Online Alcohol Assessment and Feedback for Hazardous and Harmful Drinkers: Findings From the AMADEUS-2 Randomized Controlled Trial of Routine Practice in Swedish Universities

Background Previous research on the effectiveness of online alcohol interventions for college students has shown mixed results. Small benefits have been found in some studies and because online interventions are inexpensive and possible to implement on a large scale, there is a need for further study. Objective This study evaluated the effectiveness of national provision of a brief online alcohol intervention for students in Sweden. Methods Risky drinkers at 9 colleges and universities in Sweden were invited by mail and identified using a single screening question. These students (N=1605) gave consent and were randomized into a 2-arm parallel group randomized controlled trial consisting of immediate or delayed access to a fully automated online assessment and intervention with personalized feedback. Results After 2 months, there was no strong evidence of effectiveness with no statistically significant differences in the planned analyses, although there were some indication of possible benefit in sensitivity analyses suggesting an intervention effect of a 10% reduction (95% CI –30% to 10%) in total weekly alcohol consumption. Also, differences in effect sizes between universities were seen with participants from a major university (n=365) reducing their weekly alcohol consumption by 14% (95% CI –23% to –4%). However, lower recruitment than planned and differential attrition in the intervention and control group (49% vs 68%) complicated interpretation of the outcome data. Conclusions Any effects of current national provision are likely to be small and further research and development work is needed to enhance effectiveness. Trial Registration International Standard Randomized Controlled Trial Number (ISRCTN): 02335307; http://www.isrctn.com/ISRCTN02335307 (Archived by WebCite at http://www.webcitation.org/6ZdPUh0R4).


INTRODUCTION 2a-i) Problem and the type of system/solution
Like this: "Existing evidence of the effectiveness of online interventions with students is mixed [6][7][8], partly because of unresolved methodological challenges including attrition prevention and assessment reactivity" 2a-ii) Scientific background, rationale: What is known about the (type of) system Like this: "Existing evidence of the effectiveness of online interventions with students is mixed [6][7][8], partly because of unresolved methodological challenges including attrition prevention and assessment reactivity" METHODS 3a) CONSORT: Description of trial design (such as parallel, factorial) including allocation ratio Like this: "AMADEUS-2 aims to provide a further evaluation of the national system for online alcohol intervention used in routine practice among the student health care centres in Sweden, specifically among the key target population of hazardous and harmful drinkers." 3b) CONSORT: Important changes to methods after trial commencement (such as eligibility criteria), with reasons No changes was done 3b-i) Bug fixes, Downtimes, Content Changes No change was made 4a) CONSORT: Eligibility criteria for participants Like thia " Students who were drinking twice a month or more often, 5 standard drinks (12 grams of alcohol in Sweden) or more for men or 4 standard drinks or more for women, were deemed eligible for trial participation. " 4a-i) Computer / Internet literacy No specified in this trials. Alla participants were young students who were thought to be use to computers. 4a-ii) Open vs. closed, web-based vs. face-to-face assessments: Like this: "All students in their 2nd, 4th or 6th terms (n=54,507) at the 9 colleges/universities were sent an e-mail in March 2013 inviting them to answer a single question about their drinking. If eligible for trial participation, they were provided with information permitting informed consent, making the present study unlike AMADEUS--1. " 4a-iii) Information giving during recruitment Like this: " All students in their 2nd, 4th or 6th terms (n=54,507) at the 9 colleges/universities were sent an e-mail in March 2013 inviting them to answer a single question about their drinking. If eligible for trial participation, they were provided with information permitting informed consent, making the present study unlike AMADEUS--1." 4b) CONSORT: Settings and locations where the data were collected Like this: "The study was undertaken in 9 of 26 student health care centres in Sweden, each providing services to one university or college. These centres were selected on the basis that they had not previously been involved in randomised controlled trials in our research programme" 4b-i) Report if outcomes were (self-)assessed through online questionnaires Like this: This study used a single 2-month follow-up assessment interval, after which the control group gained access to the intervention" 4b-ii) Report how institutional affiliations are displayed Like this: " The initial e-mail routinely sent from the participating student health care centre was altered so as to invite study participation" 5) CONSORT: Describe the interventions for each group with sufficient details to allow replication, including how and when they were actually administered 5-i) Mention names, credential, affiliations of the developers, sponsors, and owners Like this: followed comprehensive normative feedback with information describing participants' alcohol use compared to their peers in Swedish universities (adjusted for sex and age group), and, if applicable, personalized advice on reducing unhealthy levels or patterns of consumption. The feedback was shown on the screen and could also be printed out by the student. A pdf version of the feedback was also e-mailed to the students immediately after closing this page. A demonstration version in English of the assessment and feedback intervention can be viewed at http://demo.livsstilstest.nu." 5-ii) Describe the history/development process Like this: " AMADEUS-1 used an unconventional trial design with students unaware they were participating in a trial [16]. We preferred not to alter the intervention content of the national system on the basis of a single evaluation study with an unconventional trial design. These considerations led us to design the subsequent study, AMADEUS-2, as a conventional 2 arm randomised controlled trial (RCT) design targeting hazardous and harmful drinkers only, using a single screening question and no baseline assessment in order to minimise assessment reactivity [4,15]. "

5-iii) Revisions and updating
No change was done 5-iv) Quality assurance methods No specific checked. The follow--up was done by online self-assessment. 5-v) Ensure replicability by publishing the source code, and/or providing screenshots/screen-capture video, and/or providing flowcharts of the algorithms used Like this: A demonstration version in English of the assessment and feedback intervention can be viewed at http://demo.livsstilstest.nu." 5-vi) Digital preservation Like this: "A demonstration version in English of the assessment and feedback intervention can be viewed at http://demo.livsstilstest.nu."

5-vii) Access
The participants accessed the intervention on their own computer/tablett/mobile phone.

5-viii) Mode of delivery, features/functionalities/components of the intervention and comparator, and the theoretical framework
Like this: comprehensive normative feedback with information describing participants' alcohol use compared to their peers in Swedish universities (adjusted for sex and age group), and, if applicable, personalized advice on reducing unhealthy levels or patterns of consumption. The feedback was shown on the screen and could also be printed out by the student. A pdf version of the feedback was also e-mailed to the students immediately after closing this page. A demonstration version in English of the assessment and feedback intervention can be viewed at http://demo.livsstilstest.nu."

5-ix) Describe use parameters
Like this: " The study is a two-arm parallel group RCT in which routine provision of single-session online alcohol assessment and feedback intervention (Group 1) compared with non-intervention (Group 2), by delaying access to intervention for two months until research follow-up is completed. "

5-xi) Report any prompts/reminders used
This was a single-session interventions with no reminders.

5-xii) Describe any co-interventions (incl. training/support)
NA 6a) CONSORT: Completely defined pre-specified primary and secondary outcome measures, including how and when they were assessed Like this:" The primary outcome was total weekly alcohol consumption. This was computed as the sum of alcohol consumed in standard drinks for each of the 7 days in a typical week, with data on each day of the week separately provided. Secondary outcomes were the proportions of students still drinking above national guidelines [22], frequency of drinking (number of days per week), quantity of drinks per drinking day, frequency of HED as defined in the screening question, highest estimated blood alcohol concentration (eBAC) and stage of change." 6a-i) Online questionnaires: describe if they were validated for online use and apply CHERRIES items to describe how the questionnaires were designed/deployed The questions used is similar to the validated AUDIT questionnaire commonly used in alcohol research 6a-ii) Describe whether and how "use" (including intensity of use/dosage) was defined/measured/monitored Since this was a single session we did/could no measure how long time the individual spend on reading the feed-back. 6a-iii) Describe whether, how, and when qualitative feedback from participants was obtained No qualitative feedback was obtained in this study 6b) CONSORT: Any changes to trial outcomes after the trial commenced, with reasons NA 7a) CONSORT: How sample size was determined 7a-i) Describe whether and how expected attrition was taken into account when calculating the sample size To detect an effect size d of 0.1 standard deviations between the two groups with 5% significance level and 80% power required 1,600 individuals analysed per group. Assuming a follow-up rate of 50%, we aimed at recruiting 3,200 individuals per group i.e. 6,400 in total. We had no data on the number of screen positives who might be willing to participate in the trial but assumed approximately 70% would do so, meaning that we would need to identify approximately 8,000 hazardous and harmful drinkers. In order to identify these number of participants, we needed to send e-mails to approximately 40,000 students with an average response rate of 40% (i.e. n=16,000) and a 50% prevalence rate. We could not be confident of these estimates as, for example, patterns of e-mail use vary considerably between colleges, being compulsory in some institutions and rarely used in others. We therefore decided to undertake the study in 9 colleges/universities with a total student enrolment of 54,507 students. 7b) CONSORT: When applicable, explanation of any interim analyses and stopping guidelines NA 8a) CONSORT: Method used to generate the random allocation sequence Like this: "Randomisation was done using Java's built in random number generator (java.util.Random). Randomisation was thus fully computerized, did not employ any strata or blocks, and was not possible to subvert, as this and all subsequent study processes were fully automated (programmed by MB). " 8b) CONSORT: Type of randomisation; details of any restriction (such as blocking and block size) Like this: Randomisation was done using Java's built in random number generator (java.util.Random). Randomisation was thus fully computerized, did not employ any strata or blocks, and was not possible to subvert, as this and all subsequent study processes were fully automated (programmed by MB). 9) CONSORT: Mechanism used to implement the random allocation sequence (such as sequentially numbered containers), describing any steps taken to conceal the sequence until interventions were assigned Like this: Randomisation was done using Java's built in random number generator (java.util.Random). Randomisation was thus fully computerized, did not employ any strata or blocks, and was not possible to subvert, as this and all subsequent study processes were fully automated (programmed by MB). 10) CONSORT: Who generated the random allocation sequence, who enrolled participants, and who assigned participants to interventions Like this: Randomisation was done using Java's built in random number generator (java.util.Random). Randomisation was thus fully computerized, did not employ any strata or blocks, and was not possible to subvert, as this and all subsequent study processes were fully automated (programmed by MB). 11a) CONSORT: Blinding -If done, who was blinded after assignment to interventions (for example, participants, care providers, those assessing outcomes) and how 11a-i) Specify who was blinded, and who wasn't No blinding in this study 11a-ii) Discuss e.g., whether participants knew which intervention was the "intervention of interest" and which one was the "comparator" No blinding 11b) CONSORT: If relevant, description of the similarity of interventions NA 12a) CONSORT: Statistical methods used to compare groups for primary and secondary outcomes All regression analyses were performed first unadjusted and then adjusted for frequency of heavy episodic drinking at baseline, age, university, and gender, using the first two as continuous variables (thus allowing for dependence between individuals in the same university); the adjusted analysis was specified a priori as the primary result.
Tests for whether the intervention effect was modified by frequency of HED at baseline, age, university, and gender were undertaken for the primary outcome only, using the first two as continuous variables. A post hoc sensitivity analysis accounted for possible hetereogeneity between universities of treatment effects on weekly alcohol consumption using a two-stage approach. The treatment effects on weekly alcohol consumption were first estimated in each university separately by negative binomial regression (adjusted for frequency of heavy episodic drinking at baseline, age, and gender, using the first two as continuous variables) and were then combined in a random-effects meta-analysis. 12a-i) Imputation techniques to deal with attrition / missing values Missing outcome data were initially handled by a complete-cases analysis, assuming that the data were missing at random (MAR). If data are missing not at random then non-responders differ systematically from responders, and early responders are likely to differ systematically from late responders, who are likely to be more similar to non-responders [23]. We therefore explored the plausibility of the MAR assumption by regressing the primary outcome on the number of follow-up emails needed before an individual responded, using a negative binomial regression in responders: a significant association would cast doubt on the MAR assumption. To allow for the possibility of data being missing not at random, we fitted the repeated attempts model of Jackson et al. [26]. This model is not available in standard software for negative binomial regression, so we applied it to a linear regression of log (alcohol consumption + k), where k=24 units/week was chosen to eliminate skewness.

12b) CONSORT: Methods for additional analyses, such as subgroup analyses and adjusted analyses
Tests for whether the intervention effect was modified by frequency of HED at baseline, age, university, and gender were undertaken for the primary outcome only, using the first two as continuous variables. A post hoc sensitivity analysis accounted for possible hetereogeneity between universities of treatment effects on weekly alcohol consumption using a two-stage approach. The treatment effects on weekly alcohol consumption were first estimated in each university separately by negative binomial regression (adjusted for frequency of heavy episodic drinking at baseline, age, and gender, using the first two as continuous variables) and were then combined in a random-effects meta-analysis. RESULTS 13a) CONSORT: For each group, the numbers of participants who were randomly assigned, received intended treatment, and were analysed for the primary outcome Seen in figure 1 13b) CONSORT: For each group, losses and exclusions after randomisation, together with reasons Seen in figure 1 13b-i) Attrition diagram NA 14a) CONSORT: Dates defining the periods of recruitment and follow-up No 14a-i) Indicate if critical "secular events" fell into the study period NA 14b) CONSORT: Why the trial ended or was stopped (early) NA

15) CONSORT: A table showing baseline demographic and clinical characteristics for each group
Tabel 1 and 2 15-i) Report demographics associated with digital divide issues Table 1 and 2 16a) CONSORT: For each group, number of participants (denominator) included in each analysis and whether the analysis was by original assigned groups 16-i) Report multiple "denominators" and provide definitions NA 16-ii) Primary analysis should be intent-to-treat Like this: " All outcomes were compared between randomised groups under the intention -to -treat principle (that is, including all randomised individuals in their originally randomised groups). " 17a) CONSORT: For each primary and secondary outcome, results for each group, and the estimated effect size and its precision (such as 95% confidence interval) Seen in table 3 and figure 2 17a-i) Presentation of process outcomes such as metrics of use and intensity of use NA-singel session 17b) CONSORT: For binary outcomes, presentation of both absolute and relative effect sizes is recommended NA

18) CONSORT: Results of any other analyses performed, including subgroup analyses and adjusted analyses, distinguishing pre-specified from exploratory
In the assessment for skewness of the continuous variables we found 1 outlier in the treatment group (with weekly alcohol consumption of 1044 g/week) and 2 outliers in the control group (with weekly alcohol consumption of 1128 and 1524 g/week). The maximum reported weekly alcohol consumption of those not excluded was 552 g/week in the treatment group, and 456 g/week in the control group. We therefore performed a sensitivity analysis without these outliers. In this analysis, which was not specified a priori, the between-group difference in the primary outcome, weekly alcohol consumption, in the primary adjusted analysis, crossed the conventional threshold for statistical significance (8% reduction (95% CI 0% to 15% reduction, p=0.049 adjusted; 10% reduction (95% CI 1% to 17% reduction), p=.02 unadjusted) 18-i) Subgroup analysis of comparing only users NA 19) CONSORT: All important harms or unintended effects in each group NA

19-i) Include privacy breaches, technical problems
No technical problems in this study 19-ii) Include qualitative feedback from participants or observations from staff/researchers NA DISCUSSION

20) CONSORT: Trial limitations, addressing sources of potential bias, imprecision, multiplicity of analyses 20-i) Typical limitations in ehealth trials
Like this: We did not have statistical power to detect the effect size that we believed was worth obtaining in the planning of this study. We succeeded in recruiting only one quarter of our target sample size, and the best estimate of the effect obtained on the primary outcome (a 6% reduction in alcohol consumption) is of clear public health significance. For example, it is very close to the size of effects considered appropriate for the implementation of face-to-face brief intervention programmes [25] and measures to increase the price of alcohol [26]. " 21) CONSORT: Generalisability (external validity, applicability) of the trial findings 21-i) Generalizability to other populations Like this: The effects are also broadly similar, for example with THRIVE showing a 17% reduction in alcohol consumption in comparison with controls after one month, decreasing to 11% after 6 months [30]. Some effects among Maoris were somewhat larger, though effects among non-Maori in the parallel e-SBINZ trial were smaller [31][32].  found that feedback added little to the effects of assessment, and detailed investigations of intervention content are urgently needed, in order to ascertain whether it may be possible to develop novel interventions or intervention components capable of larger effects than have been identified to date. This under-development of online study is similar to that which pertains for face-to-face brief interventions [33]. 21-ii) Discuss if there were elements in the RCT that would be different in a routine application setting Our study used the same elements and procedure as routine provision of the interventions to students in Sweden 22) CONSORT: Interpretation consistent with results, balancing benefits and harms, and considering other relevant evidence 22-i) Restate study questions and summarize the answers suggested by the data, starting with primary outcomes and process outcomes (use) Like this: Table 3 reveals, however, that this study does not rule out an intervention effect of up to 13% reduction in total weekly alcohol consumption. The sensitivity analysis excluding outliers suggests an intervention effect on reduced total weekly alcohol consumption, though the statistical significance attained in that analysis should be treated with caution, as the particular analysis was not pre-specified. We have no reason to anticipate later occurring effects, as brief intervention effects generally wane with time [5] and the short term nature of this evaluation study is important to note in interpreting study findings."

22-ii) Highlight unanswered new questions, suggest future research
Like this: "The existing evidence suggests that these types of alcohol policies are most likely to be effective and we do not know whether or how far individual-level interventions in whole populations such as that evaluated here may enhance the anticipated effects [36].

X26-i) Comment on ethics committee approval
Like this: " Ethical approval for the study was granted by the Regional Ethical Committee in Östergötland, Sweden, number: 2013/46-31. " x26-ii) Outline informed consent procedures Like this: " If eligible for trial participation, they were provided with information permitting informed consent, making the present study unlike AMADEUS--1. " X26-iii) Safety and security procedures NA X27-i) State the relation of the study team towards the system being evaluated Declaration of interest PB and MB own a company that has developed the online intervention used in this study and that develops and distributes computerised lifestyle interventions.