnership with condomless last sex (aPR: 2.04, 1.77–2.36). Men who were diagnosed but untreated were more likely to report HIV high-risk behaviour (aPR: 2.06; 1.52–2.78), condomless casual partnership (aPR: 1.80; 1.35–2.40), condomless last sex (aPR: 1.40; 1.29–1.53) and multiple sexual partnerships with condomless last sex (aPR: 1.91; 1.51–2.43).
Among both men and women, those with LLV did not report significantly different HIV high-risk behaviour, condomless casual partnerships, transactional partnerships, or multiple sexual partnerships with condomless last sex compared to those with undetectable VL. HIV negative women were less likely to report HIV high-risk behaviour (aPR: 0.60; 0.53–0.69), transactional partnerships (aPR: 0.68; 0.64–0.73), and multiple sexual partnerships with condomless last sex. Condomless last sex was more common among all untreated PLHIV subgroups and among HIV negative men and women than those with undetectable VL. Regression results were similar in the weighted and unweighted analysis (S14 Fig).
Distribution of viremia and transmission rates across PLHIV subgroups
Mean log10 VL was similar among those undiagnosed, untreated, and on treatment with non-suppressed VL (Fig 2), ranging from 4.05 log10 copies/mL (Malawi 2015–2016, undiagnosed women) to 5.32 log10 copies/mL (Côte d’Ivoire 2017–2018, diagnosed but untreated men). Among PLHIV on treatment with LLV, mean log10 VL ranged from 2.00 (IQR: 1.78, 2.03; Kenya 2018–2019, men) to 2.37 log10 copies/mL (IQR: 2.04, 2.64; Malawi 2015–2016, men).
Note: unable to reliably estimate distribution due to small number of men diagnosed but untreated in the Ethiopia PHIA survey.
Across all countries, the predicted mean transmission rate using the Hill function was lowest among PLHIV on treatment with LLV, ranging between 0.003 to 0.007 per year. Transmission rates were similar across the other categories (undiagnosed: 0.15 to 0.24 per year; diagnosed but untreated: 0.16 to 0.24 per year; those on treatment with viral non-suppression: 0.14 to 0.26 per year) reflecting the similarity of the VL distributions (S10 and S11 Figs, S12 Table).
Contribution of PLHIV subgroups to HIV transmission
In most surveys, the majority of transmission was estimated to be among undiagnosed PLHIV, accounting for the prevalence ratio of HIV high-risk behaviour, distribution of viremia, and relative size of each PLHIV subgroup. The proportion of transmission from this group ranged from 52% (Eswatini, 2016–2017, women) to 92% (Ethiopia, 2017–2018, men; Fig 3). The subgroup with the second largest proportion of transmission varied between PLHIV diagnosed but untreated and those on ART but non-suppressed. Individuals on ART with LLV had the smallest contribution to transmission, ranging between 0.1% (Nigeria, 2018, men) to 0.9% (Namibia, 2017, women; Fig 3).
Estimated using the country-specific prevalence ratio of HIV high-risk behaviour, proportion size, and transmission rate for each PLHIV subgroup.
Using UNAIDS estimates for the proportion of PLHIV in each subgroup from 2010 to 2020, we estimated that, in 2010, undiagnosed women and men accounted for the majority of transmission (47%–94%) in 9/14 countries (Cote d’Ivoire, Nigeria, Cameroon, Tanzania, Zambia, Ethiopia, Namibia, Lesotho and Zimbabwe), and women and men diagnosed but untreated had the second largest contribution (Fig 4 and S12 Fig). In the remaining countries, diagnosed but untreated PLHIV represented the largest transmission proportion (Kenya, Uganda, Malawi, Rwanda, Eswatini). The estimated transmission proportion from PLHIV on ART but non-suppressed (1%–5%) and with LLV (0.03%–0.2%) was small.
Countries chosen as illustrative examples as their transmission proportion patterns are similar to other countries in their regions. See S12 Fig for all countries in analysis.
In 2020, the size of the undiagnosed and untreated population reduced and their estimated proportion of tranmsission also reduced since 2010, but, in most countries, the undiagnosed population still accounted for the majority of transmission (17%–75%). However, by 2020 men and women on ART but non-suppressed were the second largest transmission population (5%–64%). In Tanzania, Zambia, Ethiopia, Kenya, Malawi, and Lesotho, men and women undiagnosed and those diagnosed but untreated remained the two largest transmission groups, followed by those on ART but non-suppressed. Transmission from PLHIV on ART with LLV in 2020 remained low (0.3%–2.4%).
In sensitivity analyses, results were similar when we excluded the Nigeria survey, stratified the analysis by region, explored other measures of sexual risk behaviour, assumed equal sexual risk in all groups, higher risk among those on ART vs untreated and used a linear function to represent the relationship between viremia and transmission risk (S13 and S15–S25 Figs). However, when assuming higher levels of sexual risk among PLHIV on ART compared to those not on ART, those undiagnosed and diagnosed but untreated still account for majority of transmission across all countries in our analysis (27%–80% and 3%–32%, respectively), except among women in Ethiopia (2017–2018) and Lesotho (2016–2017), and men in Rwanda (2018–2019), where those on ART but non-suppressed accounted for the largest proportion of estimated transmission (54%, 43% and 48% respectively). Estimated transmission among those with LLV was still low and ranged between 0.3%–1.8% (S22 Fig).
Discussion
In national HIV surveys with delineation of PLHIV across the HIV care cascade, undiagnosed PLHIV and those diagnosed but untreated were more likely to report risk behaviours that could result in HIV transmission. The distribution of viremia and, consequently, estimated transmission rates were remarkably similar between undiagnosed, diagnosed but untreated, and PLHIV on ART with viral non-suppression. Taken together, we estimated that the majority of transmission continues to be from undiagnosed PLHIV and those diagnosed but untreated, reflecting both higher VL and sexual risk behaviour. Even as ART coverage reaches high levels, timely diagnosis, linkage to care, retention on ART, and effective re-engagement following interruption remain priorities to prevent HIV transmisions.
We estimated that transmission from people with LLV constituted <1% of all new HIV infections. Treated PLHIV with undetectable VL reported the lowest levels of risk behaviour, and we found no evidence of increased risk among those on ART with LLV or virally non-suppressed. This contrasts studies from South Africa and the US where higher rates of HIV risk behaviour among non-suppressed PLHIV on ART were reported, though the authors acknowledged the potential role of unmeasured confounding [15, 22]. Our findings suggest that individuals on ART with LLV or non-suppression are not more likely to engage in HIV high-risk behaviour, which may reflect effective HIV prevention counselling provided during clinical care [33]. Sex and regional differences in the prevalence of HIV risk behaviours were similar to findings from other studies in SSA [15, 24].
Our assumption about infectiousness at LLV was derived by extrapolating the Hill function parameterised by Fraser et al. [30] to VL ≤1000 copies/mL, a range in which no transmission was observed in the data used to estimate the function [9]. A new systematic review of linked HIV transmissions did not identify any confirmed transmission events in which the infectious partner had VL ≤1000 copies/mL [34]. Therefore, our assumptions about transmission at LLV are likely conservative, and likely represents an upper bound on the proportion of transmissions. This finding alleviates concerns that people with LLV may become an important transmission sub-population in settings with high treatment coverage and demonstrates the population-level implications of “Undetectable = Untransmittable”. However, studies have found PLHIV with LLV have a high risk of subsequent viral non-suppression [7, 12], which could result in HIV transmission risk and adverse individual health outcomes.
That most transmissions continue to arise from undiagnosed and untreated subgroups was consistent with a recent mathematical modelling study from South Africa, which estimated that, in 2019, 30% of sexually acquired HIV transmissions were from undiagnosed individuals, 60% from diagnosed but untreated, and 10% from non-suppressed persons on ART [35]. The similar viral load distribution between non-suppressed PLHIV on ART and untreated PLHIV was notable. This suggests similar biologic risk of onward transmission among non-suppressed PLHIV on ART as those who are untreated, highlighting the importance of ensuring that individuals achieve and sustain viral suppression. Following large reductions in undiagnosed and untreated PLHIV over the past decade, the contribution to transmission from non-suppressed PLHIV on ART was estimated to have increased (between 1%-5% in 2010 to 5%-65% in 2020). Ongoing adherence support, frequent virologic monitoring, early interventions for viremia, and timely regimen switching among PLHIV on ART are increasingly important in this group. Optimized treatment regimens that increase viral suppression among PLHIV on ART, particularly recent transition to dolutegravir-based regimes [36, 37], will help curtail this source of transmission. The variations in the transmission trends between 2010 and 2020 across countries in our analysis were largely due to differences in distribution across the undiagnosed, diagnosed but untreated, viral non-suppression cascade stages. Since 2020, several countries have repeated PHIA surveys, which have confirmed the increases in ART coverage represented in UNAIDS modelled estimates used here and demonstrated increased viral suppression among persons on ART [38–43].
Strengths of this study included the combination of data on HIV risk behaviour and infectiousness across HIV cascade stages from nationally represntative samples acros multiple countries in SSA. Availability of ARV and VL testing enabled the classification PLHIV by treatment and viremia status. Our study also has limitations. In addition to those identified above, self-reported measures of sexual behaviour are subject to recall and social desirability bias, and we cannot exclude the possibility that ART and its counselling reduced risk reporting more than actual risk taking (and may underestimate relative sexual risk among those on ART). This may have biased our estimates of the prevalence ratios of sexual risk behaviours, however, overall findings about priority transmission subgroups were relatively robust in sensitivity analyses assuming smaller differentials in sexual risk. Second, because the PHIA surveys are cross-sectional, our study is unable to identify changes in sexual behaviours over time and there is a need for further research to understand how these have evolved. Third, we were unable to distinguish between causes for non-suppression while on ART (non-adherence, treatment failure or drug resistance), nor reasons for being untreated (recent diagnosis or interrupting treatment) among those diagnosed but untreated because responses to survey questions about treatment interruptions were redacted from some PHIA survey datasets, and UNAIDS estimates did not stratify the diagnosed but untreated group according to never or previously treated. In the era of “test-and-start”, we expect that persons interrupting treatment account for most transmissions among the diagnosed but not currently on ART population, consistent with findings of the recent mathematical modelling analysis for South Africa [35]. Finally, our findings may not be representative of all countries in the regions included in our analysis due to differences in their HIV epidemiology, healthcare systems and cultural factors. Notably, our results for the Southern African region do not include the national HIV household survey in South Africa [44], which has the largest HIV epidemic in the region. The survey was implemented using different questionnaires than PHIA surveys analysed here and variables could not be sufficiently harmonised for inclusion in pooled analyses.
In conclusion, as HIV treatment coverage reaches high levels in high HIV burden settings in sub-Saharan Africa, HIV programmes require up-to-date information about HIV risk behaviour and transmission risk among subgroups to guide prevention priorities. We estimated that undiagnosed PLHIV and those diagnosed but untreated continue to account for most HIV transmission due to higher levels of both sexual risk behaviour and viral load. Reaching undiagnosed and untreated PLHIV with HIV testing, ART linkage, retention, and treatment support services to attain durable virologic suppression remains the priority to reduce HIV transmission and end the HIV epidemic. As ART coverage increases, individuals on treatment with virologic non-suppression may increasingly contribute to onward transmission and become a high-priority subgroup. Individuals on ART with LLV do not contribute substantially to HIV transmission, but may indicate future transmission risk if LLV predicts future treatment failure and higher viremia.
Supporting information
S1 Table. Summary of participant characteristics across all included PHIA surveys by sex.
https://doi.org/10.1371/journal.pgph.0003030.s001
(DOCX)
S2 Table. Crude and adjusted prevalence ratios of self-reporting HIV high-risk behaviour by sex.
https://doi.org/10.1371/journal.pgph.0003030.s002
(DOCX)
S3 Table. Crude and adjusted prevalence ratios of self-reporting condomless casual partnership by sex.
https://doi.org/10.1371/journal.pgph.0003030.s003
(DOCX)
S4 Table. Crude and adjusted prevalence ratios of self-reporting transactional partnership by sex.
https://doi.org/10.1371/journal.pgph.0003030.s004
(DOCX)
S5 Table. Crude and adjusted prevalence ratios of self-reporting condomless last sex (with any partner) by sex.
https://doi.org/10.1371/journal.pgph.0003030.s005
(DOCX)
S6 Table. Crude and adjusted prevalence ratios of self-reporting both multiple sexual partnership and condomless last sex (with any partner) by sex.
https://doi.org/10.1371/journal.pgph.0003030.s006
(DOCX)
S7 Table. Predicted prevalence ratios of self-reporting high-risk sex relative to PLHIV On ART undetectable (≤50 copies/mL) for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s007
(DOCX)
S8 Table. Predicted prevalence ratios of self-reporting condomless casual partnership for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s008
(DOCX)
S9 Table. Predicted prevalence ratios of self-reporting transactional partnership for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s009
(DOCX)
S10 Table. Predicted prevalence ratios of self-reporting condomless last sex (with any partner) for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s010
(DOCX)
S11 Table. Predicted prevalence ratios of self-reporting both multiple sexual partnership and condomless last sex for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s011
(DOCX)
S12 Table. Mean log10 viral load and transmission rate per year for all PLHIV subgroups by sex.
https://doi.org/10.1371/journal.pgph.0003030.s012
(DOCX)
S2 Fig. Flowchart of participants included in study.
https://doi.org/10.1371/journal.pgph.0003030.s014
(DOCX)
S3 Fig. Weighted prevalence and 95% confidence intervals of self-reported HIV high-risk behaviour, condomless casual partnerships, transactional partnerships, condomless last sex and both multiple sexual partnership and condomless last sex by sex across 14 PHIA surveys.
https://doi.org/10.1371/journal.pgph.0003030.s015
(DOCX)
S4 Fig. Predicted prevalence ratios reporting high-risk sex for each cascade group relative to HIV positive on ART and undetectable (≤50 copies/mL) adults.
https://doi.org/10.1371/journal.pgph.0003030.s016
(DOCX)
S5 Fig. Forest plots showing the predicted prevalence ratios and 95% confidence intervals of self-reporting condomless casual partnership for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s017
(DOCX)
S6 Fig. Forest plots showing the predicted prevalence ratios and 95% confidence intervals of self-reporting transactional partnership for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s018
(DOCX)
S7 Fig. Forest plots showing the predicted prevalence ratios and 95% confidence intervals of self-reporting condomless last sex for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s019
(DOCX)
S8 Fig. Forest plots showing the predicted prevalence ratios and 95% confidence intervals of self-reporting both multiple sexual partnership and condomless last sex for each of the 14 survey countries by sex.
https://doi.org/10.1371/journal.pgph.0003030.s020
(DOCX)
S9 Fig. Estimated log-odds ratio and 95% confidence intervals of HIV high-risk behaviour, condomless casual partnership, transactional partnership, condomless last sex and self-reporting both multiple sexual partnership and condomless last sex by age predicted by the spline effects by sex and risk behaviour.
https://doi.org/10.1371/journal.pgph.0003030.s021
(DOCX)
S10 Fig.
(A) Boxplots of the log10 viral load distribution and (B) Mean annual transmission rates and 95% confidence intervals for each PLHIV population sub-group estimated using the Hill function by sex.
https://doi.org/10.1371/journal.pgph.0003030.s022
(DOCX)
S11 Fig.
(A) Plot showing the transmission rate as a function of viral load using the Hill function. (B) Distribution of viral load for each PLHIV subgroup using data from Lesotho 2016–2017 survey.
https://doi.org/10.1371/journal.pgph.0003030.s023
(DOCX)
S12 Fig. Estimated proportion of PLHIV in each sub-group and proportion of transmission attributed to each PLHIV sub-group from 2010–2020.
https://doi.org/10.1371/journal.pgph.0003030.s024
(DOCX)
S13 Fig. Forest plots showing the adjusted prevalence ratios of self-reporting high HIV high-risk behaviour by sub-groups (sensitivity analyses excluding the Nigeria 2018 survey).
https://doi.org/10.1371/journal.pgph.0003030.s025
(DOCX)
S14 Fig. Forest plots showing the adjusted prevalence ratios of self-reporting high HIV high-risk behaviour by sub-groups (including normalized survey weight in regression mode).
https://doi.org/10.1371/journal.pgph.0003030.s026
(DOCX)
S15 Fig. Forest plots showing the region-specific adjusted prevalence ratios of self-reporting HIV high-risk behaviour, condomless casual partnerships, transactional partnerships, condomless last sex and self-reporting both multiple sexual partnership and condomless last sex for each subgroup.
https://doi.org/10.1371/journal.pgph.0003030.s027
(DOCX)
S16 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and region-specific adjusted prevalence ratio of self-reporting HIV high-risk behaviour in transmission equation).
https://doi.org/10.1371/journal.pgph.0003030.s028
(DOCX)
S17 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and adjusted prevalence ratio of self-reporting condomless casual partnerships in transmission equation).
https://doi.org/10.1371/journal.pgph.0003030.s029
(DOCX)
S18 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and adjusted prevalence ratio of self-reporting transactional partnerships in transmission equation).
https://doi.org/10.1371/journal.pgph.0003030.s030
(DOCX)
S19 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and adjusted prevalence ratio of self-reporting condomless last sex (with any partner) in transmission equation).
https://doi.org/10.1371/journal.pgph.0003030.s031
(DOCX)
S20 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and adjusted prevalence ratio of self-reporting both multiple sexual partnership and condomless last sex in transmission equation).
https://doi.org/10.1371/journal.pgph.0003030.s032
(DOCX)
S21 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and assuming the same adjusted prevalence ratio of self-reporting HIV high-risk behaviour for all subgroups).
https://doi.org/10.1371/journal.pgph.0003030.s033
(DOCX)
S22 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using Hill function of transmission and viremia relationship, and assuming the adjusted prevalence ratio of self-reporting HIV high-risk behaviour among untreated PLHIV was half that among PLHIV on ART).
https://doi.org/10.1371/journal.pgph.0003030.s034
(DOCX)
S23 Fig.
Plot showing the transmission rate as a function of viral load using the linear function for (A) Women-to-Men and (B) Men-to-Women, and distribution of viral load for each PLHIV subgroup using data from Lesotho 2016–2017 survey for (C) women and (D) men.
https://doi.org/10.1371/journal.pgph.0003030.s035
(DOCX)
S24 Fig. Sex-stratified weighted average transmission rates per sexual contact and 95% confidence intervals for each PLHIV population sub-group estimated using the linear function for each of the 14 PHIA surveys.
https://doi.org/10.1371/journal.pgph.0003030.s036
(DOCX)
S25 Fig. Proportion of transmission attributed to PLHIV sub-group by sex across all 14 PHIA surveys (sensitivity analysis using linear function of transmission and viremia relationship, and adjusted prevalence ratio of self-reporting HIV high-risk behaviour in transmission equation.
https://doi.org/10.1371/journal.pgph.0003030.s037
(DOCX)
References
- 1.
Joint United Nations Programme on HIV/AIDS (UNAIDS). HIV estimates with uncertainty bounds 1990-Present. 2023 [cited 25 May 2022]. https://www.unaids.org/en/resources/documents/2021/HIV_estimates_with_uncertainty_bounds_1990-present
- 2.
World Health Organization. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: recommendations for a public health approach, 2nd ed. 2016 [cited 12 Jun 2023]. https://apps.who.int/iris/handle/10665/208825
- 3.
World Health Organization. Updated recommendations on HIV prevention, infant diagnosis, antiretroviral initiation and monitoring. 2021 [cited 17 Jul 2021]. https://www.who.int/publications/i/item/9789240022232
- 4. Ryscavage P, Kelly S, Li JZ, Harrigan PR, Taiwo B. Significance and clinical management of persistent low-level viremia and very-low-level viremia in HIV-1-infected patients. Antimicrob Agents Chemother. 2014;58: 3585–3598. pmid:24733471
- 5. Neely MN, Benning L, Xu J, Strickler HD, Greenblatt RM, Minkoff H, et al. Cervical shedding of HIV-1 RNA among women with low levels of viremia while receiving highly active antiretroviral therapy. J Acquir Immune Defic Syndr. 2007;44: 38–42. pmid:17106279
- 6. Bareng OT, Moyo S, Zahralban-Steele M, Maruapula D, Ditlhako T, Mokaleng B, et al. HIV-1 drug resistance mutations among individuals with low-level viraemia while taking combination ART in Botswana. J Antimicrob Chemother. 2022; dkac056. pmid:35229102
- 7. Esber A, Polyak C, Kiweewa F, Maswai J, Owuoth J, Maganga L, et al. Persistent Low-level Viremia Predicts Subsequent Virologic Failure: Is It Time to Change the Third 90? Clin Infect Dis. 2019;69: 805–812. pmid:30462188
- 8. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al. Viral Load and Heterosexual Transmission of Human Immunodeficiency Virus Type 1. New England Journal of Medicine. 2000;342. pmid:10738050
- 9. Fideli ÜS, Allen SA, Musonda R, Trask S, Hahn BH, Weiss H, et al. Virologic and immunologic determinants of heterosexual transmission of human immunodeficiency virus type 1 in Africa. AIDS Res Hum Retroviruses. 2001;17. pmid:11461676
- 10. Hughes JP, Baeten JM, Lingappa JR, Magaret AS, Wald A, de Bruyn G, et al. Determinants of per-coital-act HIV-1 infectivity among African HIV-1-serodiscordant couples. J Infect Dis. 2012;205: 358–365. pmid:22241800
- 11. Eisinger RW, Dieffenbach CW, Fauci AS. HIV viral load and transmissibility of HIV infection undetectable equals untransmittable. JAMA. 2019. pmid:30629090
- 12. Chun HM, Abutu A, Milligan K, Ehoche A, Shiraishi RW, Odafe S, et al. Low-level viraemia among people living with HIV in Nigeria: a retrospective longitudinal cohort study. Lancet Glob Health. 2022. pmid:36400087
- 13. Hakim AJ, Tippett Barr BA, Kinchen S, Musuka G, Manjengwa J, Munyati S, et al. Progress Toward the 90-90-90 HIV Targets in Zimbabwe and Identifying Those Left Behind. J Acquir Immune Defic Syndr. 2021;88. pmid:34321414
- 14. Gregson S, Nyamukapa C. Did sexual behaviour differences between HIV infection and treatment groups offset the preventative biological effects of ART roll-out in Zimbabwe? Popul Stud. 2021;75. pmid:33559537
- 15. Huerga H, Venables E, Ben-Farhat J, Van Cutsem G, Ellman T, Kenyon C. Higher risk sexual behaviour is associated with unawareness of HIV-positivity and lack of viral suppression—Implications for Treatment as Prevention. Sci Rep. 2017;7. pmid:29170407
- 16. Lulseged S, Melaku Z, Habteselassie A, West CA, Gelibo T, Belete W, et al. Progress towards controlling the HIV epidemic in urban Ethiopia: Findings from the 2017–2018 Ethiopia population-based HIV impact assessment survey. PLoS One. 2022;17: e0264441-. pmid:35213668
- 17. Nakiganda LJ, Nakigozi G, Kagaayi J, Nalugoda F, Serwadda D, Sewankambo N, et al. Cross-sectional comparative study of risky sexual behaviours among HIV-infected persons initiated and waiting to start antiretroviral therapy in rural Rakai, Uganda. BMJ Open. 2017;7. pmid:28893749
- 18. Risher K, Rehle T, Simbayi L, Shisana O, Celentano DD. Antiretroviral Treatment and Sexual Risk Behavior in South Africa. AIDS Behav. 2016;20. pmid:26194426
- 19. Guigayoma J, Chen YH, Snowden JM, Santos GM, Hecht J, Raymond HF. Self-Perceived Viral Load and Sexual Risk Behavior among Known HIV-Positive MSM in San Francisco, 2014. J Acquir Immune Defic Syndr (1988). 2017;75. pmid:28604432
- 20. Stolte IG, De Wit JBF, Van Eeclen A, Coutinho RA, Dukers NHTM. Perceived viral load, but not actual HIV-1-RNA load, is associated with sexual risk behaviour among HIV-infected homosexual men. AIDS. 2004;18. pmid:15353980
- 21. Kalichman SC, Rompa D, Austin J, Luke W, DiFonzo K. Viral load, perceived infectivity, and unprotected intercourse [1]. J Acquir Immune Defic Syndr. 2001. pmid:11694841
- 22. Mattson CL, Freedman M, Fagan JL, Frazier EL, Beer L, Huang P, et al. Sexual risk behaviour and viral suppression among HIV-infected adults receiving medical care in the United States. AIDS. 2014;28. pmid:25000558
- 23. Pan Y, Metsch LR, Gooden LK, Philbin MM, Daar ES, Douaihy A, et al. Viral suppression and HIV transmission behaviors among hospitalized patients living with HIV. Int J STD AIDS. 2019;30. pmid:31159715
- 24. Schaefer R, Gregson S, Benedikt C. Widespread changes in sexual behaviour in eastern and southern Africa: Challenges to achieving global HIV targets? Longitudinal analyses of nationally representative surveys. J Int AIDS Soc. 2019;22. pmid:31225953
- 25. Porter L, Bello G, Nkambule R, Justman J. HIV General Population Surveys: Shedding Light on the Status of HIV Epidemics and Informing Future Actions. J Acquir Immune Defic Syndr (1988). 2021;87. pmid:34166307
- 26.
National Bureau of Statistics—Nigeria. Nigeria HIV-AIDS Indicator and Impact Survey (NAIIS). 2018 [cited 11 Apr 2022]. https://www.nigerianstat.gov.ng/nada/index.php/catalog/65?msclkid=94fe957fb99711ec971383c70b363679
- 27.
ICAP at Columbia University. PHIA Project: Population-based HIV Impact Assessment. 2021 [cited 11 Apr 2022]. https://phia.icap.columbia.edu/
- 28. Sachathep K, Radin E, Hladik W, Hakim A, Saito S, Burnett J, et al. Population-Based HIV Impact Assessments Survey Methods, Response, and Quality in Zimbabwe, Malawi, and Zambia. J Acquir Immune Defic Syndr. 2021;87. pmid:34166308
- 29. Patel HK, Duong YT, Birhanu S, Dobbs T, Lupoli K, Moore C, et al. A Comprehensive Approach to Assuring Quality of Laboratory Testing in HIV Surveys: Lessons Learned From the Population-Based HIV Impact Assessment Project. J Acquir Immune Defic Syndr. 2021;87. pmid:34166309
- 30. Fraser C, Hollingsworth TD, Chapman R, De Wolf F, Hanage WP. Variation in HIV-1 set-point viral load: Epidemiological analysis and an evolutionary hypothesis. PNAS. 2007;104. pmid:17954909
- 31. Wilson DP, Law MG, Grulich AE, Cooper DA, Kaldor JM. Relation between HIV viral load and infectiousness: a model-based analysis. Lancet. 2008;372. pmid:18657710
- 32. Supervie V, Breban R. Per sex-act risk of HIV transmission under antiretroviral treatment: A data-driven approach. J Acquir Immune Defic Syndr. 2018;79. pmid:30179982
- 33. Fisher JD, Cornman DH, Shuper PA, Christie S, Pillay S, Macdonald S, et al. HIV prevention counseling intervention delivered during routine clinical care reduces HIV risk behavior in HIV-infected South Africans receiving antiretroviral therapy: The izindlela zokuphila/options for health randomized trial. J Acquir Immune Defic Syndr. 2014;67. pmid:25230288
- 34. Broyles L, Luo R, Boeras D, Vojnov L. The risk of sexual transmission of HIV in individuals with low-level HIV viraemia: a systematic review. Lancet. 2023. Available: https://doi.org/10.1016/S0140-6736(23)00877-2. pmid:37490935
- 35. Johnson LF, Meyer-Rath G, Dorrington RE, Puren A, Seathlodi T, Zuma K, et al. The Effect of HIV Programs in South Africa on National HIV Incidence Trends, 2000–2019. J Acquir Immune Defic Syndr (1988). 2022;90. pmid:35125471
- 36. Esber A, Dear N, Shah N, Kibuuka H, Maswai J, Owuoth J, et al. Brief Report: Virologic Impact of the Dolutegravir Transition: Prospective Results from the Multinational African Cohort Study. J Acquir Immune Defic Syndr (1988). 2022;91. pmid:35980350
- 37. Schramm B, Temfack E, Descamps D, Nicholas S, Peytavin G, Bitilinyu-Bangoh JE, et al. Viral suppression and HIV-1 drug resistance 1 year after pragmatic transitioning to dolutegravir first-line therapy in Malawi: a prospective cohort study. Lancet HIV. 2022;9. pmid:35905753
- 38.
Malawi Ministry of Health. Malawi Population-based HIV Impact Assessment MPHIA 2020–2021: Summary Sheet. 2022 [cited 25 Sep 2023]. https://phia.icap.columbia.edu/wp-content/uploads/2022/03/110322_MPHIA_Summary-sheet-English.pdf
- 39.
Zambia Statistic Agency. 89-98-96 Zambia on track to meet all the UNAIDS 95-95-95 after exceeding HIV Treatment and Viral Suppression targets. 2022 [cited 25 Sep 2023]. https://www.zamstats.gov.zm/89-98-96-zambia-on-track-to-meet-all-the-unaids-95-95-95-after-exceeding-hiv-treatment-and-viral-suppression-targets/
- 40.
Eswatini Ministry of Health and Central Statistical Office. Eswatini Population-Based HIV Impact Assessment SHIMS3 2021. 2022 [cited 25 Sep 2023]. https://phia.icap.columbia.edu/wp-content/uploads/2022/12/53059_14_SHIMS3_Summary-sheet-Web.pdf
- 41.
Uganda Ministry of Health. Uganda Population-Based HIV Impact Assessment UPHIA 2020–2021: Summary Sheet. 2022 [cited 25 Sep 2023]. https://phia.icap.columbia.edu/wp-content/uploads/2022/08/UPHIA-Summary-Sheet-2020.pdf
- 42.
Lesotho Ministry of Health LB of S. Lesotho Population-Based HIV Assessment LePHIA 2020: Summary Sheet. 2020 [cited 25 Sep 2023]. https://phia.icap.columbia.edu/wp-content/uploads/2021/08/53059_14_LePHIA_Summary-sheet_with-coat-of-arms_WEB_v2.pdf
- 43.
Zimbabwe Ministry of Health and Child Care. Zimbabwe Population-Based HIV Impact Assessment ZIMPHIA 2020: Summary Sheet. 2020 [cited 25 Sep 2023]. https://phia.icap.columbia.edu/wp-content/uploads/2020/11/ZIMPHIA-2020-Summary-Sheet_Web.pdf
- 44.
Simbayi L, Zuma K, Zungu N. South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2017. HSRC Press; 2017.