Trial design

In the current preference trial, patients completed an 8-week ICBT program for alcohol misuse, named the Alcohol Change Course (ACC). After a brief screening interview, eligible patients were given a choice between (a) a therapist-guided version, where patients could message an assigned therapist who responded on a pre-determined day each week or (b) a self-guided version, where patients worked on the program on their own. The trial was conducted via the Online Therapy Unit (OTU) website (www.onlinetherapyuser.ca). The OTU is based at the University of Regina, SK, Canada, offering free therapist-guided ICBT to Saskatchewan residents with funding awarded by the Saskatchewan Ministry of Health. In addition, the self-guided ACC is available to all Canadian residents. The current study was registered at https://clinicaltrials.gov/ct2/show/NCT04611854 (NCT04611854) and approved by the University of Regina Ethics Review Board (approval number 2019-058). All patients signed an online informed consent form before enrolling in the study.

Patients

Patients were recruited through several sources, including online advertisements, emails, posters distributed to SK physicians/doctors, referrals from SK health regions, and word of mouth from friends and family members. These referral sources directed interested individuals to the OTU website, where they could read about the ACC and apply to take the course through an online screening questionnaire. The questionnaire included a consent form and questions regarding applicants’ contact information (e.g., telephone number, email address), demographic information (e.g., gender, ethnicity, education), relevant personal details (e.g., medical history), mental health (e.g., depression, anxiety), and alcohol use. Upon completing the screening, applicants meeting initial inclusion criteria were directed to an online booking program to schedule a telephone screening appointment with OTU staff. During the screening call, applicants answered follow-up questions to confirm their eligibility. Patients eligible for the preference trial had to: (a) be ≥ 18 years old; (b) be a SK resident; (c) have Internet access; (d) have consumed ≥ 14 drinks in the previous week (i.e., a cut-off used in similar previous research; i.e., [13]); and (e) indicate hazardous or harmful alcohol consumption by scoring ≥ 8 on the Alcohol Use Disorder Identification Test [18]. Additionally, applicants were excluded if they (a) scored ≥ 25 on the Drug Use Disorder Identification Test [DUDIT; [19]]; (b) scored a three on item 9—asking about suicidal ideation—of the 9-item Patient Health Questionnaire [PHQ-9; [20]]; (c) had unmanaged symptoms of bipolar disorder, schizophrenia, and/or psychosis; (d) had a medical condition that would inhibit active participation in the course; (e) received concurrent mental health treatment more than twice monthly in the last three months (not including taking psychotropic medication[s]); or (f) had been hospitalized for mental health reasons in the past year.

Patients that met all of the listed criteria were eligible to choose between the self-guided and therapist-guided ACC and were subsequently analyzed in the current trial. Additionally, those patients who met all of the criteria, apart from residing in SK, were offered self-guided ACC and were included in the exploratory analyses of the current study.

Intervention: the alcohol change course

The ACC is an ICBT program for alcohol misuse, developed initially in Switzerland [21,22,23]. The intervention was translated to English [24] and subsequently adapted to address young adults’ alcohol use and depression [24, 25]. As described previously [12], the ACC was further modified for use in the OTU, being revised to focus on adult populations and to align with the other evidence-based ICBT programs offered by the OTU. The updated version of the ACC contained 12 lessons—delivered consecutively over eight weeks—comprised of slide shows with psychoeducation, case stories, and downloadable worksheets for practicing skills. For the current preference trial, a patient-oriented working group (i.e., three patients with personal or familial lived experience with alcohol misuse, two Internet therapists, two operations managers, two trainees, and two group facilitators) collaborated in updating the ACC based on patient feedback, altering the language to be more inclusive and condensing the course into eight lessons, one lesson per week [26]. Furthermore, information regarding abstinence, the relationship between physical health and drinking, as well as Canada’s alcohol use prevalence rates and low-risk drinking guidelines was incorporated into the introductory lesson.

Treatment guidance

Eligible patients were given a choice between the therapist-guided and self-guided versions of the ACC. Regardless of patients’ preferred treatment condition, they received automated, weekly emails containing information regarding new lesson content, and they were also contacted by the OTU if they wanted to discontinue the ACC and/or experienced technical difficulties with the treatment platform. Moreover, patients in both the therapist- and self-guided ACC responded to weekly surveys prompting them to reflect on their challenges and learning experiences during the past week.

Apart from these specific situations, patients who chose the self-guided ACC received no regular contact with Internet therapists. In contrast, patients who chose the therapist-guided ACC were supported by one of two Internet therapists; these therapists held graduate degrees as a Master of social work and a Master of Education in counselling psychology, and had practiced for 14 and two years, respectively. Through the messaging functioning of the online treatment platform, therapists contacted patients once per week on a pre-set day, spending ~ 15 weekly minutes connecting with each therapist-guided patient to help them manage motivation, reinforce lesson completion, and answer questions. Therapists also communicated with these patients via telephone calls in rare cases where (a) patients demonstrated increased suicide risk, (b) patients requested a call, and/or (c) a misunderstanding between therapists and patients was to be addressed.

InstrumentsPrimary outcome measures

At pre-treatment (baseline screening), mid-treatment (Week 4), post-treatment (Week 8), and follow-up (Week 20), total preceding week alcohol consumption and number of heavy drinking days (HDD) were assessed using the well-standardized Timeline Follow-Back [TLFB; [27]]. Participants reported the number of standard drinks (i.e., one 12 oz can/bottle of 5% beer, cider, or alcopop/cooler; one 4.5 oz glass of 12% wine; or one 1.3–1.5 oz shot of 40% hard liquor) that they had consumed during the past seven days at each measurement period. The seven daily values were summed to calculate the number of total preceding week drinks (i.e., the TLFB variable). Cronbach’s α for the TLFB ranged from 0.74 to 0.82 in the current study. Further, separated by gender, the number of days when women consumed more than three daily drinks and when men consumed more than four daily drinks in the past seven days were summed to calculate their total number of preceding week HDDs.

Secondary outcome measures

All secondary outcome measures were assessed at pre-treatment, post-treatment, and follow-up using well-standardized self-report questionnaires.

Alcohol misuse was assessed through the Alcohol Use Disorder Identification Test [AUDIT; [18]]. Patients responded to 10 items on scales from 0 to 4; responses were summed to produce a total score from 0 to 40. Higher scores indicate greater alcohol-related difficulties and behaviours. Scores ≥ 8 indicate hazardous or harmful alcohol consumption, and scores ≥ 15 indicate a possible alcohol use disorder. In the current study, Cronbach’s α for the AUDIT ranged from 0.77 to 0.83.

Alcohol craving was assessed via the Penn Alcohol Craving Scale [PACS; [28]]. Patients responded to five items on 7-point scales from 0 to 6. Items were summed to produce a total score ranging from 0 to 30, with higher scores indicating greater alcohol craving. Cronbach’s α for the PACS ranged from 0.91 to 0.92 in the current study.

Patients’ confidence in their ability to resist alcohol cravings was assessed by the Brief Situational Confidence Questionnaire [BSCQ; [29]]. Patients responded to items referencing eight situations (i.e., negative emotional states, negative physical states, positive emotional states, testing personal control, urges and temptations, interpersonal conflict, social pressure, and positive social states), each measured on a scale from 0 to 100. Item responses were summed to produce a total score from 0 to 800. Higher scores indicate greater confidence in one’s abilities to resist alcohol cravings. In the current study, Cronbach’s α for the BSCQ ranged from 0.87 to 0.91.

Anxiety symptoms were assessed with the Generalized Anxiety Disorder-7 [GAD-7; [30]]. Patients responded to seven items on 4-point scales from 0 to 3. Responses were summed to produce a total score, ranging from 0 to 21. Higher scores indicate more severe self-reported anxiety, with scores > 9 indicating clinical levels of anxiety. Cronbach’s α for the GAD-7 ranged from 0.89 to 0.90 in the current study.

Depression symptoms were assessed via the 9-item Patient Health Questionnaire [PHQ-9; 20]. Patients responded to nine items on 4-point scales from 0 to 3. Items were summed to produce a total score from 0 to 27. Higher scores indicate greater depression symptom severity, with scores > 9 indicating a possible major depressive disorder. In the current study, Cronbach’s α for the PHQ-9 ranged from 0.87 to 0.91.

Functional impairment was assessed with three items from the Sheehan Disability Scale [SDS; [31]]. Patients responded to three items (on 11-point scales from 0 to 10) assessing work/school, social, and family life functional impairment. Responses were summed to yield a total score ranging from 0 to 30, where higher scores indicate greater total functional impairment. Cronbach’s α for the SDS ranged from 0.82 to 0.93 in the current study.

Additional measures

At baseline screening, patients completed the Drug Use Disorder Identification Test [DUDIT; [19]], an 11-item self-report scale with total scores ranging from 0 to 44. The first nine items are rated 0, 1, 2, 3, and 4, while the last two items are rated 0, 2, and 4. A score of ≥ 25 suggests significant problem with drugs and was used to exclude patients from the current trial. The Cronbach’s α was 0.74 for the DUDIT in the current trial.

At baseline screening, patients’ motivation to change their drinking and current stage of change (i.e., pre-contemplation, contemplation, action) were assessed by the revised version of the self-report Readiness to Change Questionnaire—Treatment Version [RCQ-TV; [33]]. Twelve items were scored on a 5-point scale ranging from − 2 = strongly disagree to 2 = strongly agree. Item responses were used to calculate three sum scores ranging from − 8 to 8 for pre-contemplation (Cronbach’s α = 0.83), contemplation (Cronbach’s α = 0.70), and action (Cronbach’s α = 0.92). Patients were considered to be in the stage of change where they received their highest sum score. See Heather and Hönekopp [33] for further information regarding the revised RCQ-TV scoring.

At mid-treatment, patients’ perceived treatment credibility and expected treatment success were assessed via the 6-item Credibility and Expectancy Questionnaire [CEQ; [32]]. Patients responded to the first three items, each on a 9-point scale ranging from 1 = not at all logical/useful/confident to 9 = very logical/useful/confident; these items were summed to produce a total score ranging from 3 to 27, with higher scores indicating greater perceived treatment credibility. Further, patients responded to the last three items; items 4 and 6 were on 11-point scales ranging from 0 to 100 (coded as 0–10), and item 5 was on a 9-point scale ranging from 1 to 9. These three items were summed to produce a total score ranging from 1 to 29, with higher scores indicating greater treatment expectancy. Cronbach’s α = 0.80 for the CEQ in the current study.

At post-treatment, patients’ satisfaction with and negative effects experienced during the ACC were assessed via a self-report questionnaire developed by the OTU research team [see [26]]. Patients responded to 10 items measuring their evaluations of the treatment (e.g., “Would you feel confident recommending this treatment to a friend?”) and negative effects they perceived experiencing as a result of the treatment (e.g., “Have you experienced any unwanted negative effects or events that you associate with taking part in this online treatment?”).

Further, as a proxy for engagement, patients’ course completion was recorded throughout their eight weeks of treatment. Total values representing the proportion of completed lessons and the proportion of overall treatment completion were calculated for each patient.

Statistical analyses

Statistical analyses were conducted using IBM SPSS Statistics (Version 27.0). Descriptive statistics described patients’ characteristics in percentages, means, and standard deviations. The pre-treatment characteristics and post-treatment engagement and satisfaction of the groups (i.e., therapist- versus self-guided) were compared with Chi square tests for categorical variables and t tests for continuous variables. Following previously established methodology [34,35,36], a series of mixed model analyses were conducted to examine changes in the primary (i.e., TLFB, HDD) and secondary (i.e., AUDIT, PACS, BSCQ, GAD-7, PHQ-9, SDS) outcomes over time and to assess if these changes differed between the groups. Of note, this approach was chosen as mixed-model analyses can produce accurate inferences with small samples [see 41, 42]. For each outcome, a series of models involving the fixed and random effects of intercept (i.e., symptom scores at pre-treatment) and slope (time) were conducted and included in the model to account for the correlated nature of the data. The fixed-effect models included time, the preference group (group), and their interaction (time × group). Intraclass correlation coefficients were manually calculated to determine if mixed model analyses were appropriate [34]. Various within-individual covariance structures (e.g., scaled identity, diagonal, autoregressive, unstructured) were also tested. The models with smallest Akaike’s Information Criterion and Bayesian Information Criterion were selected for the final analysis. Estimates were calculated using the full information maximum likelihood method with the Satterthwaite approximation for the denominator’s degrees of freedom. When significant differences emerged between the preference groups, baseline demographic characteristics were included in the model as covariates to adjust for potentially confounding effects of these variables on treatment outcomes. We also controlled for the potential effects of other clinically relevant independent variables such as concomitant use of psychotropic medications, number of lessons completed (weighted by the actual number of lessons completed), treatment expectancy and credibility, and readiness to change (i.e., pre-contemplation, contemplation, action) by covarying the variables in each of the final models.

Via a series of mixed-models, an exploratory analysis assessed the change in outcomes among the patients who resided outside of SK and, therefore, were automatically assigned to the self-guided ACC as they were not eligible to select their treatment preference. Further, using Chi square tests or t tests, post-treatment engagement and satisfaction were compared between the self-guided patients residing outside of SK and the preference trial groups.

Missing data management

For the preference trial patients, there were no missing values across all pre-treatment variables but there were 17 (23.0%) and up to 29 (39.2%) patients with missing values in the primary and secondary outcome measures, respectively, at post-treatment. Data were missing from 30 (40.5%) patients at follow-up. There were no significant differences in attrition between therapist- and self-guided preference groups at mid-treatment, post-treatment, and follow-up for the primary outcome variables (p range: 0.11–0.85). The analysis of missingness with Little’s Missing Completely at Random test (χ2 = 199.86, df = 206, p = 0.61) suggested that the data were missing at random [37]. Following a modified intention-to-treat approach, the missing data were imputed using the multiple imputation method, generating 10 multiply imputed datasets so that the data from all eligible preference trial patients were analyzed [38]. Pooled results are presented for all mixed model analyses. Effect sizes, Cohen’s d [39]—for the difference between pre- and post-treatment and pre-treatment and follow-up assessments on the TLFB and HDD variables—were computed using estimated marginal means.