Addictive Behaviors Reports (Jun 2023)

Effects of drug and hazardous alcohol use on having a detectable HIV viral load: An adherence mediation analysis

  • Edward R. Cachay,
  • Tesfaye S. Moges,
  • Huifang Qin,
  • Laura Bamford,
  • David J. Grelotti,
  • Wm. Christopher Mathews

Journal volume & issue
Vol. 17
p. 100486

Abstract

Read online

Objectives: People living with HIV (PWH) with substance or alcohol use often have unsuppressed plasma HIV viral loads (pVL). The degree to which substance and alcohol use effects on HIV viral suppression are mediated through medication nonadherence is incompletely understood. Methods: We included PWH prescribed antiretroviral therapy and receiving care at an academic HIV clinic between 2014 and 2018 who completed both patient-reported outcomes (PRO) questionnaires and had subsequent pVL measurements. Measures included assessments of alcohol use (AUDIT-C), drug use (NIDA-ASSIST), and self-reported adherence measured using four different methods. Substances found in bivariate analysis to predict detectable pVL were modeled separately for mediation effects through adherence. We report natural direct (NDE) and indirect effect (NIE), marginal total effect (MTE), and percentage mediated. Results: Among 3125 PWH who met eligibility criteria, 25.8% reported hazardous alcohol use, 27.1% cannabis, 13.1% amphetamines, 1.9% inhalants, 5.3% cocaine, 4.5% sedative-hypnotics, 2.9% opioids, and 2.3% hallucinogens. Excellent adherence was reported by 58% of PWH, and 10% had detectable pVL. Except for sedatives, using other substances was significantly associated with worse adherence. Bivariate predictors of detectable pVL were [OR (95% CI)]: amphetamine use 2.4 (1.8–3.2) and opioid use 2.3 (1.3–4.0). The percent of marginal total effect mediated by nonadherence varied by substance: 36% for amphetamine use, 27% for opioid use, and 39% for polysubstance use. Conclusion: Use of amphetamines, opioids, and multiple substances predicted detectable pVL. Up to 40% of their effects were mediated by self-reported nonadherence. Confirmation using longitudinal measurement models will strengthen causal inference from this cross-sectional analysis.

Keywords