Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Mar 1.
Published in final edited form as: J Acquir Immune Defic Syndr. 2024 Mar 1;95(3):222–230. doi: 10.1097/QAI.0000000000003356

Polysubstance Use Profiles and HIV Viremia in a South African Cohort of Female Sex Workers: A Latent Class Analysis

Joseph G Rosen 1,2,*, Justin R Knox 3,4,5, Katherine B Rucinski 1,2, Mfezi Mcingana 6, Ntambue Mulumba 7, Carly A Comins 2, Lillian Shipp 2, Siyanda Makama 7, S Wilson Beckham 8, Harry Hausler 4,9, Stefan D Baral 2, Sheree R Schwartz 2
PMCID: PMC10922222  NIHMSID: NIHMS1947126  PMID: 38032752

Abstract

Background:

Given intersecting social and structural factors, female sex workers (FSW) exhibit elevated risk of HIV and substance use. However, there is limited study of how distinct substance use typologies influence HIV treatment outcomes among FSW.

Setting:

A cross-sectional survey with objective viral load assessments of 1,391 FSW enrolled into a treatment optimization-focused trial in Durban, South Africa (2018–2020).

Methods:

We used latent class analysis to uncover discrete patterns in past-month self-reported use of the following substances: heavy alcohol use, cannabis, cocaine, crack, ecstasy, methamphetamine, heroin, and whoonga. We used Wald tests to identify multilevel predictors of latent class membership and multivariable mixture modeling to quantify associations of substance use classes with HIV viremia (>50 RNA copies/mL).

Results:

Substance use (87%) and HIV viremia (62%) were highly prevalent. LCA uncovered three polysubstance use profiles: Heavy Alcohol Use Only (~54%); Cannabis, Heavy Alcohol, & Crack Use (~28%); and Whoonga & Crack Use (~18%). Whoonga & Crack Use was associated with social and structural adversities, including homelessness, outdoor/public sex work, HIV stigma, and violence. Relative to Heavy Alcohol Use Only, HIV viremia was significantly higher in the Whoonga & Crack Use class (adjusted odds ratio [adjOR] 1.97, 95% confidence interval [95%CI] 1.13–3.43) but not in the Cannabis, Heavy Alcohol, & Crack Use class (adjOR 1.17, 95%CI 0.74–1.86).

Conclusion:

HIV viremia differed significantly across identified polysubstance use profiles among South African FSW. Integrating drug treatment and harm reduction services into HIV treatment programs is key to improving virologic outcomes in marginalized communities.

Keywords: sex work, people living with HIV, viral load suppression, substance use, HIV treatment, South Africa

Introduction

South Africa, where one in five adults are estimated to be living with HIV, has made major strides towards HIV epidemic control, with over 5.5 million people initiated onto lifelong antiretroviral therapy (ART) in 2021.1 Over 40% of new HIV infections in South Africa are estimated to occur in key populations, including female sex workers (FSW) and their sexual partners.2 Globally, HIV incidence is disproportionately elevated among FSW,3 who experience poorer HIV treatment outcomes relative to other populations. One study in Port Elizabeth, South Africa, for example, reported 39% ART coverage among FSW living with HIV.4 To deliver HIV services equitably and accelerate momentum towards HIV epidemic control, HIV treatment services must address the overlapping individual and structural drivers of care disengagement and ART non-adherence, which are disproportionately shouldered by FSW.

Substance use is associated with HIV service disengagement and has been linked to poor treatment outcomes, including viremia, in various populations.5–7 Studies have characterized elevated substance use, particularly polydrug use, among FSW as a coping mechanism to the syndemics of violence, psychological distress, and poverty.8,9 FSW in South Africa reporting violence victimization histories and symptoms of post-traumatic stress disorder, for instance, are more likely to report higher-risk substance use.10 Manifestations of financial instability, from food insecurity to homelessness, have also been linked to elevated substance use among women who sell sex.11–13 The confluence of these forces can disrupt HIV self-management practices (i.e., adherence to daily ART regimens) and HIV care continuity (i.e., missed appointments, treatment attrition) for people living with HIV.14–16

Less, however, is known about the clustering of FSW’s drug use around specific substances and how different patterns, or typologies, of substance use shape HIV treatment outcomes. Studies of sexually and racially diverse men living with HIV have found that specific polydrug use patterns, including poly-amphetamine use, attenuate the likelihood of viral load suppression.17,18 Importantly, there is a paucity of evidence linking substance use typologies among FSW to HIV treatment outcomes, including viral load suppression. Understanding substance use patterns and their determinants is key to optimizing HIV service delivery for viremic FSW, who exhibit elevated risks of virologic failure and onward HIV transmission.

To address this gap, we used psychometric statistical modeling techniques to identify latent patterns of substance use, examine factors associated with these discrete substance use typologies, and assess their relationship to HIV viremia in a large, cross-sectional sample of FSW living with HIV in South Africa.

Methods

Design and Population

Data are derived from the enrollment visit of Siyaphambili (“pushing forward together” in isiZulu), a pragmatic randomized trial of adaptive differentiated HIV treatment strategies for FSW living with HIV in Durban, South Africa. Siyaphambili study procedures, including recruitment and sampling methods, are reported in detail elsewhere.19–21 Briefly, peers conducted study recruitment activities from a mobile van in mapped priority sex work venues in eThekwini Municipality (Durban) and from a drop-in center managed by TB HIV Care, the largest key populations-focused HIV service provider within the district. Across community and clinical settings, peers approached potentially eligible women and invited them to screen for study enrollment.

Eligibility criteria included: (1) aged ≥18 years; (2) assigned female sex at birth; (3) reported selling sex for goods or money as the primary source of income; (4) living with HIV and diagnosed ≥6 months prior to screening; (5) residing in Durban with no intention to relocate for ≥2 months in the next year; and (6) isiZulu or English comprehension. Women who were pregnant, on second-line ART, or participating in ART adherence clubs at the time of screening were excluded.

Procedures

FSW who met inclusion criteria and provided written informed consent were enrolled on an ongoing basis from June 2018 through March 2020, when data collection was suspended following the implementation of COVID-19 lockdown measures in South Africa. At the enrollment visit, participants completed an interviewer-administered structured questionnaire, followed by CD4 cell count and HIV viral load assessments using whole blood drawn by nurses. Participants received 100 ZAR (~7 USD) for completing the enrollment visit. The Johns Hopkins Bloomberg School of Public Health Institutional Review Board, the University of Western Cape Biomedical Research Ethics Committee, and the eThekwini Municipal and Kwa-Zulu Natal Provincial Departments of Health approved the study protocol.

Measures

Our primary outcome was HIV viremia, defined using a cutpoint of ≥50 RNA copies/mL in alignment with the South African Ministry of Health guidelines.22 Quantitative serum viral load assessments were completed at the Department of Virology at Inkosi Albert Lethuli Central Hospital in Durban, part of the National Health Laboratory Service, using the Abbott RealTime HIV-1 m2000rt assay (Abbott Molecular, Inc., Des Plaines, IL).

Substance use was the primary exposure of interest and was measured dichotomously (yes or no) through self-reported use of the following illicit drugs in the past 30 days: cannabis, cocaine (or “rice”), crack (or “rock”/”tear drop”), methamphetamine (or “tik”), ecstasy or methylenedioxymethamphetamine (MDMA), heroin, and whoonga (or nyaope/”sugar”, a heroin derivative containing various cutting additives, from pharmaceuticals to chemical agents).23,24 Other unenumerated self-reported substances used in the past month included methcathinone (or “flakk”/”cat”, a psychoactive stimulant colloquially known as bath salts) and methaqualone (or “mandrax”, a synthetic sedative agent). Alcohol use was measured using the first two items of the Alcohol Use Disorders Identification Test-Concise (AUDIT-C): (1) “How many standard drinks containing alcohol do you have on a typical day where you drink?” and (2) How often do you have a drink containing alcohol?”; composite scores of ≥3 (range: 0–12) were defined as heavy alcohol use.25 Separately, recent injection drug use was ascertained from self-reported injection of any drugs in the past six months (yes or no).

Other independent variables included factors closely associated with both HIV viremia and substance use (see Figure 1), identified a priori from the literature; we sought to quantify and compare the relative frequencies of these variables across latent substance use typologies. Socio-demographics included age (measured in continuous years), race (Black or other), nationality (South African or non-South African), completed education (secondary and higher or primary and less), current relationship status (steady partnerships of any cohabiting status or unpartnered/single), parity (nulliparous or multiparous), and mobility (spent any nights outside Durban in the past six months or not).

Figure 1.

Figure 1.

Open in a new tab

Directed acyclic graph illustrating pathways from substance use to HIV viremia and theorized confounders.

Occupational factors included age at sex work initiation (measured in continuous years), number of regular and new paying sexual partners in the past 30 days (measured continuously), and consistent (always or less than always) condom use with regular and new paying partners in the past 30 days. We also taxonomized type of sex work setting into indoor (private homes, brothels, bars/clubs, private parties, hotels/guesthouses, shelters) and outdoor (streets, parks, public gardens, beaches, private vehicles, cemeteries) venues from places where FSW primarily sold sex to paying partners.

We also included several validated instruments of health and wellbeing. Depressive symptoms were measured using the Patient Health Questionnaire-9 (PHQ-9), a 9-item, 4-point instrument measuring depression severity (Cronbach’s alpha: 0.92).26 Scores were initially grouped into five PHQ-9-defined categories of symptom severity (none/minimal, mild, moderate, moderately severe, severe), then collapsed dichotomously (none-to-mild or moderate-to-severe). Self-reported health status was measured using a single question ascertaining the quality of one’s health on a continuous scale from 0 (worst) to 100 (best).

Lastly, measures of anticipated (expectations of stereotyping, discrimination, or prejudice) and enacted (experiences with stereotyping, discrimination, or prejudice) stigma related to sex work (17 items) and HIV (13 items) were separately quantified and dichotomized to distinguish FSW reporting any experience with sex work and HIV-related stigmas, respectively (yes or no).27 Individual items of anticipated and enacted stigmas are described elsewhere (see Supplemental Table S1). Lifetime physical and sexual violence victimization, respectively, were measured dichotomously (yes or no) from two questions adapted from the WHO Violence Against Women Instrument: (1) “Has someone ever pushed, shoved, slapped, hit, kicked, choked, or otherwise physically hurt you?” and (2) “Has someone ever forced you to have sex when you did not want to?”28

Analyses

First, we estimated prevalence and calculated measures of central tendency for dichotomous (frequencies, percentages) and continuous (medians, interquartile ranges) variables. We then stratified descriptive sample statistics by HIV viremia, implementing Pearson’s chi-square (χ2) tests (for dichotomous covariates) and Wilcoxon rank-sum tests (for continuous covariates) to identify statistically significant (p<0.05) differences in sample characteristics by viral load status.

To identify distinguishable substance use typologies among FSW reporting any heavy alcohol or illicit drug use in the past month, we used latent class analysis (LCA)—a person-centered modeling approach that probabilistically assigns individuals to discrete groups, or classes, emerging from underlying covariate patterns. Following standards of latent class model identifiability, we used eight systematically enumerated substances (K) from the enrollment visit survey to fit six distinct LCA models, starting with two and concluding with seven latent classes (K-1 class threshold). Model diagnostics, predicted prevalence of the smallest latent class, and interpretability of the resultant latent classes guided selection of the best-fitting LCA model.29,30 Inspected LCA model fit indices included the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), entropy, Lo-Mendell-Rubin Adjusted Likelihood Ratio Test (LMR LRT), and the Bootstrapped Likelihood Ratio Test (BLRT).

After identifying the best-fitting and most interpretable LCA model, we first implemented the Bolck, Croon, and Hagenaars (BCH) 3-step mixture modeling procedure to identify statistically significant correlates (auxiliary predictors) of LCA-identified substance use profiles.31 The BCH approach tends to outperform other 3-step approaches by applying analytic weights that correct for potential misclassification of probabilistic latent class assignments.32 We inspected Wald tests from the BCH procedure to identify statistically significant differences in covariate distributions across latent substance use classes. We then used Vermunt’s 3-step maximum likelihood approach to calculate adjusted odds ratios (adjOR) and 95% confidence intervals (95%CI) for HIV viremia across substance use typologies.32 This approach simultaneously fits a latent class model and quantifies covariate relationships with probabilistic (rather than observed) latent class membership, avoiding potential shifts in latent class formation that might have occurred in the presence of auxiliary predictors.32 The final multivariable mixture model adjusted for factors associated with both HIV viremia (identified from χ2 and Wilcoxon rank-sum tests) and substance use profiles (identified from BCH Wald tests) at the p<0.1 level.

To address missing data (~5% of all observations), we implemented LCA and mixture modeling with full-information maximum likelihood estimators, which use an imputation-like procedure to estimate missing observations from all available cases.33 We managed and descriptively analyzed data in Stata/IC 15.1 (StataCorp LLC, College Station, TX) and implemented LCA and mixture modeling in MPlus 8.4 (Muthen & Muthen, Los Angeles, CA).

Results

Sample Characteristics

Overall, 1,391 FSW were enrolled in Siyaphambili and completed a baseline assessment (see Table 1). The median age was 31 years (interquartile range [IQR] 27–37). Most identified as Black (97%), were South African nationals (98%), and completed primary school or less (81%). More than half reported no steady romantic partnerships (51%), and most had at least one child (85%). Nearly a third experienced homelessness in the past six months (30%). Few FSW reported spending an evening in a location other than Durban in the past six months (11%).

Table 1.

Descriptive sample characteristics, by viral load suppression (VLS) outcomes (N = 1,391).

Characteristics (n, %) Total
N = 1,391		 HIV Viremia
n = 857 (62.1%)		 VLS
n = 523 (37.9%)		 p-value
Demographics
Age, in years (median, IQR) 31 (27–37) 29 (26–34) 34 (29–41) <0.001
Race 0.067
 Black 1,341 (97.2) 824 (96.6) 511 (98.3)
 Other 38 (2.8) 29 (3.4) 9 (1.7)
Nationality 0.967
 South African 1,354 (98.3) 837 (98.2) 511 (98.3)
 Non-South African 24 (1.7) 15 (1.8) 9 (1.7)
Completed education 0.913
 Less than secondary 1,116 (80.9) 691 (81.0) 420 (80.8)
 Secondary or higher 263 (19.1) 162 (19.0) 100 (19.2)
Current relationship status 0.238
 Steadily partnered 680 (49.3) 410 (48.1) 267 (51.3)
 Unpartnered (single) 699 (50.7) 443 (51.9) 253 (48.7)
Parity 0.007
 Nulliparous (no children) 201 (14.6) 140 (16.4) 58 (11.2)
 Multiparous (1 child or more) 1,175 (85.4) 712 (83.6) 461 (88.8)
Homelessness, past 6 months 408 (29.7) 299 (35.2) 105 (20.2) <0.001
Mobility, past 6 months 151 (11.1) 93 (11.1) 58 (11.3) 0.894
Occupational factors
Age at sex work initiation, in years (median, IQR) 23 (20–27) 22 (20–26) 24 (20–29) <0.001
Primary sex work venue 0.695
 Indoor 1,010 (73.4) 628 (73.8) 378 (72.8)
 Outdoor 365 (26.6) 223 (26.2) 141 (27.2)
Number of regular and new paying partners§ 18 (10–30) 18 (10–30) 18 (10–30) 0.792
Consistent condom use with regular paying partners§ 682 (52.9) 406 (51.3) 273 (55.3) 0.162
Consistent condom use with new paying partners§ 823 (62.4) 494 (60.6) 327 (65.4) 0.082
Health and wellbeing
Depressive symptomatology (PHQ-9) 0.045
 None to mild 906 (66.7) 546 (64.8) 358 (70.1)
 Moderate to severe 453 (33.3) 297 (35.2) 153 (29.9)
Self-reported health status (median, IQR) 52 (50–70) 51 (46–70) 56 (50–76) <0.001
Stigma and violence
Anticipated or enacted sex work stigma, lifetime 1,239 (90.3) 781 (92.1) 454 (87.5) 0.005
Anticipated or enacted HIV stigma, lifetime 666 (48.8) 425 (50.4) 238 (46.0) 0.122
Physical violence victimization, lifetime 743 (54.0) 491 (57.6) 249 (48.0) 0.001
Sexual violence victimization, lifetime 517 (37.6) 360 (42.3) 155 (29.9) <0.001
Substance use
Any substance use, past 30 days 1,209 (86.9) 774 (90.3) 430 (82.2) <0.001
 Heavy alcohol 879 (63.2) 530 (61.8) 346 (66.2) 0.106
 Cannabis 478 (34.4) 320 (37.3) 155 (29.6) 0.003
 Crack 378 (27.2) 284 (33.1) 93 (17.8) <0.001
 Whoonga 215 (15.5) 177 (20.7) 35 (6.7) <0.001
 Cocaine 138 (9.9) 95 (11.1) 43 (8.2) 0.085
 Ecstasy/MDMA 67 (4.8) 41 (4.8) 26 (5.0) 0.875
 Heroin 31 (2.2) 24 (2.8) 7 (1.3) 0.075
 Methamphetamine 24 (1.7) 15 (1.8) 9 (1.7) 0.968
 Methcathinone 9 (0.7) 3 (0.4) 5 (1.0) 0.150
 Methaqualone 2 (0.1) n/a 2 (0.4) 0.070
Injection drug use, past 6 months 53 (3.9) 42 (5.0) 9 (1.7) 0.002
Open in a new tab

Notes:

§

Measured in the past 30 days. Dichotomous variables (reported with numbers and percentages) were compared using Pearson’s chi-square tests, and continuous variables (reported with medians and interquartile ranges) were compared using Wilcoxon rank-sum tests.

Bolded values represent statistically significant differences by viral load suppression outcomes at p<0.05 level. HIV viremia was defined as a viral load ≥50 RNA copies/mL.

The median age at sex work initiation was 23 years (IQR 20–27). Most FSW reported primarily selling sex at indoor venues (73%) and a median of 18 past-month (regular and new) paying partners (IQR 10–30). Condom use was inconsistent among FSW, fewer than two-thirds of whom reported always using condoms with regular (53%) and new (62%) paying partners in the past month.

In terms of health and wellbeing, depressive symptomology was prevalent (33%). The median self-reported health status score was 52 (IQR 50–70). Experiences with anticipated or enacted stigmas related to sex work (90%) and HIV (49%) were commonplace. Over a third of FSW reported lifetime exposure to physical (54%) and sexual (38%) violence, respectively.

Prevalence and Correlates of HIV Viremia

Most FSW exhibited HIV viremia (62%) at the time of assessment. Younger (median age: 29 vs. 34 years, p<0.001) and nulliparous (16% vs. 11%, p=0.007) FSW were significantly more likely to exhibit viremia (see Table 1). FSW who were homeless (35% vs. 20%, p<0.001) and younger at sex work initiation (median: 22 vs. 24 years, p<0.001) were more likely to exhibit viremia. Other significant bivariate correlates of viremia included moderate-to-severe depression (35% vs. 30%, p=0.045), lower self-reported health status (median score: 51 vs. 56, p<0.001), anticipated or enacted sex work stigma (92% vs. 88%, p=0.005), and lifetime physical (58% vs. 48%, p=0.001) and sexual (42% vs. 30%, p<0.001) violence victimization.

Latent Classes of Substance Use

A majority of FSW reported any past-month heavy alcohol or illicit drug use (87%) (see Table 1). The prevalence of substances used in the past month was as follows: heavy alcohol (63%), cannabis (34%), crack (27%), whoonga (16%), cocaine (10%), ecstasy/MDMA (5%), heroin (2%), and methamphetamine (2%). Fewer than 1% of FSW reported methcathinone or methaqualone use in the past month. Injection drug use in the past six months was infrequently reported (4%) but was significantly associated with HIV viremia (5% vs. 2%, p=0.002).

Following inspection of LCA model fit indices and assessment of latent class distinguishability across models, a 3-class solution emerged as the best-fitting and most interpretable model of past-month substance use. Relative to other models, the 3-class solution yielded the lowest BIC (7138.845) and adequate entropy (0.658) while maintaining a predicted sample prevalence >10% in the smallest class (see Supplemental Table S2).30

The 3-class solution uncovered discrete polysubstance use profiles (see Figure 2). The modal latent class, Heavy Alcohol Use Only (~54% predicted sample prevalence) was characterized exclusively by universal heavy alcohol use (100%). Next, Cannabis, Crack, & Heavy Alcohol Use (~28% predicted sample prevalence) was distinguishable from other classes by high polysubstance use, specifically cannabis (75%), heavy alcohol (55%), and crack (50%) use. Lastly, Whoonga & Crack Use (~18% predicted sample prevalence), relative to other classes, exhibited the highest propensities for whoonga (78%) and crack (62%) use.

Figure 2.

Figure 2.

Open in a new tab

Conditional item-response probabilities of substance use, by probabilistic latent class membership.

Characteristics of Polysubstance Use Profiles

Table 2 summarizes statistically significant global Wald tests from mixture modeling, which compare the distribution of predicted covariate means and probabilities across latent polysubstance use profiles. FSW with predicted membership in the Heavy Alcohol Use Only class were significantly older in age (χ2 190.704, p<0.001) and at time of sex work initiation (χ2 51.487, p<0.001) relative to other latent polysubstance use profiles. FSW with predicted membership in this class were also significantly more likely to report having any children (χ2 21.635, p<0.001) and better self-reported health (χ2 26.966, p<0.001) compared to FSW with predicted membership in other latent substance use classes. By comparison, FSW with predicted membership in the Whoonga & Crack Use class exhibited the highest probabilities of homelessness (χ2 124.584, p<0.001), outdoor sex work (χ2 47.593 p<0.001), anticipated or enacted HIV stigma (χ2 7.959, p=0.019), and physical (χ2 10.732, p=0.005) and sexual (χ2 33.630, p<0.001) violence victimization across latent classes. FSW in this class were also the least mobile (χ2 13.226, p=0.001) and reported significantly fewer past-month paying partners (χ2 11.073, p=0.004) relative to other latent substance use classes.

Table 2.

Correlates of latent polysubstance use profiles identified from mixture modeling.

Characteristics Predicted Means and Probabilities
 Wald Test
Heavy Alcohol Use Only Cannabis, Heavy Alcohol, & Crack Use Whoonga & Crack Use Overall χ2 p-value
Age, in years (mean) 34 29 28 190.704 <0.001
Parity 21.635 <0.001
 Nulliparous 0.105 0.205 0.248
 Multiparous 0.895 0.795 0.752
Homelessness, past 6 months 0.171 0.297 0.655 124.584 <0.001
Mobility, past 6 months 0.116 0.148 0.036 13.226 0.001
Age at sex work initiation, in years (mean) 25 22 22 51.487 <0.001
Primary sex work venue 47.593 <0.001
 Indoor 0.749 0.881 0.490
 Outdoor 0.251 0.119 0.510
Number of regular and new paying partners, past 30 days (mean) 24 33 20 11.073 0.004
Self-reported health status (mean) 60 55 51 26.966 <0.001
Anticipated or enacted HIV stigma, lifetime 0.467 0.469 0.596 7.959 0.019
Physical violence victimization, lifetime 0.532 0.578 0.674 10.732 0.005
Sexual violence victimization, lifetime 0.313 0.454 0.549 33.630 <0.001
HIV viremia (≥50 RNA copies/mL) 0.555 0.677 0.849 60.071 <0.001
Open in a new tab

Notes: Characteristics presented in the table were restricted to those that differed significantly (p<0.05) across latent substance use classes, per global Wald tests from the BCH 3-step procedure.

HIV Viremia by Polysubstance Use Profile

Relative to the Heavy Alcohol Use Only class (56%), both the Cannabis, Heavy Alcohol, & Crack Use (68%) and Whoonga & Crack Use (85%) classes exhibited significantly higher probabilities of HIV viremia (χ2 60.071, p<0.001). Figure 3 illustrates unadjusted and adjusted associations of HIV viremia with latent polysubstance use profiles, identified from mixture modeling. In multivariable analysis, the Whoonga & Crack Use class exhibited significantly higher adjusted odds of HIV viremia than the Heavy Alcohol Use Only class (adjOR 1.97, 95%CI 1.13–3.43, p=0.017). The observed association of HIV viremia with the Cannabis, Heavy Alcohol, & Crack Use class, relative to the Heavy Alcohol Use Only class, in bivariate analysis was attenuated in the presence of potential confounders, indicating viremia did not differ significantly between these latent classes (adjOR 1.17, 95%CI 0.74–1.86, p=0.494). The adjusted odds of HIV viremia was higher, albeit non-significantly, in the Whoonga & Crack Use class compared to the Cannabis, Heavy Alcohol, & Crack Use class (adjOR 1.68, 95%CI 0.84–3.35, p=0.144).

Figure 3.

Figure 3.

Open in a new tab

Unadjusted and adjusted odds ratios of HIV viremia identified in mixture modeling, by latent polysubstance use profiles.

Notes: Multivariable model adjusted for age, parity, homelessness, age at sex work initiation, self-reported health status, and lifetime physical and sexual violence. The reference group for both multivariable models was the Heavy Alcohol Use Only class. HIV viremia was defined as a viral load ≥50 RNA copies/mL.

Discussion

We uncovered three distinct patterns of polysubstance use among FSW living with HIV who use drugs, which were differentially associated with HIV viremia. FSW with predicted membership in the most prevalent class, Heavy Alcohol Use Only, exhibited significantly higher odds of viral load suppression relative to FSW with predicted membership in the Cannabis, Heavy Alcohol, & Crack Use and Whoonga & Crack Use classes, respectively. In the presence of confounders, however, HIV viremia was significantly higher in the Whoonga & Crack Use class but not the Cannabis, Heavy Alcohol, & Crack Use class, suggesting that HIV viremia clusters significantly in FSW concurrently using opioids and stimulants. Taken together, these findings indicate that specific polydrug use profiles, notably stimulant and opioid co-use, are likely reflective of underlying sources of marginalization and unmet mental health needs, amplifying risk for poor HIV treatment outcomes among FSW.

Consistent with findings from the extant literature,34–37 the Whoonga & Crack Use profile in our study was distinguished by the highest rates of poverty, stigma, violence, and poor health among latent substance use classes. FSW with predicted membership in the Whoonga & Crack Use class were most likely to report homelessness, which is closely associated with sex work initiation to support drug use and, in particular, opioid-related withdrawal symptom management.13,38 Likewise, the Whoonga & Crack Use class had the highest predicted prevalence of outdoor sex work, which has been linked to physical and sexual violence victimization in other studies of FSW.39,40 These overlapping sources of marginalization, coupled with the affordability of specific opioids and stimulants for economically disenfranchised groups, may drive heightened use of crack and whoonga, which are associated with HIV treatment challenges (i.e., suboptimal ART adherence, care disengagement and attrition).41,42 Addressing concomitant social and structural challenges experienced by FSW with discrete polydrug use patterns, through interventions like housing stabilization and stigma mitigation, may be key to increasing viral load suppression in the context of substance use.

The polydrug use profiles emerging from LCA in our study were distinct from substance use typologies identified in other studies of FSW and people living with HIV. A study of street-based FSW at risk for and living with HIV in the United States identified three substance use patterns, all of which were characterized by high probabilities of opioid (predominantly heroin) use.36 Another study of FSW in Pretoria, South Africa, found only two distinct profiles of substance use (high opioid/cocaine/cannabis use and moderate alcohol consumption), but this study did not enumerate an exhaustive list of substances that may have further delineated these latent classes.43 Likewise, a study of sexually diverse men living with HIV in the United States found that among three polydrug use typologies identified from LCA, HIV viremia was most prevalent in the class characterized by high amphetamine use.18 Despite these notable heterogeneities in polydrug use patterns across studies, opioid and stimulant co-use appear to most consistently cluster with elevated marginalization and poorer health outcomes, consistent with our study findings. Thus, contextually appropriate strategies to improve HIV care outcomes in this cohort of FSW could be guided by evidenced-based practices from other settings, including mobile substance use disorder treatment and co-located substance use, mental health, and HIV treatment service delivery.44–46

Our findings must be considered with several limitations in mind. First, unlike other studies,34,36 ours did not include measures of drug consumption routes (e.g., injecting, snorting/sniffing, smoking, or swallowing), frequencies (e.g., daily vs. non-daily use), or quantities (e.g., amount per consumption episode). The relationship between HIV viremia and the aforementioned substance use attributes in our study population is unaddressed by the present study. Second, our measure of heavy alcohol use was derived only from the first two items of the AUDIT-C, which likely underestimated the true prevalence of hazardous alcohol consumption in the study population. Third, although mixture modeling treats latent variables as predictors of a distal outcome (i.e., HIV viremia), the study’s design limits temporal inferences that can be gleaned from the observed cross-sectional associations of polysubstance use with HIV viremia. Longitudinal studies examining polydrug use and incident viral load suppression or viral rebound are, therefore, warranted. Fourth, apart from viral load, all data were self-reported and are subject to social desirability and recall biases. Lastly, our study population was restricted to FSW living with HIV in Durban, and findings may differ among FSW living without HIV or residing in settings characterized by drug markets and HIV service delivery environments that are distinguishable from those observed in urban eThekwini.

Conclusion

In our study of 1,391 FSW living with HIV in South Africa, polysubstance use and viremia were highly prevalent, and specific polysubstance use profiles were more closely associated with viremia. Even among FSW reporting only heavy alcohol use, the predicted probability of viremia was high (~54%), suggesting that even non-illicit substance use contributes to suboptimal HIV treatment outcomes. In this population characterized by both high levels of substance use and HIV viremia, interventions should prioritize reducing the harms associated with substance use as a vehicle for optimizing engagement with HIV treatment services. Documenting FSW preferences for substance use-related services and appropriately integrating these services into HIV treatment programs will be key to addressing observed gaps in viral load suppression.

Supplementary Material

Supplemental Digital Content

Acknowledgements

We wish to acknowledge the Siyaphambili study participants, community advisory board, and data collection teams—without whom this work would not be possible.

Sources of Support

This study was funded by the National Institute of Nursing Research (R01NR016650) and the Johns Hopkins University Center for AIDS Research (P30AI094189), an NIH-funded program. JGR, KBR, and SWB were supported by the National Institute of Mental Health (F31MH126796, K01MH129226, K01MH114715). JRK was supported by the National Institute on Alcohol Abuse and Alcoholism (K01AA028199) and the National Institute on Drug Abuse (R01DA054553, R21DA053156, R01DA057351). The content is solely the responsibility of the authors and does not necessarily reflect the official views of the NIH.

Footnotes

Conflicts of Interest

The authors have no conflicts of interest to disclose.

References

  • 1.Joint United Nations Programme on HIV/AIDS. UNAIDS Data 2021 UNAIDS; 2022. https://www.unaids.org/sites/default/files/media_asset/JC3032_AIDS_Data_book_2021_En.pdf [PubMed] [Google Scholar]
  • 2.Stone J, Mukandavire C, Boily MC, et al. Estimating the contribution of key populations towards HIV transmission in South Africa. Journal of the International AIDS Society 2021;24(1):e25650. doi: 10.1002/jia2.25650 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Shannon K, Strathdee SA, Goldenberg SM, et al. Global epidemiology of HIV among female sex workers: influence of structural determinants. Lancet 2015;385(9962):55–71. doi: 10.1016/S0140-6736(14)60931-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Schwartz S, Lambert A, Phaswana-Mafuya N, et al. Engagement in the HIV care cascade and barriers to antiretroviral therapy uptake among female sex workers in Port Elizabeth, South Africa: findings from a respondent-driven sampling study. Sex Transm Infect 2017;93(4):290–296. doi: 10.1136/sextrans-2016-052773 [DOI] [PubMed] [Google Scholar]
  • 5.Azar MM, Springer SA, Meyer JP, Altice FL. A Systematic Review of the Impact of Alcohol Use Disorders on HIV Treatment Outcomes, Adherence to Antiretroviral Therapy and Health Care Utilization. Drug Alcohol Depend 2010;112(3):178–193. doi: 10.1016/j.drugalcdep.2010.06.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Gonzalez A, Barinas J, O’Cleirigh C. Substance Use: Impact on Adherence and HIV Medical Treatment. Curr HIV/AIDS Rep 2011;8(4):223. doi: 10.1007/s11904-011-0093-5 [DOI] [PubMed] [Google Scholar]
  • 7.Deryabina AP, El-Sadr WM. Optimizing HIV prevention and treatment outcomes for persons with substance use in Central Asia: what will it take? Current Opinion in HIV and AIDS 2019;14(5):374–380. doi: 10.1097/COH.0000000000000565 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Parmley LE, Comins CA, Young K, et al. Occupational barriers to accessing and adhering to antiretroviral therapy for female sex workers living with HIV in South Africa. Occup Environ Med 2020;77(2):100–106. doi: 10.1136/oemed-2019-105947 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Beksinska A, Nyariki E, Kabuti R, et al. Harmful Alcohol and Drug Use Is Associated with Syndemic Risk Factors among Female Sex Workers in Nairobi, Kenya. Int J Environ Res Public Health 2022;19(12):7294. doi: 10.3390/ijerph19127294 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Yeo EJ, Hlongwane K, Otwombe K, et al. Key risk factors for substance use among female sex workers in Soweto and Klerksdorp, South Africa: A cross-sectional study. PLoS One 2022;17(1):e0261855. doi: 10.1371/journal.pone.0261855 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Pellowski JA, Huedo-Medina TB, Kalichman SC. Food Insecurity, Substance Use, and Sexual Transmission Risk Behavior Among People Living with HIV: A Daily Level Analysis. Arch Sex Behav 2018;47(7):1899–1907. doi: 10.1007/s10508-017-0942-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Rosen JG, Park JN, Opper N, Lim S, Sherman SG. Patterns and Correlates of New Drug Initiation among Female Exotic Dancers: The Contribution of Occupational and Structural Risks. Subst Use Misuse 2020;55(7):1122–1128. doi: 10.1080/10826084.2020.1729199 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ogden SN, Harris MT, Childs E, et al. “You need money to get high, and that’s the easiest and fastest way:” A typology of sex work and health behaviours among people who inject drugs. International Journal of Drug Policy 2021;96:103285. doi: 10.1016/j.drugpo.2021.103285 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Colasanti J, Stahl N, Farber EW, Del Rio C, Armstrong WS. An Exploratory Study to Assess Individual and Structural Level Barriers Associated With Poor Retention and Re-engagement in Care Among Persons Living With HIV/AIDS. J Acquir Immune Defic Syndr 2017;74 Suppl 2:S113–S120. doi: 10.1097/QAI.0000000000001242 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Kerkerian G, Kestler M, Carter A, et al. Attrition Across the HIV Cascade of Care Among a Diverse Cohort of Women Living With HIV in Canada. J Acquir Immune Defic Syndr 2018;79(2):226–236. doi: 10.1097/QAI.0000000000001775 [DOI] [PubMed] [Google Scholar]
  • 16.Degarege A, Krupp K, Tamargo J, Martinez SS, Campa A, Baum M. Polysubstance use and adherence to antiretroviral treatment in the Miami Adult Studies on HIV (MASH) cohort. AIDS Care 2022;34(5):639–646. doi: 10.1080/09540121.2021.1896660 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Cleland CM, Gwadz M, Collins LM, et al. African American/Black and Latino Adults with Detectable HIV Viral Load Evidence Substantial Risk for Polysubstance Substance Use and Co-occurring Problems: A Latent Class Analysis. AIDS Behav 2021;25(8):2501–2516. doi: 10.1007/s10461-021-03212-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Meyers-Pantele SA, Rendina HJ, Talan AJ, Shalhav O, Lammert S, Horvath KJ. Characterizing substance use typologies and their association with HIV viral load outcomes: A latent class analysis among sexual minority men living with HIV. Drug Alcohol Depend 2021;227:108928. doi: 10.1016/j.drugalcdep.2021.108928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Comins CA, Schwartz SR, Phetlhu DR, et al. Siyaphambili protocol: An evaluation of randomized, nurse-led adaptive HIV treatment interventions for cisgender female sex workers living with HIV in Durban, South Africa. Res Nurs Health 2019;42(2):107–118. doi: 10.1002/nur.21928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Wang L, Dowdy DW, Comins CA, et al. Health-related quality of life of female sex workers living with HIV in South Africa: a cross-sectional study. J Int AIDS Soc 2022;25(2):e25884. doi: 10.1002/jia2.25884 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Rock A, Comins C, Mulumba N, et al. Antiretroviral Treatment Sharing among Female sex Workers Living with HIV in eThekwini (Durban), South Africa: Drivers and Implications for Treatment Success. J Int Assoc Provid AIDS Care 2022;21:23259582221110820. doi: 10.1177/23259582221110820 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.National Department of Health. 2019 ART Clinical Guidelines for the Management of HIV in Adults, Pregnancy, Adolescents, Children, Infants and Neonates Government of the Republic of South Africa; 2020. https://www.knowledgehub.org.za/system/files/elibdownloads/2020-05/2019%20ART%20Guideline%2028042020%20pdf.pdf [Google Scholar]
  • 23.Grelotti DJ, Closson EF, Smit JA, et al. Whoonga: potential recreational use of HIV antiretroviral medication in South Africa. AIDS Behav 2014;18(3):511–518. doi: 10.1007/s10461-013-0575-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Tyree GA, Mosery N, Closson EF, et al. Trajectories of initiation for the heroin-based drug whoonga - qualitative evidence from South Africa. Int J Drug Policy 2020;82:102799. doi: 10.1016/j.drugpo.2020.102799 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA, for the Ambulatory Care Quality Improvement Project (ACQUIP). The AUDIT Alcohol Consumption Questions (AUDIT-C): An Effective Brief Screening Test for Problem Drinking. Archives of Internal Medicine 1998;158(16):1789–1795. doi: 10.1001/archinte.158.16.1789 [DOI] [PubMed] [Google Scholar]
  • 26.Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001;16(9):606–613. doi: 10.1046/j.1525-1497.2001.016009606.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Earnshaw VA, Chaudoir SR. From conceptualizing to measuring HIV stigma: a review of HIV stigma mechanism measures. AIDS Behav 2009;13(6):1160–1177. doi: 10.1007/s10461-009-9593-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Schraiber LB, Latorre M do RDO, França I, Segri NJ, D’Oliveira AFPL. Validity of the WHO VAW study instrument for estimating gender-based violence against women. Rev Saude Publica 2010;44(4):658–666. doi: 10.1590/s0034-89102010000400009 [DOI] [PubMed] [Google Scholar]
  • 29.Muthén B, Muthén LK. Integrating Person-Centered and Variable-Centered Analyses: Growth Mixture Modeling With Latent Trajectory Classes. Alcoholism: Clinical and Experimental Research 2000;24(6):882–891. doi: 10.1111/j.1530-0277.2000.tb02070.x [DOI] [PubMed] [Google Scholar]
  • 30.Nylund KL, Asparouhov T, Muthén BO. Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation Modeling: A Multidisciplinary Journal 2007;14(4):535–569. doi: 10.1080/10705510701575396 [DOI] [Google Scholar]
  • 31.Bolck A, Croon M, Hagenaars J. Estimating latent structure models with categorical variables: One-step versus three-step estimators. Polit Anal 2004;12(1):3–27. doi: 10.1093/pan/mph001 [DOI] [Google Scholar]
  • 32.Vermunt JK. Latent Class Modeling with Covariates: Two Improved Three-Step Approaches. Political Analysis 2017;18(4):450–469. doi: 10.1093/pan/mpq025 [DOI] [Google Scholar]
  • 33.Schafer JL, Olsen MK. Multiple Imputation for Multivariate Missing-Data Problems: A Data Analyst’s Perspective. Multivariate Behav Res 1998;33(4):545–571. doi: 10.1207/s15327906mbr3304_5 [DOI] [PubMed] [Google Scholar]
  • 34.Tavitian-Exley I, Boily MC, Heimer R, Uusküla A, Levina O, Maheu-Giroux M. Polydrug Use and Heterogeneity in HIV Risk Among People Who Inject Drugs in Estonia and Russia: A Latent Class Analysis. AIDS Behav 2018;22(4):1329–1340. doi: 10.1007/s10461-017-1836-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Marotta PL, Terlikbayeva A, Gilbert L, et al. Intimate relationships and patterns of drug and sexual risk behaviors among people who inject drugs in Kazakhstan: A latent class analysis. Drug Alcohol Depend 2018;192:294–302. doi: 10.1016/j.drugalcdep.2018.07.046 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Beckham SW, Glick JL, Schneider KE, et al. Latent Classes of Polysubstance Use and Associations with HIV Risk and Structural Vulnerabilities among Cisgender Women Who Engage in Street-Based Transactional Sex in Baltimore City. Int J Environ Res Public Health 2022;19(7):3783. doi: 10.3390/ijerph19073783 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Schneider KE, Park JN, Allen ST, Weir BW, Sherman SG. Patterns of polysubstance use and overdose among people who inject drugs in Baltimore, Maryland: A latent class analysis. Drug Alcohol Depend 2019;201:71–77. doi: 10.1016/j.drugalcdep.2019.03.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.White RH, Park JN, Galai N, et al. Short-term interruptions to sex work among a prospective cohort of street-based cisgender female sex workers in Baltimore. Int J Drug Policy 2020;84:102858. doi: 10.1016/j.drugpo.2020.102858 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Deering KN, Lyons T, Feng CX, et al. Client demands for unsafe sex: the socioeconomic risk environment for HIV among street and off-street sex workers. J Acquir Immune Defic Syndr 2013;63(4):522–531. doi: 10.1097/QAI.0b013e3182968d39 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Logie CH, White RH, Galai N, Tomko C, Sherman SG. Brief Report: Longitudinal Associations Between Place of Sex Work and Client Condom Coercion Among Sex Workers in Baltimore, Maryland. J Acquir Immune Defic Syndr 2020;85(5):579–583. doi: 10.1097/QAI.0000000000002494 [DOI] [PubMed] [Google Scholar]
  • 41.Azar P, Wood E, Nguyen P, et al. Drug use patterns associated with risk of non-adherence to antiretroviral therapy among HIV-positive illicit drug users in a Canadian setting: a longitudinal analysis. BMC Infectious Diseases 2015;15(1):193. doi: 10.1186/s12879-015-0913-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Varshney K, Browning SD, Debnath SK, Shet P, Shet D. A Systematic Review of Risk Factors and Consequences of Nyaope Usage: The Illicit Street Drug Containing HIV Antiretrovirals. AIDS Behav Published online July 27, 2022. doi: 10.1007/s10461-022-03791-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Wechsberg WM, Peasant C, Kline T, et al. HIV Prevention Among Women Who Use Substances And Report Sex Work: Risk Groups Identified Among South African Women. AIDS Behav 2017;21(Suppl 2):155–166. doi: 10.1007/s10461-017-1889-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Deering KN, Kerr T, Tyndall MW, et al. A peer-led mobile outreach program and increased utilization of detoxification and residential drug treatment among female sex workers who use drugs in a Canadian setting. Drug Alcohol Depend 2011;113(1):46–54. doi: 10.1016/j.drugalcdep.2010.07.007 [DOI] [PubMed] [Google Scholar]
  • 45.Haldane V, Cervero-Liceras F, Chuah FL, et al. Integrating HIV and substance use services: a systematic review. J Int AIDS Soc 2017;20(1):21585. doi: 10.7448/IAS.20.1.21585 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Regis C, Gaeta JM, Mackin S, Baggett TP, Quinlan J, Taveras EM. Community Care in Reach: Mobilizing Harm Reduction and Addiction Treatment Services for Vulnerable Populations. Frontiers in Public Health 2020;8. Accessed October 31, 2022. 10.3389/fpubh.2020.00501 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Digital Content
NIHMS1947126-supplement-Supplemental_Digital_Content.docx (18.3KB, docx)

RESOURCES