Abstract
AIMS: We assessed the feasibility and preliminary efficacy of a multilevel intervention (Kisoboka) to reduce high-risk alcohol use and improve human immunodeficiency virus (HIV) treatment engagement among fisherfolk men in Uganda. DESIGN: A parallel individually randomized controlled pilot trial with follow-up at 3 and 6 months. SETTING: Five HIV clinics near Ugandan fishing communities. PARTICIPANTS: 160 men (80 per arm), aged 18-50, living with HIV, reporting suboptimal antiretroviral therapy (ART) adherence and high-risk alcohol use, enrolled between January 2021 and March 2022. INTERVENTION: Kisoboka applies behavioral economic principles and motivational interviewing to address contextual determinants of alcohol use and suboptimal HIV treatment through counseling sessions, text reminders about savings and health goals, and a structural component of assistance setting up mobile money accounts and use of mobile money to receive work pay. The comparator arm (S&R) received only brief alcohol screening feedback, referrals for alcohol counseling and HIV treatment counseling. MEASUREMENTS: High-risk alcohol use assessed via phosphatidylethanol (PEth) level at baseline and 6 months and Alcohol Use Disorders Identification Test - Consumption (AUDIT-C) (cutoff ≥9) at baseline, 3 and 6 months. Self-reported ART adherence (≥90% versus <90%) assessed at baseline, 3 and 6 months. SECONDARY OUTCOMES: number of days with ≥5 drinks in the prior 28 days and HIV viral load. We assessed intervention fidelity by audio recording and transcribing sessions, then coding a random 20% sample for fidelity to protocol elements. FINDINGS: Retention (92.5% completed all sessions) and fidelity (92.8-100% across sessions) were high. Relative to the comparator arm, PEth values decreased from baseline to 6-month follow-up in the Kisoboka arm [difference in change -126.87 ng/ml, 95% confidence interval (CI) = -264.29 to 10.55], which was a small effect size (adjusted time* arm interaction: d = 0.14). The proportion with AUDIT-C ≥9 decreased more in the Kisoboka arm between baseline and 6 months (-26.01%, 95% CI = -36.09% to -15.93%), a medium effect size time*arm effect (d = 0.68). Kisoboka did not improve adherence but protected against declines in optimal adherence observed in the S&R arm (difference in change: 20.96%, 95% CI = 7.15% to 34.77%), which was a medium effect size time*arm effect (d = 0.51). CONCLUSIONS: The Kisoboka multilevel intervention to reduce high-risk alcohol use appears to be feasible; it showed promising clinically meaningful small to medium size effects on reducing high-risk alcohol use and protected against decreases in antiretroviral therapy adherence in a setting with strong contextual determinants of behavior.