The Body Project is a group-based ED prevention program based on the principles of cognitive dissonance. According to Festinger [27], cognitive dissonance occurs when there is an inconsistency between one’s beliefs and one’s actions. This inconsistency creates psychological discomfort, which motivates the individual to reduce cognitive discord by changing their beliefs to align with their actions. In the Body Project, participants collectively critique the thin beauty ideal through a series of verbal, written, and behavioral exercises. Engaging in these activities prompts participants to align their attitudes with their publicly displayed behaviors, which results in a reduction in pursuit of the thin beauty ideal, which in turn is theorized to reduce body dissatisfaction and ED symptoms. Consistent with this theory, reductions in pursuit of the thin beauty ideal mediated the effects of the Body Project on ED symptom reduction [28]. During the four sessions, the participants define the appearance ideal, discuss costs of pursuing this ideal, attempt to dissuade group facilitators from pursuing the appearance ideal in role-plays, and engage in discussions about the negative effects of social comparison [29] and social media. Written home exercises are conducted between sessions and debriefed with the group in the following meeting. We have previously translated the latest version of the Body Project script [15] to Norwegian for the first time and piloted a diabetes-specific version among Norwegian young women with type 1 diabetes [30]. No further adaptations were considered necessary to obtain cultural fit, which is similar to experiences reported in the Swedish virtual Body Project study [16]. The Body Project is largely participant-driven given the Socratic open questions in line with dissonance-theory, which may explain why minimal changes are required to obtain efficacy in different countries. For example, in session 1 group facilitators ask what the society tells us that young women should ideally look like, by which the participants themselves define the appearance ideal in their own cultural context.

A total of 441 young women will be included in the study. Inclusion criteria are being female, age between 16 and 25 years, and at least some level of self-reported body dissatisfaction verbally confirmed by the participant in the screening call, following procedures used in previous trials [14, 15, 31]. Participants diagnosed with an ED will also be included. These individuals have shown to display larger reductions in ED symptoms than participants without EDs in response to the Body Project d = 0.71 vs. 0.17) [32]. Exclusion criteria include being hospitalized or currently in treatment for an ED.

A Norwegian website (www.bodyproject.no) has been designed to facilitate recruitment and project profiles are posted on Facebook and Instagram. The link will be shared via our online platforms, including Facebook, Instagram, and TikTok. Also, advertisement campaigns on social media will be run nationally. Individuals express interest by sending an e-mail to a Body Project mailbox hosted by Oslo University Hospital, which leads to a screening phone call to determine eligibility based on the inclusion criteria described above. In this screening call eligible participants verbally confirm to experience at least some body dissatisfaction. After providing informed consent, participants will complete baseline assessment (pretest) before being randomized into one of three arms: (i) peer-led virtual Body Project groups; (ii) clinician-led virtual Body Project groups; and (iii) educational video control group. Assessment will be repeated post intervention (posttest) and follow-up assessment after 6-months, 1-year, and 2-years. ED diagnostic interviews are conducted at baseline, 1-year, and 2-year assessments by research staff who are masked to the condition to which each participant will be randomized. Online assessment will be conducted via an online survey (Viedoc) hosted by Clinical Trials Unit (CTU), Oslo University Hospital. Randomization will be conducted by Viedoc once baseline assessment is completed. Virtually delivered Body Project group sessions will be implemented by use of Microsoft Teams, administered by Oslo University Hospital. Adaptations required for virtual script delivery include administration of home exercises via e-mail and using the chat function for summarizing themes during in-session group discussions rather than white boards or flip-charts used for in-person Body Project groups. This study is funded by the Norwegian Health Authority Region South-East (reference 2,023,032).

Clinician group facilitators consist of nurses (psychiatric and school health nurses), clinical dietitians, and a social worker, who are licensed health care personnel associated with the Regional Department for Eating Disorders at Oslo Universal Hospital. Peer group facilitators are recruited through acquaintances of our research team, and includes university and college-level students (not within medicine, psychology, or nursing) and maximum four years older than the participants.

Facilitators (both clinicians and peer educators) will read the intervention manual that outlines the theoretical rational of the intervention, reviews the evidence-base for the Body Project, and contains a detailed intervention script. Facilitators will then attend an 8-hour virtual workshop, where the conceptual rationale and supporting evidence for the Body Project are presented, facilitators role-play delivery of the intervention using the scripted manual, and process issues are discussed (e.g., promoting session attendance and homework completion, and recording attendance, homework completion, and verbal participation in the sessions for each group participant). All Body Project sessions will be videotaped, and all sessions from the first group implemented by facilitators will be rated for intervention fidelity and therapeutic competence (see below for psychometrics); supervision based on these ratings will be provided to facilitators via e-mail (feedback on what went well, and advice on how certain aspects may be improved for future sessions), which has been found to significantly increase intervention effects compared to trials that did not provide this type of ongoing supervision [10]. This study will use the same procedures for training and supervision as past Body Project trials [31, 33].

An educational control condition was selected to control for expectancy effects and demand characteristics. Pre-existing freely available videos allow participants to view them over the Internet to match the virtual delivery modality (1 h per week over 4 consecutive weeks). The videos contain information about how to promote a positive body image, costs associated with EDs, and how to seek help if they experience concerns related to body image, eating, body pressures, and weight. Past trials of the Body Project have utilized such educational videos as the control condition [15, 33, 34]. Results indicated that participants assigned to the educational control condition showed reductions in ED symptoms compared to assessment-only controls [34], but that the reductions in ED symptoms were significantly larger for Body Project participants than for control participants. Adherence to watching the videos will be recorded by self-report at posttest assessment. In a previous trial participants showed similar adherence to attending Body Project group sessions versus watching the educational videos (typically over 80%) [35].

All measures are translated to Norwegian. Only the Eating Disorder Examination Questionnaire (EDE-Q) and Eating Disorder Assessment for DSM-5 (EDA-5) are validated in Norwegian, while the psychometric properties of the remaining measures are validated in English. See details below:

Thin-ideal internalization: the 8-item Ideal-Body Stereotype Scale-Revised (IBSS-R) [36] to assess endorsement of thin-ideal using a Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. It has shown internal consistency (α = 0.91), 2-week test-retest reliability (r = .80), predictive validity for future ED onset, and sensitivity to detecting intervention effects [33].

Body dissatisfaction: ten items from the Body Parts Scale [37] to assess dissatisfaction with body parts with a Likert scale ranging from 1 = extremely dissatisfied to 6 = extremely satisfied. It has shown internal consistency (α = 0.94), 3-week test-retest reliability (r = .90), predictive validity for future ED onset, and sensitivity to detecting intervention effects [33].

Negative affect: 20 negative items from the Positive and Negative Affect Schedule – Revised (PANAS-X) [38] to measure negative affect. It has high internal consistency (mean Cronbach’s α of 0.94), test-retest reliability (r = .78), and convergent validity [15].

ED symptoms: The Eating Disorder Examination Questionnaire (EDE-Q) [39] is a self-report 28-item questionnaire of specific eating disorder psychopathology based on the Eating Disorder Examination (EDE) diagnostic interview [40]. Reponses range from 0 to 6 with higher scores indicating higher levels of ED psychopathology. The EDE-Q consists of the four subscales eating restraint, eating concern, shape concern, and weight concern and is previously translated and validated in Norwegian men and women [41, 42].

Appearance ideal internalization and pressures: the Social Attitudes Towards Appearance Questionnaire 4 – Revised (SATAQ-4R) for women [43] consists of 31 items with subscales which measure: internalization thin/low body fat, internalization of muscularity, internalization of general attractiveness, pressures from family, pressures from peers, pressures from significant others, and pressures from media. Internal consistency (Cronbach’s α ≥ 0.82) and construct validity have been demonstrated.

Implicit attitudes: the Implicit Association Test (IAT) [44] measures implicit attitudes to the thin beauty ideal. The IAT is a widely used measure of implicit attitudes based on the strength of automatic associations between concepts and positive versus negative adjectives, which assesses attitudes toward images of the thin appearance ideal and words describing ED behaviors. It has shown good psychometric properties [45]. Participants are asked to quickly sort words into categories in a series of seven blocks that vary in whether the thin images and ED words are paired with positive or negative words. Response latencies are recorded according to standard.

ED diagnoses: the Eating Disorder Assessment for DSM-5 (EDA-5). The EDA-5 [46] is as a semi-structured, web-based diagnostic interview assessing DSM-5 feeding and eating disorders that have been present in the previous three months. The interview can be administered with limited training [47] and is significantly quicker to administer than traditional diagnostic interviews for EDs. High rates of agreement have been found between diagnoses by EDA-5 and traditional clinical interviews, such as the Eating Disorder Examination (EDE) interview, supporting its diagnostic validity [46, 48]. A recent validation study [48] supports the validity of the Norwegian EDA-5 to efficiently generate ED diagnoses without compromising diagnostic accuracy.

All sessions will be recorded and key elements of a randomly selected subset of sessions will be rated for degree of fidelity (100-point scale from 0 = No adherence; the section was skipped to 100 = Perfect; all material in the section was presented as written). Facilitator competence will be rated with 12 items (e.g., leaders express ideas clearly and at an appropriate pace) using a 100-point scale with anchors for each item (e.g., 20 = Poor; leaders are difficult to follow and session proceeds at an uncomfortable pace, 100 = Superior; leaders are unusually articulate and express ideas in ways that all group members understand; perfect pace). Fidelity ratings and competence ratings have shown inter-rater agreement (ICC = 0.92 and = 0.96, respectively) for the original Body Project [49]. The PI will be responsible to do these ratings and provide supervision to group facilitators.

Model building. Intent-to-treat (ITT) analyses of condition effects will be evaluated using mixed effects growth models fit with SAS 9.2 PROC MIXED [50]. Individual variability in outcomes from posttest to 2-year follow-up was modeled adjusting for pretest outcome values. Following Singer and Willet [51], when constructing the longitudinal models we will (a) examine empirical growth plots; (b) fit an unconditional means model; (c) fit an unconditional linear growth model; (d) fit unconditional non-linear models; and (e) compare models of longitudinal change using the Akaike Information Criterion. Condition, time (coded in months since posttest), and a condition × time interaction will be added to the unconditional model. We will estimate a growth model for each continuous outcome. We will report the results of planned contrasts from these models, contrasting each pair of conditions, reporting condition differences in the model implied least-square means derived from the time and condition × time interactions at 1- and 2-year follow-up. We will report effect sizes that are equivalent to Cohen’s d [52]. We will use full information maximum likelihood (ML) estimation to impute missing data because this intent-to-treat approach produces more accurate and efficient parameter estimates than alternative analytic approaches such as analyzing only participants who provide complete data, making multiple imputation the recommended procedure for handling missing data [53]. Missing data will be imputed using PROC MI using baseline levels of the outcomes, condition, and demographic factors (age, race, and parent education). Based on the recommendations [53], we will impute 50 data sets. Model parameters and standard errors were combined following Rubin [54] and implemented using PROC MIANALYZE. Cox proportional hazard models, fit with STATA [55] will test whether eating disorder onset over 2-year follow-up was significantly lower in each condition versus each other condition. Schoenfeld residuals will be examined to test the proportional hazards assumption. Hazard ratios and number needed to treat (NNT; [56]) are provided as measures of effect size. Because hazard models accommodate right censoring, we will not impute missing incidence data. Given that 88% of the effects from previous Body Project trials have replicated in 22 controlled trials from independent teams [15], we will use a p-value of 0.05 for all inferential tests because this usually balances risk for type I and type II errors.

Sample size calculation was made for incidence of EDs assessed over a 2-year follow up via diagnostic interviews as the primary endpoint. Incidence rates were obtained from the previous Swedish virtual Body Project RCT [16]. Based on this trial we postulated a difference in incidence of EDs of 8.8% and 2% between groups, and a hazard ratio of 0.26 during 2 years of follow-up. Then, with a significance level of 5% and a power of 90% the estimated number of patients needed was 147 in each arm, the total number needed was 441.