Testing Effectiveness of the Community Reinforcement Approach and Family Training (CRAFT) With American Indians (CRAFT-AI)

Treatment research with AI/AN is in its infancy. One route to address the substantial substance use related health disparities is to employ evidence-based treatments (EBTs) with AI/AN. EBTs are rarely tested with sufficiently large samples of racial ethnic groups, so outcomes are often unknown. Recently, an R01 DA021672 adapted two EBTs for substance use disorders (SUDs) (motivational interviewing and community reinforcement approach - MICRA) and conducted a randomized controlled trial (RCT) to MICRA to treatment as usual with the same AI community as involved in this proposal. The present study seeks to expand on that R01, which culturally tailored CRA for people with SUDs, but now turn our focus to working with families and concerned significant others (CSOs) who want to help engage their loved one with an addiction (identified person; IP) into treatment. The intervention, Community Reinforcement and Family Training (CRAFT), is built on reinforcement principles. CRAFT teaches CSOs to take better care of themselves, to cease any behaviors that are not working to address the IP's substance use, and to consistently reward the IP for any sober and treatment engagement behavior. Importantly, CRAFT has been shown to increase IP entry into treatment, which would address: (1) the high rates of SUDs and related problems among AI/AN, (2) low rates of AI/AN self-referrals to treatment, and (3) higher rates of AI entering treatment due to court mandates. The present study aims to conduct an RCT to compare an AI culturally adapted version of CRAFT (CRAFT-AI; n=20) to Twelve-step facilitation for CSO's (TSF-CSO) using principles and groups of Nar/Al-Anon (n=20) to examine acceptability of a culturally tailored EBT and estimate effect sizes for IP treatment entry and for CSO functioning pre to post intervention. Notably, this will be the first examination of potential mediators and moderators of CRAFT-AI including variables such as: demographic variables, severity of SUD, self-efficacy, and cultural risk and protective factors. Finally, we will examine the appropriateness of broadening the dependent variable of IPs entering formal treatment to include AI traditional healing. The methodology of this grant using a partnership between academic and tribal researchers is well-poised to contribute to nascent AI/AN SUD treatment research to impact the substantial health disparities AI/ANs and other indigenous populations endure. Positive study results will facilitate future cultural adaptation research and dissemination and diffusion of EBTs to interested AI/AN tribes and other culturally diverse populations.

Quality Assurance 1. For quality assurance of data collection, Roberta Chavez, who has over 25 years of experience working on clinical trials at University of New Mexico (UNM)/Center on Alcoholism, Substance Abuse, and Addiction (CASAA), will train the research assistant and conduct data quality assurance checks monthly. Data entry will be conducted with two passes.

For the intervention, all counselors will receive expert training in either CRAFT-AI or TSF-CSO. All counselors will be certified in CRAFT-AI or demonstrate proficiency in TSF-CSO prior to the RCT using a CRAFT coding manual and TSF-CSO coding forms. All counseling sessions will be digitally recorded and coded for fidelity to the treatment modalities. When counselors do not pass a CRAFT module or TSF-CSO component, they will receive individualized training and remediation to meet fidelity requirements.

Transmission

a. All data will be collected using pre-specified measures, instruments, and qualitative questions. b. Entry of data will occur within 30 days after it is collected. Second pass within 60 days. c. Statistician, Katie Witkiewitz, will conduct analyses to insure data fidelity upon completion of data entry and reconciliation of any discrepancies.

Regulatory Issues 1. Reporting mechanisms of adverse events (AEs) and serious adverse events (SAEs) to Institutional Review Board (IRB) and National Institute on Drug Abuse (NIDA). The PI will report any SAE, whether or not related to study intervention to the UNM IRB, NIDA and the Zuni Tribal Council within 48 hours of the principal investigator's (PI's) notification of the SAE. Outcomes of SAEs will be reported to NIDA and the Zuni Tribal Council and UNM IRB as they become known. A summary of the SAEs that occurred during the previous year will be included in the annual progress report to NIDA, which will also be sent to the Zuni Tribal Council.

2. Reporting mechanisms of IRB actions to NIDA The PI will be responsible to report any IRB actions to NIDA.

3. Trial Stopping Rules In the event that those participants in the intervention arm receiving CRAFT-AI counseling experience a significantly higher rate of SAE's that are probably study related than those participants in TSF-CSO arm of the study, then the trial will be stopped.

Potential risks and benefits for participants The primary risks involved in this research project are: a) loss of confidentiality, b) discomfort talking about problems caused by loved one with a substance use problems, c) potential disclosure of own substance use problems, and c) potential for conflict between the subject and loved one. Steps have been taken to minimize the likelihood of these risks and to ameliorate them if they are present. Past studies have assessed for level of conflict and instituted safety plans in case domestic violence should be imminent or occur.

Participants may potentially benefit from receiving the CRAFT-AI or TSF-CSO interventions in terms of improved psychological functioning such as decreased anxiety and depression. In addition, their loved one may enter treatment.

Trial Efficacy Plans for Interim Analysis of Efficacy Data: We have no plans for interim analyses.

Data analysis plan Analysis Plan. Data Integrity Preliminary Analyses: First we will compare demographics and primary study measures at baseline between randomized treatment groups, using ANOVA for continuous variables (or Kruskal-Wallis if parametric assumptions are violated) and using χ2 tests for categorical variables. Subsequent analyses will adjust for significant baseline differences among the randomized groups. Additionally, the data will be examined for both missing cases and outlier scores on measures. Variable distributions will be checked for normality and if necessary, transformations will be performed to normalize the distributions.

Statistical Analysis Plan Aim #1 Analyses for Adaptations: We will use a blend of CSP and the ADAPT-ITT procedures to adapt CRAFT in partnership with this tribal community. Digital recordings from the four focus groups will be transcribed and reviewed for themes and important areas for cultural adaptations using NVivo software.

Aim #2: Analyses for pilot feasibility RCT (N=8) 3 month follow-up. The primary outcome of IP treatment/healing engagement will be examined using Hedges's g effect sizes to correct for small sample size bias with 95% confidence intervals of the effect sizes.

Aim #2a RCT (N=40): To estimate the effect of CRAFT-AI vs. TSF-CSO on IP treatment/healing entry (primary outcome) from intake to 3- and 6-month follow-ups. The primary outcomes of IP treatment/healing entry, and CSO functioning will be examined using an intent-to-treat analysis with a generalized linear mixed effects model with fixed effects of treatment and random effects of time. The model for IP engagement will be estimated using the binomial distribution and a logit link function. Additional analyses of CSO treatment sessions attended will be examined using a negative binomial hurdle model with a log link function, which will allow us to simultaneously test the effect of treatment on engagement (yes vs no) outcome and the count of sessions attended for CSOs with or without IP treatment entry. Demographic, substance use, and psychological variables for which the randomized groups differed significantly at baseline will be included as covariates in all models. In addition, attendance at other mutual help groups such as Nar/Al-Anon will be assessed and used as a covariate as necessary, especially for the CRAFT-AI CSOs. Models will be estimated using an intent-to-treat approach that analyzes data from all randomized participants. We will use maximum likelihood estimation for all analyses, which provides the variance-covariance matrix for all available data. While attrition in the previous RCT was less than 7% at 12 months, attrition analyses will determine whether there are any differences in study variables between those with missing and complete data. Study variables associated with missing data will be covaried in all analyses.

Statistical power for Aim #2a: The primary hypothesis is that participation in CRAFT-AI will predict greater engagement of IPs in treatment/healing, as compared to those assigned to TSF-CSO. Effect size estimates were drawn from a recent meta-analysis of CRAFT. Results from the meta-analysis indicated that IP engagement rate following CRAFT was on average 67%, while engagement IP rate following TSF-CSO was on average 18%, which results in an overall effect size odds ratio of 9.25. Given this effect size for odds of IP engagement between groups with α =.05 we will have 89% power with 18 subjects per condition (n=36) to detect the main effect of treatment. Assuming 10% dropout, we propose to recruit 20 subjects per condition (n=40) to assure a final sample size of 36.

Analyses for Aim #2b: To estimate the effect of CRAFT-AI and TSF-CSO on CSO functioning and relationship functioning (secondary outcomes) from intake to 3- and 6-month follow-ups. The secondary outcomes of CSO functioning and relationship functioning will be examined using latent growth curve modeling. We hypothesize that the slope of CSO functioning and relationship functioning will indicate that participants in both CRAFT-AI and TSF-CSO will show significant decreases in depression and anxiety and significant improvement in relationship functioning regardless of whether their IP enters treatment or healing ceremony.

Statistical power for Aim #2b: A Monte Carlo simulation study was conducted to determine power for the latent growth curve models. The population parameter values for data generation and coverage were taken from the parameter estimates of an analysis of prior alcohol treatment outcome studies. We are primarily interested in testing the hypothesis that the slope would be significantly greater than zero. Based on a sample size of 40 across both conditions, we will have power greater than .80 (α<.05) to detect a slope that is statistically different from zero.

Analyses for Aim #2c: While underpowered to identify all but large mediation effects, we plan to obtain estimates of effect sizes for future studies that will examine possible mechanism of change for CRAFT-AI and potential treatment moderators. Effect sizes for possible mechanisms of change will be derived from mediation models estimated using the product of coefficients method. Given that the distribution of the product of coefficients can be non-normal we will use bootstrapping to obtain 95% confidence intervals of the mediated effect. We will have power greater than .80 to detect large mediating effects with a final sample size of 36. Effect sizes for potential treatment moderators will be derived by including interaction terms (treatment x moderator) in the Aim 2a and Aim 2b analyses.