Effects of an automated digital brief prevention intervention targeting adolescents and young adults with risky alcohol and other substance use: study protocol for a randomised controlled trial.

INTRODUCTION
Adolescence and young adulthood is a period in life when individuals may be especially vulnerable to harmful substance use. Several critical developmental processes are occurring in the brain, and substance use poses both short-term and long-term risks with regard to mental health and social development. From a public health perspective, it is important to prevent or delay substance use to reduce individual risk and societal costs. Given the scarcity of effective interventions targeting substance use among adolescents and young adults, cost-effective and easily disseminated interventions are warranted. The current study will test the effectiveness of a fully automated digital brief intervention aimed at reducing alcohol and other substance use in adolescents and young adults aged 15 to 25 years.


METHODS AND ANALYSIS
A two-arm, double-blind, randomised controlled trial design is applied to assess the effectiveness of the intervention. Baseline assessment, as well as 3-month and 6-month follow-up, will be carried out. The aim is to include 800 participants with risky substance use based on the screening tool CRAFFT (Car,Relax, Alone, Forget, Friends, Trouble). Recruitment, informed consent, randomisation, intervention and follow-up will be implemented online. The primary outcome is reduction in alcohol use, measured by Alcohol Use Disorders Identification Test total score. Secondary outcomes concern binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, average daily drinks per typical week, other substance use, mental health, sexual risk behaviours and perceived peer pressure. Moreover, the study involves analyses of potential moderators including perfectionism, openness to parents, help-seeking and background variables.


ETHICS AND DISSEMINATION
The study was approved by the Swedish Ethical Review Authority (no. 2019-03249). The trial is expected to expand the knowledge on digital preventive interventions for substance using adolescents and young adults. Results will be disseminated in research journals, at conferences and via the media.


TRIAL REGISTRATION NUMBER
24 September 2019, ISRCTN91048246; Pre-results.


INTRODUCTION
Substance use is a major public health concern causing individual suffering as well as societal costs [1][2][3].
Substance use in childhood and younger ages is particularly harmful since the brain is undergoing critical development during this time in life, which makes it more vulnerable to addictive substances [4,5]. In adolescence, some individuals are especially prone to various risk behaviours and the initiation of substance use often occurs during this period [6]. Peers are increasingly important in teens' social life and the influence of substance using peers is a prominent risk factor for one's own substance use [7]. After initiation, the use of substances frequently increases during adolescence and young adulthood, posing a number of risks for the individual [2,8,9]. Early onset of substance use implies a risk of severe adverse effects on psychological and physiological development [2,8,9]. Hazardous use of alcohol, cannabis and other substances can lead to regular consumption and development of addiction chronic problematic consumption patterns that can significantly influence the developmental trajectory during the transition from childhood to adulthood [10]. More acute or short-term consequences include, e.g., problems of academic adjustment [11], accidents [12], and problems with the police or legal authorities [13]. Risk behaviours, such as substance use, binge drinking, intoxication, and sexual risk behaviour often co-occur, and the use of lllicit substances, is often accompanied by alcohol consumption [1,2,11,[14][15][16][17][18]. The association between substance use and mental health problems is well established and bidirectional, sometimes as a result of self-medication but also as an increase in mental health problems due to substance use [19]. Poly-substance use, i.e., simultaneous use of different substances, has in previous research been associated with mental health problems, increased in magnitude and over time with the number of substances used [20].
Although alcohol use has declined among adolescents during recent years, some individuals drink more and binge drinking is still highly prevalent among young adults were the largest proportion of alcohol risk consumers is still found [21,22]. Moreover, a high prevalence of substance use and poly-substance use among young people in Europe has been recognized in the EU Drug Strategy 2005-2012, which calls for action [23]. Also, established cannabis users among Swedish upper secondary class students tend to use cannabis more often than before, according to recent survey data [24]. In light of these circumstances, it is of utmost importance to target and tailor attractive preventive interventions for adolescents and young adults. One way to reach large groups of young people is to disseminate interventions via the Internet. In Sweden, 98 percent of the population has access to the Internet at home and virtually 100 percent of [16][17][18][19][20][21][22][23][24][25] year-olds use the Internet in mobile phones, tablets or computers [25]. Digital interventions have several advantages over traditional ones. They may reduce the stigma around risk behaviours, are accessible at any time and place, and can be cost-effective because minimal staff resources are used for implementation, especially if they are fully automated [26,27]. Digital intervention may thus be particularly suitable for adolescents and young adults [28]. To date, digital prevention programs targeting alcohol and other substance use among adolescents and young adults that meet scientific evaluation criteria are scarce. An early example is eScreen.se, offering screening for alcohol and drug use with personalized feedback [29]. A more recent example is a fully automated brief motivational intervention, for substance-using [16][17][18] year-olds, which was tested in a randomised controlled trial (RCT) involving four European countries [30][31][32], where our research group was one of the partners. The intervention, named WISEteens, was based on Motivational Interviewing (MI) and demonstrated significant betweengroup effects for alcohol use, indicating that a targeted brief motivational intervention in a fully automated digital format can be effective to reduce drinking and lower barriers for existing substance use service in hazardous drinking adolescents [34].
Digital interventions generally offer adaptations of evidence based face-to-face interventions, e.g., MI.
The effects of MI on several health related behaviours have been evaluated in a number of studies, especially among adults [33]. Among adolescents, systematic reviews and meta-analyses on MI for various health-related behaviours show mixed results. A review of six studies of MI for reducing alcohol consumption in the emergency room setting suggested that MI was at least as effective as other brief interventions in the same setting, and suggested that MI could be even more effective than existing interventions [34]. Another review of 24 studies of different brief interventions for reducing alcohol consumption and alcohol-related problems among adolescents showed a significant effect that persisted up to one year compared to control interventions, with greater effects for MI. Effects were consistent over diverse settings and particularly effective components included decisional (juxtaposition of pros and cons of change) [35] and goal-setting [36]. Regarding MI for illicit drug use among adolescents, a review of 10 studies showed no effects on drug use behaviors; however, changes in attitudes towards drug use were found, which could be translated into intentions to change behaviors [35]. A scoping review concerning MI for reducing sexual risk behaviors among adolescents identified 29 unique studies with varying of designs, conceptualizations of MI and specific risk behaviours, making it difficult to generalize regarding outcomes but indicating the need of more research [37]. MI has most commonly been delivered as an individual face-to-face intervention, but has also been provided in other forms, e.g., via telephone or digitally, with various results [33,38,39]. Digital programs based on MI have demonstrated effect in the form of reduced alcohol consumption among young adults [40,41] and a combination of screening with a short intervention has shown similar effects on alcohol consumption among adolescents [42]. Brief motivational interventions are empirically supported individual level interventions for reducing alcohol use [43,44], which have also been digitally provided. Digital interventions can be provided with or without human guidance, i.e., more or less automated [45]. Previous research has demonstrated that even fully automated interventions can reduce alcohol-related problems for young people with risk consumption of alcohol up to 12 months after implementation [41,46], and indicated potential effect on cannabis use in certain groups [47,48]. Although various digital substance use prevention interventions have been developed and tested with promising results [30,49], studies on the effects of fully automated digitally delivered and MI-based substance use prevention programs for adolescents and young adults are still few [49] and more well-designed studies are warranted in order to obtain evidence for potential effects.

Aim
The study described in this protocol is informed by builds on results previously obtained in a European study of a brief MI-based substance use prevention intervention (WISEteens), digitally delivered to adolescents in a fully automated form [30]. In this study, the effectiveness of a modified version of this intervention will be tested across a range of outcome measures among adolescents and young adults. We aim to target individuals at an age when they commonly debut as alcohol or other substance users, recruiting [15][16][17][18][19][20][21][22][23][24][25] year-olds with hazardous alcohol or other substance use. The primary aim is reduction in frequency and quantity of substance use. The primary outcome is alcohol consumption at three-month follow-up. A secondary outcome includes reduction of alcohol consumption at six-month follow-up and reduction of other substance use at three and six-month follow-up, respectively. We hypothesize that participants in the intervention group will report reduced substance use (primarily alcohol consumption at three-month follow-up), with a larger effect size compared to an active control group receiving health information about various substances. A secondary aim involves analysis of intervention effects on mental health, sexual risk behaviours, perceived peer pressure, and analysis of a number of potential moderators including, perfectionism, openness to parents, help-seeking, and various background variables. The study will add to existing evidence regarding the effectiveness of digital interventions aimed at reducing alcohol and other substance use among young people.

METHODS AND ANALYSIS
In order to evaluate the effectiveness of the digital screening and brief motivational intervention, a doubleblind two-arm RCT study design is planned with baseline assessment at study entry and follow-up assessment at three and six-months across a number of outcome variables. Figure 1 displays the trial design. We will also explore and test for moderator effects. To outline and report the current study we used SPIRITreporting guidelines [50].

Screening and informed consent
Potential participants will be guided to a digital landing page with optional screening by an adapted version of the screening tool for use of alcohol and other substances, CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble) [54]. This six-item screening tool has demonstrated criterion validity and appropriateness for identifying substance-related problems among adolescents [55][56][57]. The primary eligibility criterion for participating in the study will be a score of 2 or more on the CRAFFT, since this value has shown satisfactory sensitivity for identifying substance use problems [54]. Those who score 2 or higher will be offered participation in the study and given information on confidentiality, voluntariness of participation, and human subject protections [53]. They will also be provided with contact information on suitable counseling service providers. Those who agree to participate will be guided to an online platform where they will be asked to give digital informed consent.

Assessment
All assessment will occur online. The selection of measures and associated instruments is based on the aim of the study and the theoretical base concerning factors influencing substance use, partly outlined above. While most of prior research on substance use related problems and mental health problems show that low socio-economic standard is a risk factor for substance use and related problems [58,59], the current study is informed by recent research on mechanisms involved in adolescence alcohol-and other substance use, showing associations between internalizing symptoms and substance use among high school students in affluent areas [60,61], as well as associations between achievement pressures (particularly excessive perfectionistic strivings), and isolation from parents (particularly low perceived closeness to mothers) [60,62]. The latter finding is in line with previous research showing that a positive parent-child relationship is a protective factor related to lower likelihood of adolescence substance use [63,64]. That said, the baseline assessment in the current study will cover behavioural and health-related questions for measuring various outcomes including alcohol and other substance use, mental health, sexual risk behaviours, and perceived peer pressure, as well as questions related to moderating variables including perfectionism, openness to parents, help-seeking due to mental health problems, and sociodemographic and personal characteristics, (i.e. sex, age, residence, occupation (student/not student), school performance (for students), and parents' education). Based on the previous European WISEteens study [30], where significant effects were found for alcohol consumption but not for other substance use, the primary outcome will be changes in alcohol use between baseline and both follow-up assessments.

Measures
Alcohol use will be measured using two instruments. One of them is the short version of the Alcohol Use Disorders Identification Test (AUDIT-C) [65], measuring frequency and amount of consumed alcohol and frequency of binge drinking, providing a widely used and valid index sum score for problematic alcohol use among adolescents [57,66], the other is the Daily Drinking Questionnaire (DDQ) [67], measuring the quantity and frequency of alcohol use a typical week. Previous research has demonstrated that the DDQ is highly correlated with other measures of self-reported alcohol consumption [68]. Other substance use will be measured by the first three items of the Drug Use Disorder Identification Test (DUDIT) [69], assessing frequency of consumption of drugs other than alcohol, frequency of different types of drugs other than alcohol used at the same occasion (1 = never to 5 = four or more times a week), and number of occasions when drugs other than alcohol are consumed on a typical day of drug use (1 = zero to 5 = seven or more).
The DUDIT has been found to be effective in screening for drug-related problems in clinically selected groups [70], and has proven useful in the context of public health surveys in Sweden [69,71]. Mental health will be measured by The WHO-5 Well-Being Index, which has demonstrated validity both as a screening tool for depression and as an outcome measure in clinical trials [72]. This scale has been successfully applied across a wide range of study fields and translated into more than 30 languages, including Swedish [73], and found psychometrically sound. Changes in sexual risk behaviors will be measured by multiple choice questions on sexual behaviour under the influence of alcohol and/or other substances and unprotected sex, previously used in a survey among Swedish visitors at youth health clinics in Stockholm county [74]. Changes in perceived peer pressure will be measured by two items retrieved and adapted from the Peer pressure inventory [75]. Perfectionism will be measured by two subscales of the Frost Multidimensional Perfectionism Scale (FMPS) [76,77], 2) openness to parents will be measured among participants who are up to 20 years old, by questions about disclosure from Stattin & Kerr [78][79][80], slightly modified into statements to be compatible with the digital format, and help-seeking due to mental health problems, will be measured by dichotomous questions regarding if, and to what healthcare provider, the participant has turned for help. Additionally, background data regarding sex, age, residence, occupation (student/not student), school performance (for students), and parents' education will be asked for, using multiple choice questions.

Randomisation
The study is double-blind, thus neither the participant or the researchers know which participant is allocated to the intervention or to the active control condition. After baseline assessment, the participants will be automatically randomised to one of the two study groups by a computer program using an

The intervention
Participants allocated to the intervention group will initially be asked to give additional information on alcohol and other substance use and to state their body weight, in order to generate personalized feedback on substance use behaviour and to estimate alcohol blood concentration on a typical drinking occasion.

Theoretical background of the intervention
The intervention relies on Motivational Interviewing (MI) [81] and models of Social Influence [82,83].
The goal of an MI approach in the present context is to enhance motivation to change by exploring and resolving ambivalence about substance-related behaviours. Important elements are personalized feedback about one's substance use behaviours in relation to normative comparisons, balancing of pros and cons of the target behaviour, and supporting self-efficacy. Personalized feedback, or tailored feedback, regarding risk-status and normative feedback relative to a relevant comparison group is perceived as more relevant for changing behavior than more general information [53]. Normative feedback include information about how a specific reference group actually consumes substances, in order to correct participants' "inflated perception" [43], and has been proven effective for reduction of alcohol consumption in young adults in previous meta-analyses [84,85]. For adolescents and young adults, a normative feedback approach may be particularly appealing, assuming they are curious about how their substance consumption compares to their peers [86]. Likewise, an important element of the intervention is a focus on substance-related social norms and training on how to avoid social high-risk situations and how to resist peer pressure; i.e., raising refusal self-efficacy [87]. In line with the social influence hypothesis, which states that a prominent risk factor for substance use in adolescence is the influence of substance using peers [7], targeting peer pressure in brief digital interventions has been an important element for reducing alcohol use among adolescents [88].

Description of the intervention content
The intervention is interactive, digitally delivered in a fully automated form, and requires approximately 20 minutes to complete. Tailored feedback is given to the participants based on their responses to previous assessment and suggestions on how to respond to this feedback are provided. This interactivity imitates a face-to face "dialogue" with techniques from MI such as an empathic approach, rolling with resistance, aiming at creating a dissonance between actual and desired behaviour, raising self-efficacy, and at the same time avoiding argumentation [39]. The intervention consists of three main components outlined below, and additional health related information.

Personalized feedback
The personalized feedback includes an estimation of the participants' blood-alcohol-concentration (BAC) and information on the associated risks concerning the participants' heaviest drinking episode during the past 30 days. The value will be based on a measure of Peak Drinking Quantity [89] and estimated using the Widmark-formula, which takes into account weight and gender [90]. The participants will receive graphed feedback regarding number of standard alcohol units per week that they think their peers consume (descriptive norms), as well as the participant's individual levels of consumption in relation to comparative data (actual drinking levels) from a reference group. Comparative data (AUDIT-C scores) will be taken from alcohol prevalence estimates found in a nationally representative sample of 16-25 yearolds undertaken by the Public Health Agency of Sweden. Comparative feedback will be available for drinking but not for substances other than alcohol.

Interactive MI-based exercises
The exercises provided in the current intervention build on the assumption that participants may hold certain levels of ambivalence about their current substance use, and that if they are willing to make a change they may not know how to, or may not be confident that they are able to do so [81].Therefore, the intervention uses importance and confidence rulers with a short summary and feedback to encourage participants to reflect on personal reasons for change and explore personal strengths and ability to change.
Furthermore, the program provides a decisional balance to pick up and graphically illustrate potential levels of ambivalence by offering the participants a list of possible pros and cons regarding the decision to change their current substance use [39,91]. Participants are instructed to choose statements that apply to them and are presented with the resulting balance sheet of their personal comparative potential gains and losses.

Practical advice
The intervention focuses on raising self-efficacy for being able to avoiding drinking in social situations, if desirable [88]. The participant will be asked to select three among 12 provided drinking situations that they consider most tempting and rank them (the situations are adapted from the adolescent version of the Drinking Refusal Self-efficacy Questionnaire (DRSEQ-RA) [92]. According to the selection, a number of strategies will be offered for each of the selected drinking situations to provide participants with a tool kit necessary for engaging in and maintaining their behavioural goal.

Health related information connected to substance use
Finally, the intervention program includes health related information associated with substance use. The information is provided optionally throughout the program behind "read more" buttons and also at the end

The control condition
The control group will receive the same general health related information as those in the intervention group, i.e., the additional information which is connected to the intervention program.

Statistical analysis
Randomisation-checks of baseline variables regarding alcohol and other substances consumption and psychological state will be conducted using multivariate analyses of variance MANOVA. To test the effectiveness of the intervention, we will assess whether participants in the intervention group report decreased substance use, sexual risk behaviours, perceived peer-pressure, and improved their mental health after three and six months, respectively, compared to participants in the active control group. Data analyses will consist of comparing outcome measurements with regards to within-group and betweengroup differences according to the per protocol as well as the intention-to-treat (ITT) principle, accounting for all included participants regardless of whether or not they completed follow-up assessments. The main analysis of effectiveness will use mixed effects regression models, which can be applied to both continuous and categorical outcomes and also non-normally distributed outcomes. Moreover, mixedeffects regression models are robust to missing data in longitudinal studies [93]. In the current study we assume a quite large dropout rate with data missing at random (MAR), which can be handled using mixed effects regression models. Separate models will be run to test each outcome, i.e., alcohol use, other substance use, mental health, sexual risk behaviours and perceived peer pressure. Effect sizes will be calculated separately for the two measures used for the primary outcome variable (AUDIT-C and DDQ).
Potential moderators, including perfectionism, openness to parents, help-seeking due to mental health problems, and socio-demographic and personal characteristics, (i.e. sex, age, residence, occupation (student/not student), school performance (for students), and parents' education) will serve as covariates in analyses connected to the secondary aim of the study.

Patient and public involvement statement
Patients and the public were not involved in the design and planning of the study.

ETHICS AND DISSEMINATION
The current study has been approved by the Swedish Ethical Review Authority (no. 2019-03249) and registered 24/09/2019, pre-result, on IRCTN, ID: ISRCTN91048246. Any important protocol modifications will be reported to IRCTN. For inclusion, all participants must give informed consent online prior to participation in the study. In order to use the web-based intervention program, the participants will need to state a username (which may be fictitious or a pseudonym) and an email address. The email address is necessary for the registration process and for sending one (new) password to use when logging in to the program. Regarding the actual study of the intervention program, all data collection will be done without collecting personal identification information, such as 10-digit personal identification numbers assigned to all residents in Sweden or participants' names, only personal e-mail addresses. The e-mail address will be used to connect data from the baseline measurement to the follow-up measurements. At a later stage, the raw data file will be unidentified and each person assigned a number instead of the e-mail address. The research group is responsible for ensuring that code lists and the data are kept safe in line with routines for handling and storing research data at Karolinka Institutet. All data is handled confidentially and will not be forwarded to third parties. Participating in the present study means that the participants need to reflect on their alcohol and other substance consumption. In addition, questions will be asked concerning personal circumstances, including the participants' mental health and family relationships. These issues may be perceived as somewhat unpleasant. However, in the information that potential participants receive, they are informed that participation is voluntary and that they at any time can interrupt their participation without explaining why. In addition, there will be reference to other types of official support (web pages and telephone numbers), if the advertisement or participation leads to concern about own substance use or related problems. Any issues brought up of the participants during the study will be documented and handled properly. Moreover, the research team includes professionals, such as a psychiatrist and a nurse, with possibility to refer participants to health care clinics if needed. The project's basic hypothesis is that the intervention will have positive effects, with regard to alcohol and other substance consumption among young people with risk use, and potential and benefit may include decreased or ceased risk behaviours. Overall, the benefit for the research persons is considered to exceed any risk of discomfort. Results will be disseminated in scientific peer reviewed journals, at conferences and via the media.

DISCUSSION OF STRENGTHS AND LIMITATIONS
The current study has a number of strengths. The protocol describes a two-arm double-blind RCT, considered to be the most robust experimental design controlling for participant allocation bias, selection bias, performance bias and placebo effect, and with the possibility to make causal inferences. Moreover, an active control condition controls for performance bias and placebo effect. The collection and analysis of information on potential moderators, allowing for control of these factors, facilitates the understanding of the possible effects. Thus, the present study will contribute to the literature on digital substance use prevention interventions among adolescens and young adults in several ways. The choice to study a digital intervention can from a public health perspective be regarded as positive since digital brief interventions have several advantages over face-to-face approaches, e.g., the reduction of stigma around help-seeking for substance use and the way they can be easily disseminated to large groups of people and thus be costeffective. Moreover, digital interventions have previously proven effective in addressing alcohol and other substance use in the general population and some studies have shown promising results also among adolescents and young adults [94,95]. The current intervention is well grounded in theory and incorporates elements of MI and social influence theory that has been shown to be effective in reducing problematic substance use in prior research [39]. Importantly, there are some limitations to the current study. One limitation concerns selection bias and thereby external validity, as recruitment requires either that individuals click on our ads at social media to be considered for inclusion, or that they attend upper secondary schools, or have been in contact with existing clinics, aware of the opportunity to participate in the study. Thus, our participants may have certain personality traits, or are especially prone to help-seeking help compared to a broader audience. The exclusion of people not understanding Swedish is also a limitation with regard to generalizability.

Author contributions
PK, AKS, JG, and THE obtained funding for the study. PK, THE, and CS further developed/modified the intervention. THE designed the study with contribution from PK, AKS, JG, AHB, and, CS. PK wrote this paper. All authors commented on successive manuscript drafts and approved the final version of the manuscript.

Funding statement
The work was supported by the National Public

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.
In your methods section, say that you used the SPIRITreporting guidelines, and cite them as:  Effects of an automated digital brief prevention intervention targeting adolescents and young adults with risky substance use: Study protocol for a randomised controlled trial

ABSTRACT Introduction
Adolescence and young adulthood is a period in life when individuals may be especially vulnerable to harmful substance use. Several critical developmental processes are occurring in the brain, and substance use poses both short-and long-term risks with regard to mental health and social development. From a public health perspective, it is important to prevent or delay substance use to reduce individual risk and societal costs. Given the scarcity of effective interventions targeting substance use among adolescents and young adults, cost-effective and easily disseminated interventions are warranted. The current study will test the effectiveness of a fully automated digital brief intervention aimed at reducing substance use in adolescents and young adults aged 15-25 years.

Methods and analysis
A two-arm double-blind randomised controlled trial (RCT) design is applied to assess the effectiveness of an online digital brief intervention. Baseline assessment and two follow-up assessments at three and six months will be carried out. The aim is to include 800 participants with risky substance use based on the screening tool CRAFFT. Recruitment, informed consent, randomisation, intervention, and follow-up will be implemented online. The primary outcome is reduction in alcohol use. Secondary outcomes concern binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, other substance use, mental health, sexual risk behaviours, and perceived peer pressure. Moreover, the study involves analyses of potential moderators including perfectionism, openness to parents, help-seeking, and background variables.

Ethics and dissemination
The study was approved by the Swedish Ethical Review Authority (no. 2019-03249). Prior to inclusion, participants must give informed consent online. The trial is expected to expand the knowledge on digital preventive interventions for substance using adolescents and young adults.
Results will be disseminated in research journals, at conferences and via the media.

Key words
Motivational Interviewing, Decisional balance, Substance use, Web-based intervention, Normative feedback.

Strengths and limitations of this study
 A double-blind randomised controlled trial (RCT) is considered to be the most robust experimental design, enabling causal inferences, controlling for selection bias and participant allocation bias.
 An active control condition, blinded to the study condition, controls for expectation, detection and performance bias.
 Analysis of several potential moderators, i.e., perfectionism, openness to parents, help-seeking and background variables will contribute to the understanding of possible effects.
 A convenience sample may reduce generalizability due to possible selection bias at recruitment.

INTRODUCTION
Substance use is a major public health concern causing individual suffering as well as societal costs. [1][2][3] Substance use in childhood and younger ages is particularly harmful since the brain is undergoing critical development during this time in life, which makes it more vulnerable to addictive substances. 4 5 In adolescence, some individuals are especially prone to various risk behaviours, increasing the risk of e.g., substance use. 6 Additionally, having substance using peers is a prominent risk factor for one's own substance use at this age, since peers are increasingly important in teenagers' social life. 7 Thus, the initiation of substance use often occurs during this period. 6 After initiation, the use of substances frequently increases during adolescence and young adulthood, posing a number of risks for the individual. 2 8 9 Early onset of substance use implies a risk of severe adverse effects on psychological and physiological development. 2 8 9 Hazardous use of alcohol, cannabis and other substances can lead to chronic problematic consumption patterns and addiction that can significantly influence the developmental trajectory during the transition from childhood to adulthood. 10 More acute consequences include, e.g., mental or psychiatric problems, problems of academic adjustment, 11 accidents, 12 and problems with the police or legal authorities. 1314 Risk behaviours, such as substance use, binge drinking, intoxication, and sexual risk behaviour often co-occur, and the use of lllicit substances, is e.g., often accompanied by alcohol consumption. 1 2 11 15-19 Poly-substance use, i.e., simultaneous use of different substances, may imply an increased risk of various negative outcomes, e.g., mental health problems, shown to increase in magnitude and over time, with the number of substances used. 20 The well-established association between substance use and mental health problems is bidirectional, sometimes manifested in self-medication for 4 mental health problems, and sometimes as an increase in such problems due to substance use. 21 While most of prior research shows that low socio-economic standard is a risk factor for substance use and related problems, 22 23 the current study is informed by recent research on mechanisms involved in adolescence alcohol and other substance use among upper secondary school students from affluent areas, showing associations between internalizing symptoms and substance use, 24 25 as well as associations between achievement pressures (particularly excessive perfectionistic strivings), and isolation from parents (particularly low perceived closeness to mothers). 24 26 The latter finding is in line with previous research showing that a positive parent-child relationship is a protective factor related to lower likelihood of adolescence substance use. 27 28 Although alcohol use has declined among adolescents during recent years, some individuals drink more and binge drinking is still highly prevalent among young adults, where the largest proportion of alcohol risk consumers is still found. 29 30 Moreover, a high prevalence of substance use and poly-substance use among young people in Europe was recognized in the EU Drug Strategy 2005-2012, which calls for action. 31 Also, established cannabis users among Swedish upper secondary school students tend to use cannabis more often than before, according to recent survey data. 32 In light of these circumstances, it is of utmost importance to target and tailor attractive preventive interventions for adolescents and young adults.
One way to reach large groups of young people is to disseminate digital interventions online (i.e., via the Internet). In Sweden, 98 percent of the population has access to the Internet at home and virtually 100 percent of [16][17][18][19][20][21][22][23][24][25] year-olds use the Internet in mobile phones, tablets or computers. 33 Digital interventions have several advantages over traditional ones. They may reduce the stigma around risk behaviours, they are accessible at any time and place, and can be cost-effective because minimal staff resources are used for implementation, especially if they are fully automated. 34 41 and models of Social Influence. 42 43 The goal of the MI approach was to enhance motivation to change by exploring and resolving ambivalence about substance-related behaviours. An important element in the intervention was personalized feedback about one's substance use behaviours in relation to normative comparisons, as personalized feedback is perceived as more relevant for changing behavior than more general information. 44 The normative feedback included information about how a specific which has proven effective for reduction of alcohol consumption in young adults in previous metaanalyses. 46 47 For adolescents and young adults, a normative feedback approach may be particularly appealing, assuming they are curious about how their substance consumption compares to their peers. 48 Likewise, an important element of the intervention was a focus on substance-related social norms and training on how to avoid social high-risk situations and how to resist peer pressure; i.e., raising refusal self-efficacy. 49 In line with the social influence hypothesis stating that a prominent risk factor for substance use in adolescence is the influence of substance using peers, 7 targeting peer pressure in brief digital interventions has been an important element for reducing alcohol use among adolescents. 50 Wiseteens demonstrated significant between-group effects for alcohol use, indicating that a targeted brief motivational intervention in a fully automated digital format can be effective to reduce drinking and lower barriers for accessing substance use service in hazardous drinking adolescents. 38 Digital interventions, often delivered online, generally offer adaptations of evidence based face-to-face interventions, e.g., MI. The effects of MI on several health related behaviours have been evaluated in a number of other studies, especially among adults. 51 Among adolescents, systematic reviews and metaanalyses on MI for various health-related behaviours show mixed results. A review of six studies of MI for reducing alcohol consumption in the emergency room setting suggested that MI was at least as effective as other brief interventions in the same setting. 52 Another review of 24 studies of different brief interventions for reducing alcohol consumption and alcohol-related problems among adolescents showed a significant effect that persisted up to one year compared to control interventions, with greater effects for MI. Effects were consistent over diverse settings and particularly effective components included decisional balance (juxtaposition of pros and cons of change) 53 and goal-setting 54 . Regarding MI for illicit drug use among adolescents, a review of 10 studies showed no effects on drug use behaviors; however, changes in attitudes towards drug use were found, which could be translated into intentions to change behaviors 53 . A scoping review concerning MI for reducing sexual risk behaviors among adolescents identified 29 unique studies with varying designs and conceptualizations of MI and specific risk behaviours, making it difficult to generalize regarding outcomes but indicating the need of more research. 55 MI has most commonly been delivered as an individual face-to-face intervention 51 , but has also been provided in other forms, e.g., via telephone or digitally, with various results on e.g., substance use behaviours. 56 57 Digital programmes based on MI have demonstrated effect in the form of reduced alcohol consumption among young adults. 58 59 Additionally, a combination of screening with a short intervention has shown similar effects on alcohol consumption among adolescents. 60 Brief motivational interventions are empirically supported individual level interventions for reducing alcohol consumption. 45  interventions have in the last decades also been digitally provided. 62 63 Digital interventions can be provided with or without human guidance, i.e., more or less automated. 64 Previous research has demonstrated that even fully automated interventions can reduce alcohol-related problems for young people with risky alcohol consumption up to 12 months after implementation, 59 65 and indicated potential effect on cannabis use in certain groups. 66 67 Although various digital substance use prevention interventions have been developed and tested with promising results, 38 68 studies on the effects of fully automated digitally delivered and MI-based substance use prevention programmes for adolescents and young adults are still few 68 and more well-designed studies are warranted in order to obtain evidence for potential effects.

Aim
The study described in this protocol builds on results previously obtained in a European study of a digital brief MI-based substance use prevention intervention (WISEteens), delivered online to adolescents in a fully automated form. 38 In this study, the effectiveness of a modified version of this intervention will be tested across a range of outcome measures among 15-25 year-olds with hazardous alcohol or other substance use. The primary aim is reduction in frequency and quantity of substance use.

Rationale and hypothesis
The current study will add to existing evidence regarding the effectiveness of digital interventions aimed at reducing alcohol and other substance use among young people. We hypothesize that participants in the intervention group will report reduced substance use (primarily alcohol consumption at three-month follow-up), with a larger effect size compared to an active control group receiving health information about various substances. A number of potential moderators will be assessed to confirm or contradict previous research on factors influencing the development of substance use patterns over time and effects of prevention interventions in different target groups.

METHODS AND ANALYSIS
In order to evaluate the effectiveness of the digital screening and brief motivational intervention, a doubleblind two-arm RCT study design is planned with baseline assessment at study entry and follow-up assessment at three and six-months across a number of outcome variables. Figure 1 displays the trial design. We will also explore and test for moderator effects. To outline and report the current study we used SPIRIT reporting guidelines. 69

Recruitment
The target group for the current study is adolescents and young adults aged 15-25 years with self-reported substance use, a capacity to understand the Swedish language, and access to the Internet. Based on a power calculation to detect a small effect size (Cohen's d=0.2), 70 the aim is to include 800 participants. 71 The participants will be recruited during January-October 2020 by offline marketing at youth health clinics, other health care units for youth, and at upper secondary schools in Stockholm county, as well as by social networks, using online banner advertisements. This will result in a convenience sample from the general Swedish population. 44 Results from follow-up measures are to be completely received in 2021. If sufficient number of participants are not included according to plan, we will extend the recruitment period and intensify the advertising. To enhance participation and follow-up rates, an incentive in the form of a movie ticket, with a value up to 15 Euros, for each completed follow-up assessment will be provided, i.e., up to 3 tickets. Up to three reminders will be sent out if participants do not respond to follow-up assessments.

Screening and informed consent
Potential participants will be guided to an online landing page with screening by an adapted version of the screening tool for use of alcohol and other substances, CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble). 72 This six-item screening tool has demonstrated criterion validity and appropriateness for identifying substance-related problems among adolescents. [73][74][75] The primary eligibility criterion for participating in the study will be a score of 2 or more on the CRAFFT, since this value has shown satisfactory sensitivity for identifying substance use problems. 72 Those who score 2 or higher will be offered participation in the study and given information on confidentiality, voluntariness of participation, and human subject protections. 44 They will also be provided with contact information on suitable counselling service providers. Those who agree to participate will be asked to give digital informed consent.

Assessment
All assessment will occur online. The selection of measures and associated instruments is based on the aim of the study and the theoretical base concerning factors influencing substance use. Based on results from the previous European WISEteens study, 38 showing significant effects on alcohol consumption but not on other substance use, the primary outcome is alcohol consumption at three-month follow-up.
Secondary outcomes include alcohol consumption at six-month follow-up, binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed at three-and six-months follow-ups, other substance use, mental health, sexual risk behaviours, and perceived peer pressure at three and six-month follow-ups. The moderating variables include perfectionism, openness to parents, help-seeking due to mental health problems, and socio-demographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for students), and parents' education).

Measures
Alcohol use will be measured using two instruments. One of them is the short version of the Alcohol Use Disorders Identification Test (AUDIT-C), 76 measuring frequency and amount of consumed alcohol and frequency of binge drinking, providing a widely used and valid index sum score for problematic alcohol use among adolescents. 75 77 The other is the Daily Drinking Questionnaire (DDQ), 78 measuring the quantity and frequency of alcohol use a typical week. Previous research has demonstrated that the DDQ is highly correlated with other measures of self-reported alcohol consumption. 79 Other substance use will be measured by the first four items of the Drug Use Disorder Identification Test (DUDIT), 80 assessing frequency of consumption of drugs other than alcohol, frequency of different types of drugs other than alcohol used at the same occasion (1 = never to 5 = four or more times a week), and number of occasions when drugs other than alcohol are consumed on a typical day of drug use (1 = zero to 5 = seven or more).
The DUDIT has been found to be effective in screening for drug-related problems in clinically selected groups, 81 and has proven useful in the context of public health surveys in Sweden. 80 82 Mental health will be measured by The WHO-5 Well-Being Index, which has demonstrated validity both as a screening tool for depression and as an outcome measure in clinical trials. 83 This scale has been successfully applied across a wide range of study fields, translated into more than 30 languages, including Swedish, 84 and found psychometrically sound. Changes in sexual risk behaviours will be measured by multiple choice questions on sexual behaviour under the influence of alcohol and/or other substances and unprotected sex, previously used in a survey among Swedish visitors at youth health clinics in Stockholm county. 85 Changes in perceived peer pressure will be measured by two items retrieved and adapted from the Peer pressure inventory. 86 Perfectionism will be measured by two subscales of the Frost Multidimensional

Randomisation
The study is double-blind, thus neither the participant nor the researchers will know which participant is allocated to the intervention or to the active control condition. After baseline assessment, the participants will be automatically randomised to one of the two study groups by a computer programme using an unrestricted randomisation protocol. Participants will then be informed about the name of their programme and given access immediately.

The intervention
Participants allocated to the intervention group will initially be asked to give additional information on alcohol and other substance use and to state their body weight, in order to generate personalized feedback on substance use behaviour and to estimate alcohol blood concentration on a typical drinking occasion.

Description of the intervention content
The intervention is interactive, digitally delivered online in a fully automated form, and requires approximately 20 minutes to complete. Personalized feedback is given to the participants based on their responses to previous assessment and suggestions on how to respond to this feedback are provided. This interactivity imitates a face-to face "dialogue" with techniques from MI such as an empathic approach, rolling with resistance, aiming at creating a dissonance between actual and desired behaviour, raising selfefficacy, and at the same time avoiding argumentation. 57 The intervention consists of three main components outlined below, and additional health related information.

Personalized feedback
The personalized feedback includes an estimation of the participants' blood-alcohol-concentration (BAC) and information on the associated risks concerning the participants' heaviest drinking episode during the past 30 days. The value will be based on a measure of Peak Drinking Quantity 92 and estimated using the Widmark-formula, which takes into account weight and sex. 93 The participants will receive graphed feedback regarding number of standard alcohol units per week that they think their peers consume (descriptive norms), as well as the participant's individual levels of consumption in relation to comparative data (actual drinking levels) from a reference group. Comparative data (AUDIT-C scores) will be taken from alcohol prevalence estimates found in a nationally representative sample of 16-25 yearolds undertaken by the Public Health Agency of Sweden. Comparative feedback will be available for drinking but not for substances other than alcohol.

Interactive MI-based exercises
The exercises provided in the current intervention build on the assumption that participants may hold certain levels of ambivalence about their current substance use, and that if they are willing to make a change they may not know how to, or may not be confident that they are able to. 41 Therefore, the intervention uses importance and confidence rulers with a short summary and feedback to encourage participants to reflect on personal reasons for change and explore personal strengths and ability to change.
Furthermore, the programme provides a decisional balance to pick up and graphically illustrate potential levels of ambivalence by offering the participants a list of possible pros and cons regarding the decision to change their current substance use. 57 94 Participants are instructed to choose statements that apply to them and are presented with the resulting balance sheet of their personal comparative potential gains and losses.

Practical advice
The intervention focuses on raising self-efficacy for being able to avoid drinking in social situations, if desirable. 50 The participant will be asked to select three among 12 provided drinking situations that they consider most tempting and rank them. The situations are adapted from the adolescent version of the Drinking Refusal Self-efficacy Questionnaire (DRSEQ-RA). 95 According to the selection, a number of strategies will be offered for each of the selected drinking situations to provide participants with a tool kit necessary for engaging in and maintaining their behavioural goal.

Health related information connected to substance use
Finally, the intervention programme includes health related information associated with substance use.
The information is provided optionally throughout the program behind "read more" buttons and also at the end of the programme. The information contains statements regarding risks connected to substance use

The control condition
The control group will receive the same general health related information as the intervention group, i.e., the additional information which is connected to the intervention programme.

Statistical analysis
Randomisation-checks of baseline variables regarding alcohol and other substances use and psychological state will be conducted using multivariate analyses of variance MANOVA. To test the effectiveness of the intervention, we will assess whether participants in the intervention group to a larger extent report decreased substance use, sexual risk behaviours, perceived peer-pressure, and improved mental health after three and six months, respectively, compared to participants in the active control group. Data analyses will consist of comparing outcome measurements with regards to within-group and betweengroup differences according to the intention-to-treat (ITT) principle in the primary analysis, accounting for all included participants regardless of whether or not they completed follow-up assessments. We will also perform per protocol analyses. The reason for choosing ITT in the primary analysis is the ambition to maintain the power of the study and also to avoid biased results due to selection of those completing all follow-ups, as they may have special characteristics not representative for the whole study group, which might influence study results. The main analysis of effectiveness will use mixed effects regression models, which can be applied to both continuous and categorical outcomes and also non-normally distributed outcomes. Moreover, mixed-effects regression models are robust to missing data in longitudinal studies. 96 In the current study we assume a quite large dropout rate with data missing at random (MAR), which can be handled using mixed effects regression models. Separate models will be run to test each outcome, i.e., alcohol consumption, binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed at three-and six-months follow-ups, other substance use, mental health, sexual risk behaviours and perceived peer pressure. Effect sizes will be calculated separately for the two measures used for the primary outcome variable (AUDIT-C and DDQ). Potential moderators, including perfectionism, openness to parents, help-seeking due to mental health problems, and sociodemographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 12 students), and parents' education) will serve as covariates in analyses of intervention effects. If moderator effects are found, we will carry out post-hoc analyses stratified by the detected variable.

Patient and public involvement statement
Patients and the public were not involved in the design and planning of the study, except for persons at the age of the target group reading manuscripts for the revised version of Wiseteens and also participating in a pilot-test.

ETHICS AND DISSEMINATION
The current study was approved by the Swedish Ethical Review Authority (no. 2019-03249) and registered 24/09/2019, pre-result, on IRCTN, ID: ISRCTN91048246. Any important protocol modifications will be reported to IRCTN. For inclusion, all participants must give informed consent online prior to participation in the study. In order to participate, the participants will need to state a username (which may be fictitious or a pseudonym) and an email address. All data collection will be done without collecting personal identification information, only personal e-mail addresses. The e-mail address will be used to connect data from the baseline measurement to the follow-up measurements. At a later stage, the raw data file will be anonymized and each person assigned a number instead of the e-mail address. The data will be stored in line with routines for handling and storing research data at Karolinska Institutet. All data is handled confidentially and will not be forwarded to third parties. Participating in the present study means that the participants need to reflect on their alcohol consumption and other substance use. In addition, questions will be asked concerning personal circumstances, including the participants' mental health and family relationships. These issues may be perceived as somewhat unpleasant. However, in the information that potential participants receive, they are informed that participation is voluntary and that they at any time can end their participation without explaining why. In addition, there will be reference to other types of official support (web pages and telephone numbers), if the advertisement or participation leads to concern about own substance use or related problems. Any issues brought up by the participants during the study will be documented and handled properly. Moreover, the research team includes professionals, such as a psychiatrist and a nurse, with possibility to refer participants to health care clinics if needed. The project's basic hypothesis is that the intervention will have positive effects, with regard to alcohol and other substance use among young people with risk use, and potential benefit may include decreased or ceased risk behaviours. Overall, the benefit for the research persons is considered to exceed

DISCUSSION OF STRENGTHS AND LIMITATIONS
The current study has a number of strengths. The protocol describes a two-arm double-blind RCT, adolescents and young adults. 97 98 The current intervention is well grounded in theory and incorporates elements of MI and social influence theory that has been shown to be effective in reducing problematic substance use in prior research. 57 Importantly, there are also some limitations to the current study. One limitation concerns selection bias and thereby external validity, as recruitment requires either that individuals click on our ads at social media to be considered for inclusion, or that they attend upper secondary schools, or have been in contact with existing clinics, aware of the opportunity to participate in the study. Thus, our participants may have certain personality traits, or are especially prone to helpseeking compared to a broader audience. The exclusion of people not understanding Swedish is also a limitation with regard to generalizability.

STATEMENTS Author contributions
PK, AKS, JG, and THE obtained funding for the study. PK, THE, and CS further developed/modified the intervention. THE designed the study with contribution from PK, AKS, JG, AHB, and, CS. PK wrote this paper. All authors commented on successive manuscript drafts and approved the final version of the manuscript.

Data statement
Collected data will be available from the Centre for Psychiatry Research, a collaboration between the Karolinska Institute and Region Stockholm, but restrictions apply to their availability, as they were used under ethical permission for the current study, and so are not publicly available. However, data are available from the authors upon reasonable request and with permission from the Centre for Psychiatry Research.

Competing interests
The authors declare no competing interests.

Methods and analysis
A two-arm double-blind randomised controlled trial (RCT) design is applied to assess the effectiveness of the intervention. Baseline assessment, as well as three-and six-month follow-up, will be carried out. The aim is to include 800 participants with risky substance use based on the screening tool CRAFFT.
Recruitment, informed consent, randomisation, intervention, and follow-up will be implemented online.
The primary outcome is reduction in alcohol use, measured by AUDIT-C total score and average total drinks per week assessed by DDQ. Secondary outcomes concern binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, other substance use, mental health, sexual risk behaviours, and perceived peer pressure. Moreover, the study involves analyses of potential moderators including perfectionism, openness to parents, help-seeking, and background variables.

Ethics and dissemination
The study was approved by the Swedish Ethical Review Authority (no. 2019-03249). The trial is expected to expand the knowledge on digital preventive interventions for substance using adolescents and young adults. Results will be disseminated in research journals, at conferences and via the media.

Key words
Motivational Interviewing, Decisional balance, Substance use, Web-based intervention, Normative feedback.

Strengths and limitations of this study
 A double-blind randomised controlled trial (RCT) is considered to be the most robust experimental design, enabling causal inferences, controlling for selection bias and participant allocation bias.
 An active control condition, blinded to the study condition, controls for expectation, detection and performance bias.
 Analysis of several potential moderators, i.e., perfectionism, openness to parents, help-seeking and background variables will contribute to the understanding of possible effects.
A convenience sample may reduce generalizability due to possible selection bias at recruitment.

INTRODUCTION
Substance use is a major public health concern causing individual suffering as well as societal costs. [1][2][3] Substance use in childhood and younger ages is particularly harmful since the brain is undergoing critical development during this time in life, which makes it more vulnerable to addictive substances. 4 5 In adolescence, some individuals are especially prone to various risk behaviours, increasing the risk of e.g., substance use. 6 Additionally, having substance using peers is a prominent risk factor for one's own substance use at this age, since peers are increasingly important in teenagers' social life. 7 Thus, the initiation of substance use often occurs during this period. 6 After initiation, the use of substances frequently increases during adolescence and young adulthood, posing a number of risks for the individual. 2 8 9 Early onset of substance use implies a risk of severe adverse effects on psychological and physiological development. 2 8 9 Hazardous use of alcohol, cannabis and other substances can lead to chronic problematic consumption patterns and addiction that can significantly influence the developmental trajectory during the transition from childhood to adulthood. 10 More acute consequences include, e.g., mental or psychiatric problems, problems of academic adjustment, 11 accidents, 12 and problems with the police or legal authorities. 13 14 Risk behaviours, such as substance use, binge drinking, intoxication, and sexual risk behaviour often co-occur, and the use of illicit substances, is e.g., often accompanied by alcohol consumption. 1 2 11 15-19 Poly-substance use, i.e., simultaneous use of different substances, may imply an increased risk of various negative outcomes, e.g., mental health problems, shown to increase in magnitude and over time, with the number of substances used. 20 The well-established association between substance use and mental health problems is bidirectional, sometimes manifested in self-medication for mental health problems, and sometimes as an increase in such problems due to substance use. 21 While most of prior research shows that low socio-economic standard is a risk factor for substance use and related problems, 22 23 the current study is informed by recent research on mechanisms involved in adolescence alcohol and other substance use among upper secondary school students from affluent areas, showing associations between internalizing symptoms and substance use, 24 25 as well as associations between achievement pressures (particularly excessive perfectionistic strivings), and isolation from parents (particularly low perceived closeness to mothers). 24 26 The latter finding is in line with previous research showing that a positive parent-child relationship is a protective factor related to lower likelihood of adolescence substance use. 27 28 Although alcohol use has declined among adolescents during recent years, some individuals drink more and binge drinking is still highly prevalent among young adults, where the largest proportion of alcohol risk consumers is still found. 29 30 Moreover, a high prevalence of substance use and poly-substance use among young people in Europe was recognized in the EU Drug Strategy 2005-2012, which calls for action. 31 Also, established cannabis users among Swedish upper secondary school students tend to use cannabis more often than before, according to recent survey data. 32 In light of these circumstances, it is of utmost importance to target and tailor attractive preventive interventions for adolescents and young adults.
One way to reach large groups of young people is to disseminate digital interventions online (i.e., via the Internet). In Sweden, 98 percent of the population has access to the Internet at home and virtually 100 percent of [16][17][18][19][20][21][22][23][24][25] year-olds use the Internet in mobile phones, tablets or computers. 33 Digital interventions have several advantages over traditional ones. They may reduce the stigma around risk behaviours, they are accessible at any time and place, and can be cost-effective because minimal staff resources are used for implementation, especially if they are fully automated. 34 35 Digital interventions may thus be particularly suitable for adolescents and young adults. 36 To date, digital prevention programmes targeting alcohol and other substance use among adolescents and young adults that meet scientific evaluation criteria are scarce. An early example is eScreen.se, offering screening for alcohol and drug use with personalized feedback. 37 A more recent example is a fully automated brief motivational intervention for substance-using [16][17][18] year-olds, tested in an RCT in four European countries, [38][39][40] where our research group was one of the partners. The intervention, named WISEteens, relied on Motivational Interviewing (MI) 41 and models of Social Influence. 42 43 The goal of the MI approach was to enhance motivation to change by exploring and resolving ambivalence about substance-related behaviours. An important element in the intervention was personalized feedback about one's substance use behaviours in relation to normative comparisons, as personalized feedback is perceived as more relevant for changing behavior than more general information. 44 The normative feedback included information about how a specific reference group actually consumed substances, in order to correct participants' "inflated perception", 45 which has proven effective for reduction of alcohol consumption in young adults in previous metaanalyses. 46 47 For adolescents and young adults, a normative feedback approach may be particularly appealing, assuming they are curious about how their substance consumption compares to their peers. 48 Likewise, an important element of the intervention was a focus on substance-related social norms and training on how to avoid social high-risk situations and how to resist peer pressure; i.e., raising refusal self-efficacy. 49 In line with the social influence hypothesis stating that a prominent risk factor for substance use in adolescence is the influence of substance using peers, 7 targeting peer pressure in brief digital interventions has been an important element for reducing alcohol use among adolescents. 50 WISEteens demonstrated significant between-group effects for alcohol use, indicating that a targeted brief motivational intervention in a fully automated digital format can be effective to reduce drinking and lower barriers for accessing substance use service in hazardous drinking adolescents. 38 Digital behaviours, making it difficult to generalize regarding outcomes but indicating the need of more research. 55 MI has most commonly been delivered as an individual face-to-face intervention 51 , but has also been provided in other forms, e.g., via telephone or digitally, with various results on e.g., substance use behaviours. 56 57 Digital programmes based on MI have demonstrated effect in the form of reduced alcohol consumption among young adults. 58 59 Additionally, a combination of screening with a short intervention has shown similar effects on alcohol consumption among adolescents. 60 Brief motivational interventions are empirically supported individual level interventions for reducing alcohol consumption. 45 61 Such interventions have in the last decades also been digitally provided. 62 63 Digital interventions can be provided with or without human guidance, i.e., more or less automated. 64 Previous research has demonstrated that even fully automated interventions can reduce alcohol-related problems for young people with risky alcohol consumption up to 12 months after implementation, 59 65 and indicated potential effect on cannabis use in certain groups. 66 67 Although various digital substance use prevention interventions have been developed and tested with promising results, 38 68 studies on the effects of fully automated digitally delivered and MI-based substance use prevention programmes for adolescents and young adults are still few 68 and more well-designed studies are warranted in order to obtain evidence for potential effects.

Aim
The study described in this protocol builds on results previously obtained in a European study of a digital brief MI-based substance use prevention intervention (WISEteens), delivered online to adolescents in a fully automated form. 38 In this study, the effectiveness of a modified version of this intervention, with new graphics and population based feedback-generating data on alcohol consumption covering all included age groups, will be tested across a range of outcome measures among 15-25 year-olds with hazardous alcohol or other substance use. The primary aim is reduction in frequency and quantity of substance use.

Rationale and hypothesis
The current study will add to existing evidence regarding the effectiveness of digital interventions aimed at reducing alcohol and other substance use among young people. We hypothesize that participants in the intervention group will report a significant reduction in substance use (primarily alcohol use), compared to an active control group receiving health information about various substances. A number of potential moderators will be assessed to confirm or contradict previous research on factors influencing the development of substance use patterns over time, and effects of prevention interventions in different target groups.

METHODS AND ANALYSIS
In order to evaluate the effectiveness of the digital screening and brief motivational intervention, a doubleblind two-arm RCT study design is planned with baseline assessment at study entry and follow-up assessment at three and six-months across a number of outcome variables. Figure 1 displays the trial design. We will also explore and test for moderator effects. To outline and report the current study we used SPIRIT reporting guidelines. 69

Recruitment
The target group for the current study is adolescents and young adults aged 15-25 years with self-reported substance use, a capacity to understand the Swedish language, and access to the Internet. Based on a power calculation to detect a small effect size (Cohen's d=0.2), 70 the aim is to include 800 participants. 71 The participants will be recruited during January-October 2020 by offline marketing at youth health clinics, other health care units for youth, and at upper secondary schools in Stockholm county, as well as by social networks, using online banner advertisements. This will result in a convenience sample from the general Swedish population. 44 Results from follow-up measures are to be completely received in 2021. If sufficient number of participants have not been recruited according to plan, we will extend the recruitment period and intensify the advertising by increasing visibility on social platforms. To enhance participation and follow-up rates, an incentive in the form of a movie ticket, with a value up to 15 Euros, for each completed follow-up assessment will be provided, i.e., up to 3 tickets. Up to three reminders will be sent out if participants do not respond to follow-up assessments.

Screening and informed consent
Potential participants will be guided to an online landing page with screening by an adapted version of the screening tool for use of alcohol and other substances, CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble). 72 This six-item screening tool has demonstrated criterion validity and appropriateness for identifying substance-related problems among adolescents. [73][74][75] The primary eligibility criterion for participating in the study will be a score of 2 or more on the CRAFFT, since this value has shown satisfactory sensitivity for identifying substance use problems. 72 Those who score 2 or higher will be offered participation in the study and given information on confidentiality, voluntariness of participation, and human subject protections. 44 They will also be provided with contact information on suitable counselling service providers. Those who agree to participate will be asked to give digital informed consent.

Measures
All assessment will occur online. The selection of measures and associated instruments is based on the aim of the study and the theoretical base concerning factors influencing substance use. Based on results  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   8 from the previous European WISEteens study, 38 showing significant effects on alcohol consumption but not on other substance use, the primary outcome is alcohol use, measured at three-and six-month followups. Secondary outcomes include binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed at three-and six-months follow-ups, other substance use, mental health, sexual risk behaviours, and perceived peer pressure at three and six-month follow-ups.
The moderating variables include perfectionism, openness to parents, help-seeking due to mental health problems, and socio-demographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for students), and parents' education).
Alcohol use, including the primary outcome, will be measured using two instruments. One of them is the short version of the Alcohol Use Disorders Identification Test (AUDIT-C), 76 measuring frequency and amount of consumed alcohol and frequency of binge drinking, providing a widely used and valid index sum score for problematic alcohol use among adolescents. 75 77 The other is the Daily Drinking Questionnaire (DDQ), 78 measuring a variety of parameters of alcohol use in a typical week. Previous research has demonstrated that the DDQ is highly correlated with other measures of self-reported alcohol consumption. 79 The primary outcome will be measured by AUDIT-C total score and average total drinks per week, assessed by DDQ. The two primary outcome measures complement one another in that the AUDIT-C primarily offers an indication of hazardous or harmful use, 80 whereas the DDQ quantity measure will indicate the level of alcohol consumption in grams per week. Other substance use will be measured by a short version of the Drug Use Disorder Identification Test (DUDIT), 81 including the first four items (DUDIT-C) which assess frequency of consumption of drugs other than alcohol, frequency of different types of drugs other than alcohol used at the same occasion (1 = never to 5 = four or more times a week), and number of occasions when drugs other than alcohol are consumed on a typical day of drug use (1 = zero to 5 = seven or more). The DUDIT has been found to be effective in screening for drug-related problems in clinically selected groups, 82 and both the DUDIT and the DUDIT-C have proven useful in the context of public health surveys in Sweden. 81 83 Mental health will be measured by The WHO-5 Well- Being Index, which has demonstrated validity both as a screening tool for depression and as an outcome measure in clinical trials. 84 This scale has been successfully applied across a wide range of study fields, translated into more than 30 languages, including Swedish, 85 and found psychometrically sound. Changes in sexual risk behaviours will be measured by multiple choice questions on sexual behaviour under the influence of alcohol and/or other substances and unprotected sex, previously used in a survey among Swedish visitors at youth health clinics in Stockholm county. 86 Changes in perceived peer pressure will be  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   9 measured by two items retrieved and adapted from the Peer pressure inventory. 87 Perfectionism will be measured by two subscales of the Frost Multidimensional Perfectionism Scale (FMPS). 88 89 Openness to parents will be measured among participants who are up to 20 years old with questions about disclosure from Stattin & Kerr, 90-92 slightly modified into statements to be compatible with the digital online format.
Help-seeking due to mental health problems will be measured by dichotomous questions regarding if, and to what healthcare provider, the participant has turned for help. Additionally, background data will be asked for using multiple choice questions: sex (man, women, other), age (15, 16…25), residence (name of municipality), occupation (secondary school, upper secondary school, university, not student, working, practicing, unemployed, other), school performance (merit value for students), and parents' education (secondary school, upper secondary school, university, don't know).

Randomisation
The study is double-blind, thus neither the participant nor the researchers will know which participant is allocated to the intervention or to the active control condition. After baseline assessment, the participants will be automatically randomised to one of the two study groups by a computer programme using an unrestricted randomisation protocol. Participants will then be informed about the name of their programme and given access immediately.

The intervention
Participants allocated to the intervention group will initially be asked to give additional information on alcohol and other substance use and to state their body weight, in order to generate personalized feedback on substance use behaviour and to estimate alcohol blood concentration on a typical drinking occasion.

Description of the intervention content
The intervention is interactive, digitally delivered online in a fully automated form, and requires approximately 20 minutes to complete. Personalized feedback is given to the participants based on their responses to previous assessment and suggestions on how to respond to this feedback are provided. This interactivity imitates a face-to face "dialogue" with techniques from MI such as an empathic approach, rolling with resistance, aiming at creating a dissonance between actual and desired behaviour, raising self-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 10 efficacy, and at the same time avoiding argumentation. 57 The intervention consists of three main components outlined below, and additional health related information.

Personalized feedback
The personalized feedback includes an estimation of the participants' blood-alcohol-concentration (BAC) and information on the associated risks concerning the participants' heaviest drinking episode during the past 30 days. The value will be based on a measure of Peak Drinking Quantity 93 and estimated using the Widmark-formula, which takes into account weight and sex. 94 The participants will receive graphed feedback regarding number of standard alcohol units per week that they think their peers consume (descriptive norms), as well as the participant's individual levels of consumption in relation to comparative data (actual drinking levels) from a reference group. Comparative data (AUDIT-C scores) will be taken from alcohol prevalence estimates found in a nationally representative sample of 16-25 yearolds undertaken by the Public Health Agency of Sweden. Comparative feedback will be available for drinking but not for substances other than alcohol.

Interactive MI-based exercises
The exercises provided in the current intervention build on the assumption that participants may hold certain levels of ambivalence about their current substance use, and that if they are willing to make a change they may not know how to, or may not be confident that they are able to. 41 Therefore, the intervention uses importance and confidence rulers with a short summary and feedback to encourage participants to reflect on personal reasons for change and explore personal strengths and ability to change.
Furthermore, the programme provides a decisional balance to pick up and graphically illustrate potential levels of ambivalence by offering the participants a list of possible pros and cons regarding the decision to change their current substance use. 57 95 Participants are instructed to choose statements that apply to them and are presented with the resulting balance sheet of their personal comparative potential gains and losses.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 11

Practical advice
The intervention focuses on raising self-efficacy for being able to avoid drinking in social situations, if desirable. 50 The participant will be asked to select three among 12 provided drinking situations that they consider most tempting and rank them. The situations are adapted from the adolescent version of the Drinking Refusal Self-efficacy Questionnaire (DRSEQ-RA). 96 According to the selection, a number of strategies will be offered for each of the selected drinking situations to provide participants with a tool kit necessary for engaging in and maintaining their behavioural goal.

Health related information connected to substance use
Finally, the intervention programme includes health related information associated with substance use.
The information is provided optionally throughout the program behind "read more" buttons and also at the end of the programme. The information contains statements regarding risks connected to substance use that the participant can reflect on, as well as optional links for more information. There are also two FAQsections about alcohol and cannabis, respectively, as well as information about how different substances may affect the individual, both physically and mentally. Finally, the information also contains cases, were the participant can read about some typical young persons who have used substances, why they used and what negative effects they have noted, and why they chose to stop using.

The control condition
The control group will receive the same general health related information as the intervention group, i.e., the additional information which is connected to the intervention programme.

Statistical analysis
Randomisation-checks of baseline variables regarding alcohol and other substances use and psychological state will be conducted using multivariate analyses of variance MANOVA. To test the effectiveness of the intervention, we will assess whether participants in the intervention group to a greater extent report decreased substance use, sexual risk behaviours, perceived peer-pressure, and improved mental health after three and six months, respectively, compared to participants in the active control group. Data analyses will consist of comparing outcome measurements with regards to within-group and betweengroup differences according to the intention-to-treat (ITT) principle in the primary analysis, accounting for all included participants regardless of whether or not they completed follow-up assessments. We will also perform per protocol analyses. The reason for choosing ITT in the primary analysis is the ambition to  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y maintain the power of the study and also to avoid biased results due to selection of those completing all follow-ups, as they may have special characteristics not representative for the whole study group, which might influence study results. The main analysis of effectiveness will use mixed effects regression models, which can be applied to both continuous and categorical outcomes and also non-normally distributed outcomes. Also, mixed-effects regression models are robust to missing data in longitudinal studies. 97 In the current study, we assume a quite large dropout rate, based on previous research. For example, a metaanalysis comprising 17 studies of unsupported (fully automated) interventions for depression showed 74% dropout rate at post-treatment. 98 Moreover, we expect data missing at random (MAR), which can be handled using mixed effects regression models. Separate models will be run to test each outcome, i.e., alcohol consumption, binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed at three-and six-months follow-ups, other substance use, mental health, sexual risk behaviours and perceived peer pressure. Effect sizes will be calculated separately for the two measures used for the primary outcome variable (AUDIT-C and DDQ). Potential moderators, including perfectionism, openness to parents, help-seeking due to mental health problems, and sociodemographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for students), and parents' education) will serve as covariates in analyses of intervention effects. If moderator effects are found, we will carry out post-hoc analyses stratified by the detected variable.

Patient and public involvement statement
Patients and the public were not involved in the design and planning of the study, except for persons at the age of the target group reading manuscripts for the revised version of WISEteens and also participating in a pilot-test.

ETHICS AND DISSEMINATION
The current study was approved by the Swedish Ethical Review Authority (no. 2019-03249) and registered 24/09/2019, pre-results, on IRCTN, ID: ISRCTN91048246. Any important protocol modifications will be reported to IRCTN. For inclusion, all participants must give informed consent online prior to participation in the study. In order to participate, the participants will need to state a username (which may be fictitious or a pseudonym) and an email address. All data collection will be done without collecting personal identification information, only personal e-mail addresses. The e-mail address will be used to connect data from the baseline measurement to the follow-up measurements. At a later stage, the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 13 raw data file will be anonymized and each person assigned a number instead of the e-mail address. The data will be stored in line with routines for handling and storing research data at Karolinska Institutet. All data is handled confidentially and will not be forwarded to third parties. Participating in the present study means that the participants need to reflect on their alcohol consumption and other substance use. In addition, questions will be asked concerning personal circumstances, including the participants' mental health and family relationships. These issues may be perceived as somewhat unpleasant. However, in the information that potential participants receive, they are informed that participation is voluntary and that they at any time can end their participation without explaining why. In addition, there will be reference to other types of official support (web pages and telephone numbers), if the advertisement or participation leads to concern about own substance use or related problems. Any issues brought up by the participants during the study will be documented and handled properly. Moreover, the research team includes professionals, such as a psychiatrist and a nurse, with possibility to refer participants to health care clinics if needed. The project's basic hypothesis is that the intervention will have positive effects, with regard to alcohol and other substance use among young people with risk use, and potential benefit may include decreased or ceased risk behaviours. Overall, the benefit for the research persons is considered to exceed any risk of discomfort. Results will be disseminated in scientific peer reviewed journals, at conferences and via the media.

DISCUSSION OF STRENGTHS AND LIMITATIONS
The current study has a number of strengths. The protocol describes a two-arm double-blind RCT, considered to be the most robust experimental design, controlling for selection bias and participant allocation bias and with the possibility to make causal inferences. Moreover, an active control condition, blinded to the study condition, controls for expectation, detection and performance bias. The collection and analysis of information on potential moderators, allowing for control of these factors, facilitates the understanding of the possible effects. Thus, the present study will contribute to the literature on digital substance use prevention interventions among adolescents and young adults in several ways. The choice to study a digital intervention can from a public health perspective be regarded as positive. Digital brief interventions have several advantages over face-to-face approaches, e.g., the reduction of stigma around help-seeking for substance use and easy dissemination to large groups of people, which makes them costeffective. Moreover, digital interventions have previously proven effective in addressing alcohol and other substance use in the general population and some studies have shown promising results also among adolescents and young adults. 99 100 The current intervention is well grounded in theory and incorporates elements of MI and social influence theory that has been shown to be effective in reducing problematic  57 Importantly, there are also some limitations to the current study. One limitation concerns selection bias and thereby external validity, as recruitment requires either that individuals click on our ads at social media to be considered for inclusion, or that they attend upper secondary schools, or have been in contact with existing clinics, aware of the opportunity to participate in the study. Thus, our participants may have certain personality traits, or are especially prone to helpseeking compared to a broader audience. The exclusion of people not understanding Swedish is also a limitation with regard to generalizability.

STATEMENTS Author contributions
PK, AKS, JG, and THE obtained funding for the study. PK, THE, and CS further developed/modified the intervention. THE designed the study with contribution from PK, AKS, JG, AHB, and, CS. PK wrote this paper. All authors commented on successive manuscript drafts and approved the final version of the manuscript. bodies had no role in study design, data collection, analysis, data interpretation or writing manuscripts.

Data statement
Collected data will be available from the Centre for Psychiatry Research, a collaboration between the Karolinska Institute and Region Stockholm, but restrictions apply to their availability, as they were used under ethical permission for the current study, and so are not publicly available. However, data are available from the authors upon reasonable request and with permission from the Centre for Psychiatry Research.

Methods and analysis
A two-arm double-blind randomised controlled trial (RCT) design is applied to assess the effectiveness of the intervention. Baseline assessment, as well as three-and six-month follow-up, will be carried out. The aim is to include 800 participants with risky substance use based on the screening tool CRAFFT.
Recruitment, informed consent, randomisation, intervention, and follow-up will be implemented online.
The primary outcome is reduction in alcohol use, measured by AUDIT-C total score. Secondary outcomes concern binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, average daily drinks per typical week, other substance use, mental health, sexual risk behaviours, and perceived peer pressure. Moreover, the study involves analyses of potential moderators including perfectionism, openness to parents, help-seeking, and background variables.

Ethics and dissemination
The study was approved by the Swedish Ethical Review Authority (no. 2019-03249). The trial is expected to expand the knowledge on digital preventive interventions for substance using adolescents and young adults. Results will be disseminated in research journals, at conferences and via the media.

Key words
Motivational Interviewing, Decisional balance, Substance use, Web-based intervention, Normative feedback.

Strengths and limitations of this study
 A double-blind randomised controlled trial (RCT) is considered to be the most robust experimental design, enabling causal inferences, controlling for selection bias and participant allocation bias.
 An active control condition, blinded to the study condition, controls for expectation, detection and performance bias.
 Analysis of several potential moderators, i.e., perfectionism, openness to parents, help-seeking and background variables will contribute to the understanding of possible effects.
 A convenience sample may reduce generalizability due to possible selection bias at recruitment.

INTRODUCTION
Substance use is a major public health concern causing individual suffering as well as societal costs. [1][2][3] Substance use in childhood and younger ages is particularly harmful since the brain is undergoing critical development during this time in life, which makes it more vulnerable to addictive substances. 4 5 In adolescence, some individuals are especially prone to various risk behaviours, increasing the risk of e.g., substance use. 6 Additionally, having substance using peers is a prominent risk factor for one's own substance use at this age, since peers are increasingly important in teenagers' social life. 7 Thus, the initiation of substance use often occurs during this period. 6 After initiation, the use of substances frequently increases during adolescence and young adulthood, posing a number of risks for the individual. 2 8 9 Early onset of substance use implies a risk of severe adverse effects on psychological and physiological development. 2 8 9 Hazardous use of alcohol, cannabis and other substances can lead to chronic problematic consumption patterns and addiction that can significantly influence the developmental trajectory during the transition from childhood to adulthood. 10 More acute consequences include, e.g., mental or psychiatric problems, problems of academic adjustment, 11 accidents, 12 and problems with the police or legal authorities. 13 14 Risk behaviours, such as substance use, binge drinking, intoxication, and sexual risk behaviour often co-occur, and the use of illicit substances, is e.g., often accompanied by alcohol consumption. 1 2 11 15-19 Poly-substance use, i.e., simultaneous use of different substances, may imply an increased risk of various negative outcomes, e.g., mental health problems, shown to increase in magnitude and over time, with the number of substances used. 20 The well-established association between substance use and mental health problems is bidirectional, sometimes manifested in self-medication for mental health problems, and sometimes as an increase in such problems due to substance use. 21 While most of prior research shows that low socio-economic standard is a risk factor for substance use and related problems, 22 23 the current study is informed by recent research on mechanisms involved in adolescence alcohol and other substance use among upper secondary school students from affluent areas, showing associations between internalizing symptoms and substance use, 24 25 as well as associations between achievement pressures (particularly excessive perfectionistic strivings), and isolation from parents (particularly low perceived closeness to mothers). 24 26 The latter finding is in line with previous research showing that a positive parent-child relationship is a protective factor related to lower likelihood of adolescence substance use. 27 28 Although alcohol use has declined among adolescents during recent years, some individuals drink more and binge drinking is still highly prevalent among young adults, where the largest proportion of alcohol risk consumers is still found. 29 30 Moreover, a high prevalence of substance use and poly-substance use among young people in Europe was recognized in the EU Drug Strategy 2005-2012, which calls for action. 31 Also, established cannabis users among Swedish upper secondary school students tend to use cannabis more often than before, according to recent survey data. 32 In light of these circumstances, it is of utmost importance to target and tailor attractive preventive interventions for adolescents and young adults.
One way to reach large groups of young people is to disseminate digital interventions online (i.e., via the Internet). In Sweden, 98 percent of the population has access to the Internet at home and virtually 100 percent of [16][17][18][19][20][21][22][23][24][25] year-olds use the Internet in mobile phones, tablets or computers. 33 Digital interventions have several advantages over traditional ones. They may reduce the stigma around risk behaviours, they are accessible at any time and place, and can be cost-effective because minimal staff resources are used for implementation, especially if they are fully automated. 34 35 Digital interventions may thus be particularly suitable for adolescents and young adults. 36 To date, digital prevention programmes targeting alcohol and other substance use among adolescents and young adults that meet scientific evaluation criteria are scarce. An early example is eScreen.se, offering screening for alcohol and drug use with personalized feedback. 37 A more recent example is a fully automated brief motivational intervention for substance-using [16][17][18] year-olds, tested in an RCT in four European countries, [38][39][40] where our research group was one of the partners. The intervention, named WISEteens, relied on Motivational Interviewing (MI) 41 and models of Social Influence. 42 43 The goal of the MI approach was to enhance motivation to change by exploring and resolving ambivalence about substance-related behaviours. An important element in the intervention was personalized feedback about one's substance use behaviours in relation to normative comparisons, as personalized feedback is perceived as more relevant for changing behavior than more general information. 44 The normative feedback included information about how a specific reference group actually consumed substances, in order to correct participants' "inflated perception", 45 which has proven effective for reduction of alcohol consumption in young adults in previous metaanalyses. 46 47 For adolescents and young adults, a normative feedback approach may be particularly appealing, assuming they are curious about how their substance consumption compares to their peers. 48 Likewise, an important element of the intervention was a focus on substance-related social norms and training on how to avoid social high-risk situations and how to resist peer pressure; i.e., raising refusal self-efficacy. 49 In line with the social influence hypothesis stating that a prominent risk factor for substance use in adolescence is the influence of substance using peers, 7 targeting peer pressure in brief digital interventions has been an important element for reducing alcohol use among adolescents. 50 WISEteens demonstrated significant between-group effects for alcohol use, indicating that a targeted brief motivational intervention in a fully automated digital format can be effective to reduce drinking and lower barriers for accessing substance use service in hazardous drinking adolescents. 38 Digital interventions, often delivered online, generally offer adaptations of evidence based face-to-face interventions, e.g., MI. The effects of MI on several health-related behaviours have been evaluated in a number of other studies, especially among adults. 51 Among adolescents, systematic reviews and metaanalyses on MI for various health-related behaviours show mixed results. A review of six studies of MI for reducing alcohol consumption in the emergency room setting suggested that MI was at least as effective as other brief interventions in the same setting. 52 Another review of 24 studies of different brief interventions for reducing alcohol consumption and alcohol-related problems among adolescents showed a significant effect that persisted up to one year compared to control interventions, with greater effects for MI. Effects were consistent over diverse settings and particularly effective components included decisional balance (juxtaposition of pros and cons of change) 53 and goal-setting 54 . Regarding MI for illicit drug use among adolescents, a review of 10 studies showed no effects on drug use behaviors; however, changes in attitudes towards drug use were found, which could be translated into intentions to change behaviors 53 . A scoping review concerning MI for reducing sexual risk behaviors among adolescents identified 29 unique studies with varying designs and conceptualizations of MI and specific risk behaviours, making it difficult to generalize regarding outcomes but indicating the need of more research. 55 MI has most commonly been delivered as an individual face-to-face intervention 51 , but has also been provided in other forms, e.g., via telephone or digitally, with various results on e.g., substance use behaviours. 56 57 Digital programmes based on MI have demonstrated effect in the form of reduced alcohol consumption among young adults. 58 59 Additionally, a combination of screening with a short intervention has shown similar effects on alcohol consumption among adolescents. 60 Brief motivational interventions are empirically supported individual level interventions for reducing alcohol consumption. 45 61 Such interventions have in the last decades also been digitally provided. 62 63 Digital interventions can be provided with or without human guidance, i.e., more or less automated. 64 Previous research has demonstrated that even fully automated interventions can reduce alcohol-related problems for young people with risky alcohol consumption up to 12 months after implementation, 59 65 and indicated potential  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   6 effect on cannabis use in certain groups. 66 67 Although various digital substance use prevention interventions have been developed and tested with promising results, 38 68 studies on the effects of fully automated digitally delivered and MI-based substance use prevention programmes for adolescents and young adults are still few 68 and more well-designed studies are warranted in order to obtain evidence for potential effects.

Aim
The study described in this protocol builds on results previously obtained in a European study of a digital brief MI-based substance use prevention intervention (WISEteens), delivered online to adolescents in a fully automated form. 38 In this study, the effectiveness of a modified version of this intervention, with new graphics and population based feedback-generating data on alcohol consumption covering all included age groups, will be tested across a range of outcome measures among 15-25 year-olds with hazardous alcohol or other substance use. The primary aim is reduction in frequency and quantity of substance use.

Rationale and hypothesis
The current study will add to existing evidence regarding the effectiveness of digital interventions aimed at reducing alcohol and other substance use among young people. We hypothesize that participants in the intervention group will report a significant reduction in substance use (primarily alcohol use), compared to an active control group receiving health information about various substances. A number of potential moderators will be assessed to confirm or contradict previous research on factors influencing the development of substance use patterns over time, and effects of prevention interventions in different target groups.

Recruitment
The target group for the current study is adolescents and young adults aged 15-25 years with self-reported substance use, a capacity to understand the Swedish language, and access to the Internet. Based on a power calculation to detect a small effect size (Cohen's d=0.2), 70 the aim is to include 800 participants. 71 The participants will be recruited during January-October 2020 by offline marketing at youth health clinics, other health care units for youth, and at upper secondary schools in Stockholm county, as well as by social networks, using online banner advertisements. This will result in a convenience sample from the general Swedish population. 44 Results from follow-up measures are to be completely received in 2021. If sufficient number of participants have not been recruited according to plan, we will extend the recruitment period and intensify the advertising by increasing visibility on social platforms. To enhance participation and follow-up rates, an incentive in the form of a movie ticket, with a value up to 15 Euros, for each completed follow-up assessment will be provided, i.e., up to 3 tickets. Up to three reminders will be sent out if participants do not respond to follow-up assessments.

Screening and informed consent
Potential participants will be guided to an online landing page with screening by an adapted version of the screening tool for use of alcohol and other substances, CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble). 72 This six-item screening tool has demonstrated criterion validity and appropriateness for identifying substance-related problems among adolescents. [73][74][75] The primary eligibility criterion for participating in the study will be a score of 2 or more on the CRAFFT, since this value has shown satisfactory sensitivity for identifying substance use problems. 72 Those who score 2 or higher will be offered participation in the study and given information on confidentiality, voluntariness of participation, and human subject protections. 44 They will also be provided with contact information on suitable counselling service providers. Those who agree to participate will be asked to give digital informed consent.

Measures
All assessment will occur online. The selection of measures and associated instruments is based on the aim of the study and the theoretical base concerning factors influencing substance use. Based on results  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 38 showing significant effects on alcohol consumption but not on other substance use, the primary outcome is alcohol use, measured at three-and six-month followups. Secondary outcomes include binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, average daily drinks per typical week, at three-and six-months follow-ups, other substance use, mental health, sexual risk behaviours, and perceived peer pressure at three and six-month follow-ups. The moderating variables include perfectionism, openness to parents, help-seeking due to mental health problems, and socio-demographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for students), and parents' education).
Alcohol use, including the primary outcome, will be measured using two instruments. One of them is the short version of the Alcohol Use Disorders Identification Test (AUDIT-C), 76 measuring frequency and amount of consumed alcohol and frequency of binge drinking, providing a widely used and valid index sum score for problematic alcohol use among adolescents. 75 77 The other is the Daily Drinking Questionnaire (DDQ), 78 measuring a variety of parameters of alcohol use in a typical week. Previous research has demonstrated that the DDQ is highly correlated with other measures of self-reported alcohol consumption. 79 The primary outcome will be measured by AUDIT-C total score. The two measures of alcohol use complement one another in that the AUDIT-C primarily offers an indication of hazardous or harmful use, 80 whereas the DDQ quantity measure (average daily drinks per typical week) will indicate the level of alcohol consumption in grams per week. Other substance use will be measured by a short version of the Drug Use Disorder Identification Test (DUDIT), 81 including the first four items (DUDIT-C) which assess frequency of consumption of drugs other than alcohol, frequency of different types of drugs other than alcohol used at the same occasion (1 = never to 5 = four or more times a week), and number of occasions when drugs other than alcohol are consumed on a typical day of drug use (1 = zero to 5 = seven or more). The DUDIT has been found to be effective in screening for drug-related problems in clinically selected groups, 82 and both the DUDIT and the DUDIT-C have proven useful in the context of public health surveys in Sweden. 81 83 Mental health will be measured by The WHO-5 Well-Being Index, which has demonstrated validity both as a screening tool for depression and as an outcome measure in clinical trials. 84 This scale has been successfully applied across a wide range of study fields, translated into more than 30 languages, including Swedish, 85 and found psychometrically sound. Changes in sexual risk behaviours will be measured by multiple choice questions on sexual behaviour under the influence of alcohol and/or other substances and unprotected sex, previously used in a survey among Swedish visitors at youth health clinics in Stockholm county. 86 Changes in perceived peer pressure will be measured by  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   9 two items retrieved and adapted from the Peer pressure inventory. 87 Perfectionism will be measured by two subscales of the Frost Multidimensional Perfectionism Scale (FMPS). 88 89 Openness to parents will be measured among participants who are up to 20 years old with questions about disclosure from Stattin & Kerr, 90-92 slightly modified into statements to be compatible with the digital online format. Help-seeking due to mental health problems will be measured by dichotomous questions regarding if, and to what healthcare provider, the participant has turned for help. Additionally, background data will be asked for using multiple choice questions: sex (man, women, other), age (15, 16…25), residence (name of municipality), occupation (secondary school, upper secondary school, university, not student, working, practicing, unemployed, other), school performance (merit value for students), and parents' education (secondary school, upper secondary school, university, don't know).

Randomisation
The study is double-blind, thus neither the participant nor the researchers will know which participant is allocated to the intervention or to the active control condition. After baseline assessment, the participants will be automatically randomised to one of the two study groups by a computer programme using an unrestricted randomisation protocol. Participants will then be informed about the name of their programme and given access immediately.

The intervention
Participants allocated to the intervention group will initially be asked to give additional information on alcohol and other substance use and to state their body weight, in order to generate personalized feedback on substance use behaviour and to estimate alcohol blood concentration on a typical drinking occasion.

Description of the intervention content
The intervention is interactive, digitally delivered online in a fully automated form, and requires approximately 20 minutes to complete. Personalized feedback is given to the participants based on their responses to previous assessment and suggestions on how to respond to this feedback are provided. This interactivity imitates a face-to face "dialogue" with techniques from MI such as an empathic approach, rolling with resistance, aiming at creating a dissonance between actual and desired behaviour, raising self-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 10 efficacy, and at the same time avoiding argumentation. 57 The intervention consists of three main components outlined below, and additional health related information.

Personalized feedback
The personalized feedback includes an estimation of the participants' blood-alcohol-concentration (BAC) and information on the associated risks concerning the participants' heaviest drinking episode during the past 30 days. The value will be based on a measure of Peak Drinking Quantity 93 and estimated using the Widmark-formula, which takes into account weight and sex. 94 The participants will receive graphed feedback regarding number of standard alcohol units per week that they think their peers consume (descriptive norms), as well as the participant's individual levels of consumption in relation to comparative data (actual drinking levels) from a reference group. Comparative data (AUDIT-C scores) will be taken from alcohol prevalence estimates found in a nationally representative sample of 16-25 yearolds undertaken by the Public Health Agency of Sweden. Comparative feedback will be available for drinking but not for substances other than alcohol.

Interactive MI-based exercises
The exercises provided in the current intervention build on the assumption that participants may hold certain levels of ambivalence about their current substance use, and that if they are willing to make a change they may not know how to, or may not be confident that they are able to. 41 Therefore, the intervention uses importance and confidence rulers with a short summary and feedback to encourage participants to reflect on personal reasons for change and explore personal strengths and ability to change.
Furthermore, the programme provides a decisional balance to pick up and graphically illustrate potential levels of ambivalence by offering the participants a list of possible pros and cons regarding the decision to change their current substance use. 57 95 Participants are instructed to choose statements that apply to them and are presented with the resulting balance sheet of their personal comparative potential gains and losses.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   11 Practical advice The intervention focuses on raising self-efficacy for being able to avoid drinking in social situations, if desirable. 50 The participant will be asked to select three among 12 provided drinking situations that they consider most tempting and rank them. The situations are adapted from the adolescent version of the Drinking Refusal Self-efficacy Questionnaire (DRSEQ-RA). 96 According to the selection, a number of strategies will be offered for each of the selected drinking situations to provide participants with a tool kit necessary for engaging in and maintaining their behavioural goal.

Health related information connected to substance use
Finally, the intervention programme includes health related information associated with substance use.
The information is provided optionally throughout the program behind "read more" buttons and also at the end of the programme. The information contains statements regarding risks connected to substance use that the participant can reflect on, as well as optional links for more information. There are also two FAQsections about alcohol and cannabis, respectively, as well as information about how different substances may affect the individual, both physically and mentally. Finally, the information also contains cases, were the participant can read about some typical young persons who have used substances, why they used and what negative effects they have noted, and why they chose to stop using.

The control condition
The control group will receive the same general health related information as the intervention group, i.e., the additional information which is connected to the intervention programme.

Statistical analysis
Randomisation-checks of baseline variables regarding alcohol and other substances use and psychological state will be conducted using multivariate analyses of variance MANOVA. To test the effectiveness of the intervention, we will assess whether participants in the intervention group to a greater extent report decreased substance use, sexual risk behaviours, perceived peer-pressure, and improved mental health after three and six months, respectively, compared to participants in the active control group. Data analyses will consist of comparing outcome measurements with regards to within-group and betweengroup differences according to the intention-to-treat (ITT) principle in the primary analysis, accounting for all included participants regardless of whether or not they completed follow-up assessments. We will also perform per protocol analyses. The reason for choosing ITT in the primary analysis is the ambition to  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 F o r p e e r r e v i e w o n l y maintain the power of the study and also to avoid biased results due to selection of those completing all follow-ups, as they may have special characteristics not representative for the whole study group, which might influence study results. The main analysis of effectiveness will use mixed effects regression models, which can be applied to both continuous and categorical outcomes and also non-normally distributed outcomes. Also, mixed-effects regression models are robust to missing data in longitudinal studies. 97 In the current study, we assume a quite large dropout rate, based on previous research. For example, a metaanalysis comprising 17 studies of unsupported (fully automated) interventions for depression showed 74% dropout rate at post-treatment. 98 Moreover, we expect data missing at random (MAR), which can be handled using mixed effects regression models. Separate models will be run to test each outcome, i.e., alcohol consumption, binge drinking, frequency of alcohol consumption, amount of alcohol consumed a typical day when alcohol is consumed, average daily drinks per typical week, other substance use, mental health, sexual risk behaviours and perceived peer pressure at three-and six-month follow-ups. Effect size will be calculated for the primary outcome variable (AUDIT-C total score). Potential moderators, including perfectionism, openness to parents, help-seeking due to mental health problems, and sociodemographic and personal characteristics, (i.e. sex, age, residence, occupation, school performance (for students), and parents' education) will serve as covariates in analyses of intervention effects. If moderator effects are found, we will carry out post-hoc analyses stratified by the detected variable.

Patient and public involvement statement
Patients and the public were not involved in the design and planning of the study, except for persons at the age of the target group reading manuscripts for the revised version of WISEteens and also participating in a pilot-test.

ETHICS AND DISSEMINATION
The current study was approved by the Swedish Ethical Review Authority (no. 2019-03249) and registered 24/09/2019, pre-results, on IRCTN, ID: ISRCTN91048246. Any important protocol modifications will be reported to IRCTN. For inclusion, all participants must give informed consent online prior to participation in the study. In order to participate, the participants will need to state a username (which may be fictitious or a pseudonym) and an email address. All data collection will be done without collecting personal identification information, only personal e-mail addresses. The e-mail address will be used to connect data from the baseline measurement to the follow-up measurements. At a later stage, the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 13 raw data file will be anonymized and each person assigned a number instead of the e-mail address. The data will be stored in line with routines for handling and storing research data at Karolinska Institutet. All data is handled confidentially and will not be forwarded to third parties. Participating in the present study means that the participants need to reflect on their alcohol consumption and other substance use. In addition, questions will be asked concerning personal circumstances, including the participants' mental health and family relationships. These issues may be perceived as somewhat unpleasant. However, in the information that potential participants receive, they are informed that participation is voluntary and that they at any time can end their participation without explaining why. In addition, there will be reference to other types of official support (web pages and telephone numbers), if the advertisement or participation leads to concern about own substance use or related problems. Any issues brought up by the participants during the study will be documented and handled properly. Moreover, the research team includes professionals, such as a psychiatrist and a nurse, with possibility to refer participants to health care clinics if needed. The project's basic hypothesis is that the intervention will have positive effects, with regard to alcohol and other substance use among young people with risk use, and potential benefit may include decreased or ceased risk behaviours. Overall, the benefit for the research persons is considered to exceed any risk of discomfort. Results will be disseminated in scientific peer reviewed journals, at conferences and via the media.

DISCUSSION OF STRENGTHS AND LIMITATIONS
The current study has a number of strengths. The protocol describes a two-arm double-blind RCT, considered to be the most robust experimental design, controlling for selection bias and participant allocation bias and with the possibility to make causal inferences. Moreover, an active control condition, blinded to the study condition, controls for expectation, detection and performance bias. The collection and analysis of information on potential moderators, allowing for control of these factors, facilitates the understanding of the possible effects. Thus, the present study will contribute to the literature on digital substance use prevention interventions among adolescents and young adults in several ways. The choice to study a digital intervention can from a public health perspective be regarded as positive. Digital brief interventions have several advantages over face-to-face approaches, e.g., the reduction of stigma around help-seeking for substance use and easy dissemination to large groups of people, which makes them costeffective. Moreover, digital interventions have previously proven effective in addressing alcohol and other substance use in the general population and some studies have shown promising results also among adolescents and young adults. 99 100 The current intervention is well grounded in theory and incorporates elements of MI and social influence theory that has been shown to be effective in reducing problematic  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 14 substance use in prior research. 57 Importantly, there are also some limitations to the current study. One limitation concerns selection bias and thereby external validity, as recruitment requires either that individuals click on our ads at social media to be considered for inclusion, or that they attend upper secondary schools, or have been in contact with existing clinics, aware of the opportunity to participate in the study. Thus, our participants may have certain personality traits, or are especially prone to helpseeking compared to a broader audience. The exclusion of people not understanding Swedish is also a limitation with regard to generalizability.

STATEMENTS Author contributions
PK, AKS, JG, and THE obtained funding for the study. PK, THE, and CS further developed/modified the intervention. THE designed the study with contribution from PK, AKS, JG, AHB, and, CS. PK wrote this paper. All authors commented on successive manuscript drafts and approved the final version of the manuscript. bodies had no role in study design, data collection, analysis, data interpretation or writing manuscripts.

Results
13a For each group, the numbers of participants who were randomly assigned, received intended treatment, and were analysed for the primary outcome Sources of funding and other support (such as supply of drugs), role of funders 13 *We strongly recommend reading this statement in conjunction with the CONSORT 2010 Explanation and Elaboration for important clarifications on all the items. If relevant, we also recommend reading CONSORT extensions for cluster randomised trials, non-inferiority and equivalence trials, non-pharmacological treatments, herbal interventions, and pragmatic trials.