Study design and setting
This study reports cross-sectional baseline characteristics of the Buprenorphine to Improve HIV Care Engagement and Outcomes (BRAVO) Randomized Trial (ClinicalTrials.govNCT01936857). The BRAVO trial enrolled people with HIV and moderate-to-severe opiate use disorder to receive either methadone maintenance therapy (MMT) or buprenorphine/naloxone (BUP/NX) at six Vietnamese HIV clinics between 2015 and 2019. The study clinics were chosen in the epicenter of Vietnam’s HIV and opiate epidemics; four clinics are located in Hanoi, one in Tanh Hoa Province, and one in Bac Giang Province. Clinics were selected based on the high prevalence of untreated opiate use disorder among newly engaging people with a history of intravenous drug use, ability to enroll sufficiently for the study, availability of MMT, and support of local authorities.
This study was conducted in partnership with Hanoi Medical University, the Provincial AIDS Control authorities of Hanoi, Thanh Hoa and Bac Giang, the Vietnamese National Institute of Mental Health and Oregon Health & Science University in Portland, Oregon.
Consent and data collection
Research assistants reviewed the written Vietnamese language informed consents with participants in a confidential space in the HIV clinic. Participants were instructed their study participation had no bearing on HIV treatment or other usual care at their clinic. Those agreeing to participate signed the consent form which was securely filed. Consents were periodically audited by the study quality assurance monitor for appropriate completion.
Research assistants administered surveys to participants in confidential settings using secured, web-based electronic data entry. All survey instruments were administered in Vietnamese and adapted for appropriateness during previous research. Blood specimens for HIV viral load PCR were collected at HIV clinics and sent to the National Hospital for Tropical Diseases in Hanoi, Vietnam for HIV-1 PCR testing using Abbott m2000rt RealTime HIV-1 PCR amplification assay [19].
Participants
Participants were eligible for inclusion if they were HIV positive, had current moderate-to-severe DSM-V opioid use disorder (OUD), a positive urine drug screen for opioids at the time of enrollment, interest in receiving treatment for OUD, were age 18 or older, and were willing to practice birth control, if female. Study research staff prioritized patients new to HIV care or registered for care but not receiving ART, though participants already receiving ART were allowed to enroll.
Participants were ineligible if they had a known hypersensitivity to buprenorphine or naloxone, an aspartate aminotransferase or alanine aminotransferase greater than five times the upper limit of normal, were currently pregnant or breastfeeding, had a serious medical or psychiatric illness in the past 30 days that precluded safe participation in the opinion of the study physician, or received MMT within the past 30 days. Participants from known vulnerable populations, including decisionally-impaired adults, prisoners, and children, were also excluded.
Measures
We assessed two critical steps in the HIV care continuum as study outcome variables: 1) ART initiation at time of enrollment, classified as a self-report of currently taking ART at the time of study enrollment, and 2) HIV viral suppression at time of enrollment, defined as HIV RNA PCR < 200 copies/mL.
Participant characteristics included gender (male/female), greater than 9th grade education (yes/no), currently employed (yes/no), currently married (yes/no), history of 06 rehabilitation (yes/no), substance use (positive urine drug screen (UDS) for methamphetamine or amphetamine, methadone, or buprenorphine), and self-reported daily tobacco use and risky alcohol use (defined as an AUDIT-C score greater than 3 in women or greater than 4 in men) [20,21,22,23,24]. Continuous covariates included self-reported age, lifetime number of arrests, years since HIV diagnosis, and the Depression Anxiety Stress Scales (DASS) [25] scores for depression, anxiety and stress symptoms, which has been validated for use in Vietnamese [26].
Data analysis
Primary analysis
We assessed bivariate associations (t-tests, Pearson’s chi-squared tests, and Fisher exact tests) of participant characteristics and separately, ART initiation and HIV viral suppression. Potential covariates were selected based on a priori hypotheses and previous studies. We then conducted bivariate analyses among those covariates and the study outcome (viral suppression). We chose to include covariates in our model that had p-values of < 0.10 in bivariate analyses. Regardless of statistical significance, we retained age and years since HIV diagnosis each final model. Two separate multivariable logistic regression models estimated the relationship of baseline participant characteristics with study outcomes. We used an estimated 10 events per degree of freedom ratio to estimate the number of covariates we could include to avoid overfitting [27]. We evaluated our continuous covariates for linearity in the log-odds using Lowess scatter plots, and also evaluated covariates for multi-collinearity, using a cutoff of > 0.80 as a marker for covariate review. Finally, we used Hosmer-Lemeshow tests (p > 0.05) to evaluate model goodness of fit. We dropped participants missing both age and gender, and replaced the three participants missing viral load at baseline as non-suppressed.
Sensitivity analysis
After completing our primary analyses, we identified influential observations using Pregibon’s Delta-Beta statistic and removed observations with a value greater than 0.20 from the model. If the directionality of associations changed, we planned to report these results alongside primary study results.