AAT-App Outpatient Trial

Overview

Cognitive biases contribute to the difficulty experienced by heavy drinkers wishing to reduce their alcohol use. Recent interventions designed to reduce cognitive biases demonstrate efficacy for Approach Bias Modification (ApBM). Reductions in the likelihood of relapse have been found after ApBM in Alcohol Use Disorder (AUD) patients during residential treatment. Current methods of ApBM are usually delivered by computer and joystick and come with several limitations, including accessibility. If ApBM could be shown to be feasible in other settings, such as outpatient treatment, it could benefit a much larger population with AUD. This randomised controlled trial will test the efficacy of a recently-developed ApBM smartphone app called "AAT-App" ("Alcohol Avoidance Training App"). We aim to test whether AAT-App, relative to a minimal version of the app which excludes ApBM training, is effective at reducing alcohol use, cravings, severity of dependence, and approach bias (a measure of a person's automatic tendency to automatically approach alcohol-related stimuli), and to explore user experiences of AAT-App to guide future improvements to the app and its implementation.

Aims: We aim to determine whether providing the "active" version of AAT-App (including ApBM) to people accessing outpatient alcohol treatment from AOD services, is effective, relative to a minimal version of the app which does not include ApBM training, at reducing alcohol use, cravings, and severity of dependence. We also aim to test the psychometric properties of the mobile phone approach bias measure we developed, and whether engaging in AAT-App significantly reduces approach bias, relative to the minimal control condition. We also aim to explore user experiences of AAT-App's, perceived alignment with treatment, and suggested improvements, using qualitative interviews in a subset of participants, to guide further future improvements to the app and its implementation. Design \& Setting: We will conduct a double-blind, randomised, parallel-group controlled superiority trial with alcohol treatment outpatients in Melbourne, Victoria. Participants will be randomised to one of 2 groups with a 1:1 allocation ratio. Both participants and researchers completing the follow-up interviews will be unaware (blinded) of the condition to which they have been randomly allocated. Recruitment: Staff at participating services will briefly explain the opportunity to participate in a study testing a "brain-training" app to clients who they believe may meet eligibility criteria. To preserve blinding, no specific details will be provided regarding approach bias modification, and participants will merely be informed that the study tests "a new smartphone 'brain-training app'" that "involves doing brief game-like tasks on your phone each week for 4 weeks". Clients who are interested will need to provide consent to be contacted by the research team, who will conduct screening for eligibility, provide more complete participant information, and conduct recruitment. Screening: Research staff will phone clients who have provided consent to be contacted to confirm eligibility (administering the Alcohol Use Disorders Identification Test (AUDIT), checking age and phone operating system, confirming that they have no plans to enter residential/inpatient treatment in the next month, confirming absence of past-month residential rehabilitation and any past-week inpatient treatment (e.g., hospitalisation or residential withdrawal), as well as asking about number of days spent in inpatient treatment in the past month if the person underwent any inpatient treatment prior to the previous week). Randomisation and blinding of allocation: Six computer-generated randomisation sequences (one for each recruitment site) will be produced by a data scientist who is not otherwise involved in recruitment or data collection, processing, or analysis, using a 1:1 allocation ratio, based on blocks of variable size (ranging from 2-6). As such, randomisation will be stratified by site. The app developers (ANT Development Studios Ltd.) will provide the data scientist with two lists of app access codes which will direct participants to either the intervention or minimal version of AAT-App when they first download and open the app. Using these lists and the randomisation sequence, the data scientist will generate a separate spreadsheet of access codes for each site (based on the site-stratified randomisation sequence). Researchers will send these codes to participants as they are recruited from each respective site. Researchers involved in recruitment will only have access to a spreadsheet displaying a single list of codes to be sent to participants at each respective site, while the randomisation sequence will be stored in a password-protected file provided to the trial statistician and to one research officer who will remain unblinded to assist with coordinating qualitative interviews. Neither the randomisation file, nor its password will be provided to any other staff involved in recruitment or in pursuing follow-ups or quantitative data management until all data analysis is complete. Data collection windows for post-intervention and follow up assessments: 1. Post-intervention (28 days post-baseline): Participants will be provided with an in-app link to a Qualtrics survey, in which the link will expire if the survey is not completed within 7 days (i.e., 28-35 days post-baseline). 2. 1-month follow up (56 days post-baseline): Participants will be prompted with an app notification reminder to report their past-month alcohol consumption via the in-app calendar and complete a Qualtrics survey, in which the link will expire after 2 weeks (i.e., 56-70 days post-baseline). 3. 3-month follow up (112 days post-baseline): Participants will be prompted with an app notification reminder to report their past-month alcohol consumption and complete a Qualtrics survey, in which the link will expire after 28 days (i.e.,112-140 days post-baseline). Statistical analyses: Primary outcome: A linear mixed-effects model (LMM) will be used to compare change in mean standard drinks consumed per week between groups across 4 time points (baseline, post-intervention, 1-month follow-up, and 3-month follow-up). This model will test the main effects of time and group and (most crucially for determining efficacy) the group x time interaction. Planned follow-up comparisons between groups at post-intervention, 1-month follow-up, and 3-month follow-up time-points will be conducted using t-tests, with post-intervention being the primary endpoint. A secondary sensitivity analysis will be conducted excluding a participant's data from any time-point where they had been in residential/inpatient treatment in the past week (i.e., restricting analyses to time-points where a participant's opportunity to drink had not been limited by hospitalisation, rehabilitation, etc.). If a difference between groups is found post-intervention, a secondary LMM analysis of difference between groups in change in weekly standard drinks during the intervention period (i.e., 5 levels of time: Baseline, week 1, week 2, week 3, post-intervention) will be conducted to examine how quickly differences between groups emerge, with t-tests used to compare groups at week 1, 2, 3, and post-intervention time-points. This secondary analysis will be conducted excluding participants who had engaged in any residential/inpatient treatment within the intervention period. Secondary outcomes: Continuous outcome variables (Craving Experience Questionnaire frequency scale (CEQ-F) scores, Severity of Dependence Scale scores, AUDIT scores, past-week heavy drinking days (HDDs), past-month drinking days, Australian Treatment Outcome Profile quality of life items, and approach bias) will be analysed in a similar manner to the primary analysis described above. Sensitivity analysis of past-week drinking days to control for the possible effect of past-week residential/inpatient treatment will be conducted as described for the primary outcome. Past-month drinking days will be expressed as a percentage of the total number of days on which a participant had the opportunity to drink (i.e., if a participant was in residential/inpatient treatment for 10 days at a certain time point, then for that time point, their past-month drinking days will be expressed as a proportion of the remaining 18 days on which they had the opportunity to drink), although data will be excluded at any time point where the participant did not have at least 14 days on which they had the opportunity to drink alcohol (i.e., if they were within residential/inpatient treatment on 15 or more days). For CEQ-F, secondary analyses will be conducted for each of its 3 subscales. Analyses of AUDIT scores will use 2 levels of time (baseline, 3-month follow-up). Approach bias analyses will also use 2 levels of time (baseline, post-intervention) and approach bias scores will be analysed separately for alcohol and positive images. As past-month HDDs can only be calculated at 1 time point (post-intervention), they will simply be compared between groups using a t-test post-intervention after converting scores to a proportion of days on which the participant had an opportunity to drink (as described for past-month drinking days). Proportions of groups reporting complete past-week and past-month abstinence will be compared between groups at post-intervention, 1-month follow-up, and 3-month follow-up using Pearson's chi-squared. Mobile Application Rating Scale 'functionality', 'aesthetics', and 'subjective quality' scores, and participants' subjective ratings regarding AAT-App's effect on drinking and cravings will be explored within each group separately using descriptive data (e.g., mean, median, quartile cut-offs, percentages scoring above 3) to quantify typical ratings and proportions of participants providing favourable ratings. Where relevant, exploratory analyses will also compare mean scores of uMARS scales between groups using t-tests. Qualitative study: Interview transcripts will be subjected to a thematic qualitative analysis in order to identify underlying themes and patterns within each respondent's discourse. Thematic analysis will proceed according to the six stage process described by Braun and Clark. The coding process will be primarily conducted by two researchers, and a third researcher will oversee and verify coding decisions to ensure agreement and consistency throughout the process. The digital qualitative data analysis software NVivo 11.4.0 will be used to facilitate qualitative data analysis. Free-text responses to app acceptability questions in the post-intervention survey will be reviewed primarily to identify reports of functionality issues (e.g., "bugs" in the app) and safety issues (e.g., reports of triggering). Any such issues will be catalogued to inform further improvements to the app (if necessary) and its delivery/implementation in future research and/or treatment contexts.