Body Project

Study 1

Summary

The study sought to determine the effectiveness of the dissonance intervention compared to three other conditions: healthy weight intervention, expressive-writing control intervention, and assessment-only control intervention. A total of 481 participants were randomly assigned to one of four conditions: dissonance intervention (n=115), healthy weight intervention (n=117), expressive-writing control intervention (n=123), or assessment-only control intervention (n=126). Data were collected at pretest, posttest, and at 6-month, 1-year, 2-year, and 3-year follow-ups after posttest.

The study found that the dissonance intervention relative to assessment-only or expressive writing controls produced significantly greater improvement in:

• Thin-ideal internalization at posttest, 6-month, 1-year, and 2-year follow-up
• Body dissatisfaction at posttest, 6-month, 2-year, and 3-year follow-up
• Dieting at posttest, 6-month, and 1-year follow-up
• Negative affect at posttest, 6-month, 2-year, and 3-year follow-up
• Bulimic symptoms at posttest, 6-month, 1-year, and 2-year follow-up
• Risk of obesity onset over the 1-year follow-up
• Psychosocial impairment at 2-year and 3-year follow-up
• Risk for eating pathology onset over the 3-year follow-up.

Evaluation Methodology

Design: Using a randomized controlled design, researchers sought to determine the effectiveness of the dissonance intervention (Body Project) compared to three other conditions: healthy weight intervention, expressive-writing control intervention, and assessment-only control intervention. The healthy weight intervention had the same structure as the dissonance intervention but used motivational interviewing, behavior modification, journalizing and goal setting to challenge the thin-ideal, find benefits for, and reinforce, lifestyle changes, and provide solutions to barriers encountered when changing. The expressive writing condition consisted of 3 weekly 45-minute individual writing sessions.

Participants were recruited for this study from high schools and one university using mailings, flyers and leaflets that invited female participants between the ages of 14 and 19 with body image concerns to participate in the study to help them accept their bodies. The only exclusion criterion was meeting the diagnosis for an eating disorder.

A total of 566 female participants were assessed for eligibility, 44 (7.7%) refused to participate and 41 (7%) did not meet criteria. The remaining 481 participants were randomly assigned to one of four conditions: dissonance intervention (n=115), healthy weight intervention (n=117), expressive-writing control intervention (n=123), or assessment-only control intervention (n=126). Data were collected at pretest, posttest, and at 6-month, 1-year, 2-year, and 3-year follow-ups after posttest. The mediation analysis also reported on data collected weekly from baseline to posttest.

Attrition was low with only 5% of participants not completing assessments through the 3-year follow-up.

Sample Characteristics: The sample was adolescent girls with a mean age of 17 years and with the following racial identities: 10% Asian/Pacific Islander, 6% African American, 19% Hispanic, 58% White, and 7% who specified other or mixed racial heritage. Parental education ranged from high school graduate or less (17%) to graduate/professional degree (28%).

Measures: Eight areas were measured in Stice et al. (2006). In the long-term follow-up results (Stice, Marti, Spoor et al., 2008), dieting and health service utilization were left out of the tables and social functioning was renamed psychosocial impairment. Assessments were conducted through self-report and interview using assessors blinded to the condition. The mediation analysis (Stice et al., 2007) examined two other mediating variables - healthy eating and physical activity.

Thin-ideal internalization was assessed using The Ideal-Body Stereotype Scale-Revised which uses a response format ranging from 1 = strongly disagree to 5 = strongly agree. This scale has shown internal consistency (α = .91), test-retest reliability (r = .80), and predictive validity for bulimic symptom onset.

Body dissatisfaction was assessed with 8 items from of the Satisfaction and Dissatisfaction with Body Parts Scale that assesses body parts that are often of concern to females (e.g., stomach, thighs, and hips). Participants indicate their level of satisfaction with body parts on scales ranging from 1 = extremely satisfied to 6 = extremely dissatisfied. This scale has shown internal consistency (α = .94), 3-week test-retest reliability (r = .90), and predictive validity for bulimic symptom onset.

Dieting was assessed using The Dutch Restrained Eating Scale (DRES) where participants indicate the frequency of dieting behaviors using scales ranging from 1 = never to 5 = always. The DRES has shown internal consistency (α = .95), 2-week test-retest reliability (r = .82), convergent validity with self-reported caloric intake, and predictive validity for bulimic symptom onset.

Negative affect was assessed with the sadness, guilt, and fear/anxiety subscales from the Positive Affect and Negative Affect Scale-Revised (PANAS-X). Participants reported the extent to which they had felt various negative emotional states on scales ranging from 1 = very slightly or not at all to 5 = extremely. This scale has shown internal consistency (α = .95), 3-week test-retest reliability (r = .78), convergent validity, and predictive validity for bulimic symptom onset.

Bulimic symptoms were assessed using the diagnostic items from the Eating Disorder Examination (EDE), a semi-structured investigator-based interview. Items assessing the symptoms in the past month were summed to create an overall bulimic symptom composite for each assessment point. The adapted symptom composite showed internal consistency (α = .86 at pretest) and 1-month test-retest reliability for assessment-only controls (r = .81).

BMI was used to reflect obesity and adiposity and was calculated from height (measured to the nearest millimeter using stadiometers) and weight (assessed to the nearest 0.1 kg using digital scales). Two measures of each were obtained and averaged.

Social functioning (or psychological impairment) was assessed using items adapted from the Social Adjustment which demonstrated internal consistency (α= .76) at pretest.

Health service utilization was measured using two items assessing the frequency of visits to health and mental health care providers (e.g., How often have you seen a doctor (physician) because of illness, injury, long-term health problems, or for regular check-ups in the last 6 months). The items showed 1-year test-retest reliability for assessment-only controls (r= .82 and r= .89 respectively).

Healthy eating was measured using four items that were averaged to form an overall healthy eating composite. This scale was internally consistent (α = .82) at pretest and temporally reliable (1-month test-retest r = .63 for assessment-only controls).

Physical activity was measured using two items assessing weekly frequency and duration of exercise and the two items were multiplied to form a quantity × frequency measure of exercise behavior at each assessment. This scale was temporally reliable (1-month test-retest r = .61 for assessment-only controls).

Analysis: Stice et al. (2006) used omnibus repeated measures ANOVA models to test whether there were differential changes in the outcomes across the four conditions up to one-year follow-up (condition was a 4-level between-subjects factor and time was a 4-level within-subject factor). Next, the researchers examined ANOVA models that contrasted pairs of conditions at each of the follow-up periods to determine precisely which groups differed significantly and how long these significant differences persisted. In each of these repeated measures models condition was a 2-level between-subjects factor and time was a 2-level within-subjects factor (pretest to posttest, pretest to 6-month follow-up, and pretest to 1-year follow-up). Full information maximum likelihood estimation was used to impute missing data.

Stice, Marti, Spoor et al. (2008) examined outcomes at 2- and 3-year follow-ups using mixed model analyses that included all five waves of post-intervention data as dependent variables (posttest, and follow-ups at 6-month, 1-year, 2-year and 3-year). Independent variables included condition, wave, condition × wave interaction, pretest level of the outcome, the pretest of the outcome × wave interaction, posttest to one-year slope of the outcome, and the posttest to one-year slope × wave interaction. There were 5 contrasts that tested for the hypothesized effects at both 2-year follow-up and 3-year follow-up in each model: dissonance vs. assessment only, dissonance vs. expressive writing, healthy weight vs. assessment only, healthy weight vs. expressive writing, and dissonance vs. healthy weight. Baseline versions of the outcomes and change in the outcomes from post-test to 1-year follow-up were controlled for in an effort to ensure that differences in the marginal means at 2- and 3-year follow-up were a result of the interventions rather than these covariates.

Stice et al. (2007) used multilevel analysis and multiple imputation of missing data but included a full mediation analysis.

The studies adhered with the intent-to-treat principle by gathering data and analyzing results from all participants, regardless of dose of intervention received.

Outcomes

Implementation Fidelity: Facilitators received a detailed and scripted manual for each of the intervention conditions (dissonance and healthy weight) as well as training and direct observation by the first author. A random sample of sessions (10%) were audio taped and coded by research assistants who were blinded to intervention condition to determine whether the facilitators covered the main exercises listed in the relevant treatment manual and did not cover the main exercises listed in the other treatment manual (each session contains 4-6 main exercises). Results indicated that group leaders covered 100% of the main exercises from the relevant manual and included 0% of exercises from the other manual.

Further, the study reported that session attendance was high and did not differ across conditions. Among dissonance participants, 91% attended all three sessions, 6% attended two sessions, and 3% attended one session. Parallel attendance rates were 91%, 6%, and 3% for the healthy weight intervention and 98%, 1%, and 1% for the expressive writing condition. The proportion of participants who completed the homework assignments was not significantly different in the dissonance (80%) versus the healthy weight condition (87%).

Baseline Equivalence: The four groups did not differ significantly on any of the demographic or outcome variables.

Differential Attrition: The study reported low levels of attrition (4% at 1-year follow-up and 5% at 3-year follow-up) and reported that the attriters did not differ from participants who completed all the assessments on any demographic or outcome variables at pretest. Further, the study reported that attrition did not differ across conditions. The mediation analysis examined a shorter time span and had attrition of 4%. Checks showed no differences across dropouts and completers.

Posttest and 1-year follow-up (Stice et al., 2006): Across all four conditions, significantly differential change over time was found for thin-ideal internalization (p < .001, r = .20), body dissatisfaction (p < .001, r = .20), dieting (p = .001, r = .14), negative affect (p = .006, r = .13), and bulimic symptoms (p < .007, r = .13). When looking at specific contrasts, the dissonance intervention relative to assessment-only or expressive writing controls produced:

• significantly greater reductions in thin-ideal internalization, body dissatisfaction, dieting, negative affect and bulimic symptoms at posttest and 6-month follow-up
• significantly greater decreases at 1-year follow-up in thin-ideal internalization, dieting, and bulimic symptoms

Further, the dissonance intervention produced significantly greater decreases in thin-ideal internalization, body dissatisfaction, dieting, negative affect, and bulimic symptoms relative to the healthy weight intervention from pretest to posttest; and it produced significantly greater decreases in negative affect at the 6-month and 1-year follow-ups.

Additional analyses showed significantly higher reliable change rates in bulimic symptoms and lower risk of obesity onset for the dissonance intervention condition than the control groups over the 1-year follow-up.

2- and 3-year follow-up (Stice, Marti, Spoor et al., 2008): Compared to assessment-only controls, dissonance intervention participants showed significantly lower thin-ideal internalization, body dissatisfaction, negative affect, bulimic symptoms and psychosocial impairment at 2-year follow-up and significantly lower body dissatisfaction, negative affect and psychosocial impairment at 3-year follow-up.

Compared to expressive writing controls, dissonance intervention participants showed significantly lower thin-ideal internalization and body dissatisfaction at 2-year follow-up and significantly lower body dissatisfaction and psychosocial impairment at 3-year follow-up.

Further, dissonance participants showed significantly greater reductions in psychosocial impairment than healthy weight participants by the 3-year follow-up.

Lastly, dissonance participants showed significantly lower risk for eating pathology onset than assessment-only controls (p = .022) over the 3-year follow-up.

Eating Disorder Onset Through 3-Year Follow-Up (Stice, Rohde, Gau et al., 2012). This article used classification tree analysis to identify the most important risk factors for eating disorder onset and to test for moderation of intervention effects. The authors concluded that "Denial of costs of pursuing the thin-ideal was the most potent risk factor for eating disorder onset during the 3-year follow-up (OR = 5.0). The dissonance program mitigated the effect of this risk factor."

Mediation Analysis (Stice et al., 2007): The mediation analysis used three of the conditions, comparing both dissonance intervention and healthy weight to expressive writing and excluding the no treatment control group. The sample size of 355 was assessed at four time points from baseline to posttest: before the intervention (baseline), after session 1, after session 2, and after session 3 (posttest). The analysis did not use the follow-up data.

The mediation model for the dissonance condition tested thin-ideal internalization as a mediator of four outcomes - body dissatisfaction, dieting, negative affect, and bulimic symptoms - that were linked theoretically to the program goals.

The results indicated that thin-ideal internalization partially mediated the effects of the intervention on the outcomes. Specifically, 1) the intervention group experienced significantly greater improvement in all four outcomes, 2) the intervention group experienced significantly greater improvement in the mediator, 3) changes in thin-ideal internalization significantly affected the outcomes, and 4) the effect of the intervention on the outcomes was significantly reduced after controlling for change in the mediator which accounted for 26% to 48% reduction in the main effects.

Moderation Analysis (Stice, Marti, Shaw & O'Neil, 2008). This study examined intervention moderators for the period from baseline to the one-year follow-up. The effects of the Body Project intervention on bulimic symptoms were significantly stronger for participants with initial elevations in body image distress, bulimic symptoms, and thin-ideal internalization.

Study 2

Summary

The study evaluated the effectiveness of the dissonance intervention against a healthy weight intervention and a waitlist control. A total of 148 female students were randomized to one of three conditions resulting in the following distribution after attrition: dissonance intervention (n=48); healthy weight intervention (n=51); or waitlist control (n=36). Data were collected at baseline, post-intervention, and 1-, 3- and 6-month follow-ups.

The study found significant differences between dissonance intervention participants and waitlist control participants for reductions in:

• Negative affect
• Bulimic symptoms.

Evaluation Methodology

Design: Using a randomized controlled design, researchers sought to evaluate the effectiveness of the dissonance intervention (Body Project) against a healthy weight intervention and a waitlist control. Using flyers and announcements in large classes, participants were recruited from three high schools and one university. A total of 148 female students were randomized to one of three conditions, however 14 students dropped out during the intervention resulting in the following distribution: dissonance intervention (n=48); healthy weight intervention (n=51); or, waitlist control (n=36). No information was provided on the number recruited to obtain the initial sample size of 148. Further, even though the study reports that 14 students (9%) dropped out during the intervention leaving a sample size of 134, the analyzed sample size included 135 students. Data were collected at baseline, post-intervention, and 1-, 3- and 6-month follow-up.

Sample Characteristics: The sample of adolescent girls had a mean age of 17.4 years and the following racial identities: 13% Asian/Pacific Islander, 2% African American, 16% Hispanic, 63% White, and 6% who specified other. Parental education ranged from high school graduate or less (2% of fathers and 1% of mothers) to advanced degree (20% of mothers and 39% of fathers).

Measures: Five measures previously used in other studies were gathered in this study:

Thin-ideal internalization was assessed using the Ideal-Body Stereotype Scale-Revised, which uses a response format ranging from 1 = strongly disagree to 5 = strongly agree. This scale has shown internal consistency (α = .89), test-retest reliability (r = .63), and predictive validity.

Body dissatisfaction was assessed with 8 items from of the Satisfaction and Dissatisfaction with Body Parts Scale that assesses body parts that are often of concern to females (e.g., stomach, thighs, and hips). Participants indicate their level of satisfaction with body parts on scales ranging from 1 = extremely satisfied to 6 = extremely dissatisfied. This scale has shown internal consistency (α = .94), 3-week test-retest reliability (r = .90), and predictive validity.

Dieting was assessed using the Dutch Restrained Eating Scale (DRES), where participants indicate the frequency of dieting behaviors using scales ranging from 1 = never to 5 = always. The DRES has shown internal consistency (α = .95), 2-week test-retest reliability (r = .82), convergent validity with self-reported caloric intake, and predictive validity.

Negative affect was assessed with the sadness, guilt, and fear/anxiety subscales from the Positive Affect and Negative Affect Scale-Revised (PANAS-X). Participants reported the extent to which they had felt various negative emotional states on scales ranging from 1 = very slightly or not at all to 5 = extremely. This scale has shown internal consistency (α = .95), convergent validity, and predictive validity for bulimic symptom onset.

Bulimic symptoms were assessed using the diagnostic items from the Eating Disorder Examination (EDE), a semi-structured investigator-based interview. Items assessing the symptoms in the past month were summed to create an overall bulimic symptom composite for each assessment point. The adapted symptom composite showed internal consistency (α = .84 at pretest) and test-retest reliability (r = .89).

Analysis: Repeated measures ANOVA models were used to test whether participants in the intervention conditions improved over time across the five measures compared to waitlist control participants. In separate follow-up models, condition was a three-level between-subjects factor and time was a four-level within-subjects factor.

The study may not have adhered to the intent-to-treat principle. It is not clear whether it attempted to follow up with attriters.

Outcomes

Implementation Fidelity: No information was provided on implementation fidelity.

Baseline Equivalence: No significant differences between conditions were found for demographic or outcome variables.

Differential Attrition: The study reported that the attriters did not differ from completers on any demographic or outcome variables at pretest. Further, the study reported that attrition did not differ across conditions.

Posttest: Significant differences between dissonance intervention participants and waitlist control participants were found for two of the five outcomes: reductions in negative affect (p=.001) and bulimic symptoms (p=.015). It is unclear if the significant group-by-time coefficients produced statistically significant differences at all four time points.

The study reported r coefficients of .25 and .21 for the significant interactions but did not interpret the coefficients in terms of strength.

Study 3

Summary

The study evaluated the effectiveness of the dissonance intervention against a healthy weight placebo control condition. A total of 87 female students were randomized to the dissonance intervention (n=48) and the healthy weight placebo control (n=39). Data were collected at baseline, post-intervention (week 3), and at 4-week follow-up (week 7).

The study found significant differences between dissonance intervention participants and healthy weight placebo control participants for reductions in:

• Thin-ideal internalization.

Evaluation Methodology

Design: Using a randomized controlled design, researchers sought to evaluate the effectiveness of the dissonance intervention (Body Project) against a healthy weight placebo control condition. Using flyers and email messages, participants were recruited from two universities and one community college. A total of 87 female students were randomized to the dissonance intervention (n=48) and the healthy weight placebo control (n=39). Data were collected at baseline, post-intervention (week 3), and at 4-week follow-up (week 7). The attrition rate was 10%; 9 participants dropped out of the study and did not complete posttest or follow-up surveys.

Sample Characteristics: The sample of female participants had a modal age of 19 years and the following racial identities: 13% Asian/Pacific Islander, 5% African American, 17% Hispanic, 59% White, and 6% who specified other. Parental education ranged from high school graduate or less (2%) to advanced degree (13%).

Measures: Eight measures, several used in other studies, were gathered in this study:

Thin-ideal internalization was assessed using the Ideal-Body Stereotype Scale-Revised, which uses a response format ranging from 1 = strongly disagree to 5 = strongly agree. This scale has shown internal consistency (α = .91), test-retest reliability (r = .80), and predictive validity.

Body dissatisfaction was assessed with 8 items from of the Satisfaction and Dissatisfaction with Body Parts Scale that assesses body parts that are often of concern to females (e.g., stomach, thighs, and hips). Participants indicate their level of satisfaction with body parts on scales ranging from 1 = extremely satisfied to 6 = extremely dissatisfied. This scale has shown internal consistency (α = .94), 3-week test-retest reliability (r = .90), and predictive validity.

Dieting was assessed using the Dutch Restrained Eating Scale (DRES), where participants indicate the frequency of dieting behaviors using scales ranging from 1 = never to 5 = always. The DRES has shown internal consistency (α = .95).

Negative affect was assessed with the sadness, guilt, and fear/anxiety subscales from the Positive Affect and Negative Affect Scale-Revised (PANAS-X). Participants reported the extent to which they had felt various negative emotional states on scales ranging from 1 = very slightly or not at all to 5 = extremely. This scale has shown internal consistency (α = .95), convergent validity, and predictive validity for bulimic symptom onset.

Bulimic symptoms were assessed using the diagnostic items from the Eating Disorder Examination (EDE), a semi-structured investigator-based interview. Items assessing the symptoms in the past month were summed to create an overall bulimic symptom composite for each assessment point. The adapted symptom composite showed internal consistency (α = .84 at pretest).

Fat consumption was assessed using an adapted version of the Fat-Related Diet Habits Questionnaire which asks participants to rate how often they eat high-fat food using a 5-point scale ranging from never or almost never to five or more times a week. This scale had a Cronbach's alpha of .79 at baseline and a 3-week test-retest coefficient of .76.

Body Mass was measured using BMI calculated based on self-reports of height and weight.

Exercise behavior was assessed with an adapted form of the Past Year Leisure Physical Activity Scale, which asks participants to indicate how many days they did nine activities for at least 20 minutes at a time. Items use an 8-point response format ranging from 0 days to 7 days. The 3-week test-retest coefficient was .61.

Analysis: Repeated measures ANOVA models were used to test whether participants in the intervention conditions improved over time across the outcomes compared to placebo control participants. In the models, condition was a two-level between-subjects factor and time was a three-level within-subjects factor.

The study may not have adhered to the intent-to-treat principle. It is not clear whether it attempted to follow up with attriters.

Outcomes

Implementation Fidelity: All groups were audiotaped and reviewed by a research assistant blind to the condition. The study reported that all sessions covered the material contained in the manual for the intervention and placebo control conditions.

Baseline Equivalence: No significant differences between conditions were found for demographic or outcome variables.

Differential Attrition: The study reported that the attriters did not differ from completers on any demographic or outcome variables at pretest. Further, the study reported that attrition did not differ across conditions.

Posttest: The analysis referred to tests for fat consumption, BMI, and exercise as manipulation checks for the quality of the control condition, and the tests for the other outcomes as analysis of intervention effects. Significantly better results for dissonance intervention participants relative to healthy weight placebo control participants were found for one of the eight outcomes: reductions in thin-ideal internalization (p<.001). The authors attributed the limited effects in part to the unexpected improvement in the Healthy Weight control group.

Study 4

Summary

The study evaluated the effectiveness of the four-session version of the intervention delivered by high school nurses, teachers and counselors rather than by the investigators compared to a control condition. A total of 306 students were randomly assigned to the intervention (n=139) or control (n=167) conditions. Assessments were conducted at pretest, immediate posttest, 6-month follow-up, and at 1-, 2-, and 3-year follow-ups.

The study found positive effects for:

• Eating disorder symptoms at posttest, 6-month follow-up, 1-year, and 3-year follow-up
• Thin-ideal internalization at posttest
• Body dissatisfaction at posttest, 6-month follow-up, 1-year, and 3-year follow-up.

Evaluation Methodology

Design:

Recruitment/Sample size: Between April 2005 and November 2007, high school students were recruited using mailings and flyers from three school districts. The recruitment targeted female students between the ages of 14 and 19 with body image concerns. A total of 386 students (9.2% of the population) responded to the recruitment effort and were assessed for eligibility. Students were excluded if they did not have body image concerns (n=43) or met criteria for an eating disorder (n=2). An additional 35 students (9%) refused to participate in the study resulting in participation of 306 students (79.3%).

Study type/intervention: Researchers used a randomized controlled design in this evaluation. The 306 students were randomly assigned to the intervention (n=139) or control (n=167) conditions. Unlike previous evaluations, the intervention was delivered in four sessions by high school nurses, teachers and counselors rather than the investigators. A minimal intervention psychoeducational brochure was used as the control condition.

Assessment/Attrition: Assessments were conducted at pretest, immediate posttest, 6-month follow-up, and at 1-, 2-, and 3-year follow-ups by evaluators blind to the condition. Attrition rates at each assessment were as follows: 3% at posttest, 7% at 6-month follow-up, 8% at 1-year follow-up, 10% at 2-year follow-up and 16% at 3-year follow-up.

Sample Characteristics: The sample of girls had an average age of 15.7 years, mean BMI of 24.9, and was 2% Asian/Pacific Islander, 2% African American, 9% Hispanic, 81% Caucasian, and 6% who specified other or mixed racial heritage. Parental education was 17% high school graduate or less, 24% some college, 38% college graduate, and 21% advanced graduate/professional degree.

Measures: Six measures, several used in other studies, were gathered in this study.

Thin-ideal internalization was assessed using the Ideal-Body Stereotype Scale-Revised, which uses a response format ranging from 1 = strongly disagree to 5 = strongly agree. This scale has shown internal consistency (α = .91), test-retest reliability (r = .80), and predictive validity.

Body dissatisfaction was assessed with 8 items from the Satisfaction and Dissatisfaction with Body Parts Scale that assesses body parts that are often of concern to females (e.g., stomach, thighs, and hips). Participants indicate their level of satisfaction with body parts on scales ranging from 1 = extremely satisfied to 6 = extremely dissatisfied. This scale has shown internal consistency (α = .94), 3-week test-retest reliability (r = .90), and predictive validity.

Dieting was assessed using the Dutch Restrained Eating Scale (DRES), where participants indicate the frequency of dieting behaviors using scales ranging from 1 = never to 5 = always. The DRES has shown internal consistency (α = .95).

Depressive symptoms was assessed using the Center for Epidemiologic Studies Depression Scale, where participants select among four response options reflecting increasing levels of symptom severity (0=never/less than 1 day in past week to 3=most of the time/5-7 days in the past week). The CESD has shown internal consistency (α =.74 -.91) and reliability (test-retest r= .57-.59).

Eating pathology was examined using the Eating Disorder Diagnostic Interview, a semi-structured interview adapted from the Eating Disorder Examination which assesses DSM-IV eating disorder symptoms. Items assessing the symptoms in the past month were summed to form an overall eating disorder symptom composite for each assessment. The symptom composite showed internal consistency (α =.92) and 1-week test-retest reliability (r= .90).

Body Mass was measured using BMI calculated based on height and weight measures taken using stadiometers and digital scales.

Stice, Marti et al. (2011) reported gathering two additional measures:

Functional impairment was examined using items from the Social Adjustment Scale, which measures psychosocial functioning in the family, peer group, school, and work spheres.

Health care utilization during the previous 6 months was examined using two items: the number of times a person saw a doctor because of illness, injury, long-term health problems, or regular check-ups; and the number of times a person saw a psychologist, psychiatrist, or counselor/therapist because of mental health problems. Responses to these two items were averaged.

Analysis: Omnibus repeated measures analyses of variance were conducted to test whether changes in outcomes for the two conditions were significantly different from baseline to 1-year follow-up and from baseline to 3-year follow-up. This was followed by repeated measure ANOVA models for each of the assessment points. Reliable change score analysis was conducted to examine the clinical significance of changes in eating disorder symptoms.

The study complied with the intent-to-treat principle by including all available data from participants regardless of dosage received. It used all subjects in the analysis with full information maximum likelihood estimation.

OUTCOMES

Implementation Fidelity: A total of 67% of dissonance participants attended all four sessions, 18% attended three sessions, 5% attended two sessions, and 10% attended one session. Implementation fidelity ratings (10-point scale) had a mean of 6.9 and facilitator competence ratings (10-point scale) had a mean of 6.2.

Baseline Equivalence: The study reported no differences in demographic variables between conditions. Significant differences between conditions were found at baseline for two outcome variables: students in the control condition reported significantly higher thin-ideal internalization scores and significantly lower body dissatisfaction scores.

Differential Attrition: The study reported that the participants who did not complete the assessments through the 3-year follow-up did not differ from the 84% of participants retained in the trial on any demographic or outcome measures at pretest, and attrition did not differ significantly across conditions.

Posttest: Of the three outcomes examined in the study (depression, eating disorder symptoms, BMI), eating disorder symptoms were significantly improved at posttest and this effect was sustained at 6-month follow-up.

Program Effects on R&P Factors: Of the five risk/protective factors examined in the study, three were significantly improved at posttest. Thin-ideal internalization, body dissatisfaction and dieting were significantly improved for the intervention participants compared to the control participants at immediate posttest; however the effect was sustained at the 6-month follow-up only for body dissatisfaction.

Long-term Follow-up: Differences in improvement in eating disorder symptoms between conditions was significant at one-year follow-up and three-year follow-up, but not at 2-year follow-up.

Program Effects on R&P Factors: Differences in improvement in body dissatisfaction and dieting between conditions were significant at one-year follow-up. At two-year follow-up, only differences in body dissatisfaction were maintained, however these differences were not significant at 3-year follow-up.

Mediation Tests (Stice, Marti et al., 2011): The study examined data at pretest, posttest, 6-month follow-up, and 1-year follow-up to test for the indirect effects of the intervention on eating disorder symptoms via thin-ideal internalization and body dissatisfaction. The latent growth curve models used maximum likelihood estimation and bias-corrected bootstrap standard errors. Because both the mediators and the outcomes were measured as the change from baseline to the 12-month follow-up, the mediators coincided with rather than preceded the outcomes.

The tests revealed three significant indirect effects:

• Growth in thin-ideal internalization fully mediated the effect of the intervention on body dissatisfaction
• Growth in thin-ideal internalization partially mediated the effect of the intervention on eating disorder symptoms
• Growth in body dissatisfaction partially mediated the effect of the intervention on eating disorder symptoms

The indirect effects were all small in magnitude.

Study 5

Summary

The study evaluated the effectiveness of the four-session, enhanced version of the intervention delivered by clinicians to female college students compared to a control condition. A total of 408 students were randomly assigned to the intervention (n=203) or control (n=205) conditions. Assessments were conducted at pretest, immediate posttest, 1-year follow-up, 2-year follow-up, and 3-year follow-up.

The study found that the following outcomes were significantly improved from baseline to 1-year follow-up and 2- and 3-year follow-ups:

• Negative affect
• Eating disorder symptoms.

Evaluation Methodology

Design:

Recruitment/Sample size: Between October 2009 and October 2011, female students at seven universities in Oregon, Texas and Pennsylvania were recruited by facilitators using emails and posters. The recruitment targeted female students with body image concerns. A total of 682 women responded to the recruitment effort and were screened for eligibility. Respondents were excluded (10%) if they did not have body image concerns (n=33), were not affiliated with a university (n=29) or met criteria for an eating disorder (n=4). An additional 208 respondents (30%) refused to participate in the study, resulting in participation of 408 students (60%).

Study type/Randomization/Intervention: Researchers used a randomized controlled design in this evaluation. The 408 students were randomly assigned to the intervention (n=203) or control (n=205) conditions. The intervention differed from previous evaluations in the following three ways: facilitators were trained using more role-plays and received feedback to improve delivery; supervision of the facilitators was improved; and, participation, effort and accountability for taking an anti-thin-ideal perspective were more strongly emphasized in the curriculum. A minimal intervention psychoeducational brochure was used as the control condition.

Assessment/Attrition: Assessments were conducted at pretest, immediate 4-week posttest, and 1-year, 2-year, and 3-year follow-ups (although it is unclear if the follow-up periods lasted from baseline or the end of the 4-week program). Attrition rates increased over time: 1.0% at posttest, 4.7% at 1-year follow-up, 8% at the 2-year follow-up, and 15% at the 3-year follow-up.

Sample Characteristics: The sample of female participants had an average age of 21.6 years, mean BMI of 24.4, and was 58% European American, 17% Asian, 13% Hispanic, 7% African American, 4% American Indian/Alaska Native, and 1% Native Hawaiian/Pacific Islander. Parental education was 14% high school graduate or less, 25% some college, 32% college graduate, and 28% advanced graduate/professional degree.

Measures: Ten measures, several used in other studies, were gathered from subjects in this study. All assessments were conducted by evaluators blind to the condition.

Thin-ideal internalization was assessed using the Ideal-Body Stereotype Scale-Revised, which uses a response format ranging from 1 = strongly disagree to 5 = strongly agree. This scale has shown internal consistency (α = .91), test-retest reliability (r = .80), and predictive validity.

Body dissatisfaction was assessed with 8 items from of the Satisfaction and Dissatisfaction with Body Parts Scale that assesses body parts that are often of concern to females (e.g., stomach, thighs, and hips). Participants indicate their level of satisfaction with body parts on scales ranging from 1 = extremely satisfied to 6 = extremely dissatisfied. This scale has shown internal consistency (α = .94), 3-week test-retest reliability (r = .90), and predictive validity.

Dieting was assessed using the Dutch Restrained Eating Scale (DRES), where participants indicate the frequency of dieting behaviors using scales ranging from 1 = never to 5 = always. The DRES has shown internal consistency (α = .95).

Negative affect was assessed using the Beck Depression Inventory, which has shown internal consistency (α =.73 -.95) and reliability (test-retest r= .60-.90).

Eating disorder symptoms were examined using the Eating Disorder Diagnostic Interview, a semi-structured interview adapted from the Eating Disorder Examination which assesses DSM-IV eating disorder symptoms. Items assessing the symptoms in the past month were summed to form an overall eating disorder symptom composite for each assessment. The symptom composite showed internal consistency (α =.74) and 1-week test-retest reliability (ICC= .95).

Body Mass was measured using BMI calculated based on height and weight measures taken using stadiometers and digital scales.

Unhealthy weight transitions were also examined as an outcome variable and were coded if participants made unhealthy changes in BMI (i.e., from healthy weight to overweight or obese and from overweight to obese).

Psychosocial impairment was examined using items from the Social Adjustment Scale, which measures psychosocial functioning in the family, peer group, school, and work spheres.

Health care utilization was assessed using the Patterns of Help Seeking Behavior where participants were asked the number of hours they spoke with a doctor or nurse in the past month.

Mental health care utilization was assessed using the Patterns of Help Seeking Behavior where participants were asked the number of hours they spoke with a psychiatrist, a therapist, psychologist, or other counselor, or attended a support group in the past month.

Analysis: Multilevel, linear and logistic mixed effects models were used to analyze the data. An effect for site (level 3) was found for dieting and eating disorder symptoms only and therefore these two outcomes were modeled with three levels (time, person, and site) and the rest with two levels (time and person). Tests examined group-by-time interactions that controlled for baseline outcomes.

The study complied with the intent-to-treat principle by including all available data from participants regardless of dosage received. It used all subjects in the analysis with multiple imputation of missing data.

OUTCOMES

Implementation Fidelity: A total of 62% of dissonance participants attended all four sessions and 5% attended less than two sessions; participants attended an average of 3.4 sessions. Participants completed 87% of the assigned homework exercises. Implementation fidelity ratings (10-point scale) had a mean of 7.37 and facilitator competence ratings (10-point scale) had a mean of 7.12. Very low rates of inadequate fidelity (1.6%) or competence (2.3%) were found.

Baseline Equivalence: The study reported no differences in demographic or outcome variables between conditions.

Differential Attrition: Attrition was minimal at posttest (1%) and was 4.7% at 1-year follow-up. After the 2-year and 3-year follow-ups, when attrition was higher, the study reported that none of the pretest outcomes were significantly associated with missingness. However, the results did not test for differential attrition across conditions

Results: Of the four outcomes examined in the study (negative affect, eating disorder symptoms, BMI, unhealthy weight transitions), negative affect and eating disorder symptoms were significantly improved from baseline to 1-year follow-up (p's<.001, d = -.38 to -.40).

Program Effects on R&P Factors: Of the six risk/protective factors examined in the study, four were significantly improved at posttest. Thin-ideal internalization, body dissatisfaction, dieting and psychosocial functioning were significantly improved for the intervention participants compared to the control participants from baseline to 1-year follow-up. Effect sizes ranged from -.32 to -.54. The other two, health care utilization and mental health care utilization, were not significantly improved.

Long-term Follow-up: Similarly to the posttest, the study reported significant improvement in the treatment group as compared to the control group for negative affect (p<.001) and eating disorder symptoms (p<.001) through the 3-year follow-up. Effect sizes ranged from -.22 to -.39.

Long-term effects on R&P Factors: The study reported 7 risk and protective factors, of which 4 were significantly improved through the 3-year follow-up. Thin-ideal internalization, body dissatisfaction, dieting, and psychosocial impairment continued to show significant improvement among the treatment group as compared to the control group. Effect sizes ranged from -.07 to -.54.

Study 6

Summary

This study used a randomized controlled trial to examine 78 undergraduate women with high scores on a measure of thin idealization. It evaluated a three-session version of the dissonance-based Body Project program relative to an active control group. Assessments at one month after the end of the intervention measured negative affect and eating disorder symptoms.

The study found no significant condition differences.

Evaluation Methodology

Design:

Recruitment: The study prescreened 724 college women who were recruited through introductory psychology classes and fliers at the University of North Dakota. The 322 eligible participants scored high on measures of thin-ideal internalization (i.e., an "emphasis on appearance and socially defined ideals of attractiveness and weight") and were at risk of an eating disorder but did not report bulimic symptoms such as binging and purging that required clinical treatment. From the eligible subjects, the author randomly selected 78 women for the study.

Assignment: The 78 participants were randomly assigned to either the dissonance-based program or an active psychoeducational control group that focused on increasing knowledge of nutrition and understanding the negative consequences of eating disorders. A female clinical psychology doctoral student facilitated all sessions, but other psychology graduate students performed the interventions. 

Assessments/Attrition: Assessments occurred at baseline, the intervention end, and one month after the intervention end. Three of the initial 78 participants (4%) dropped out of the study during the intervention, and 49 (63%) did not complete the one-month follow-up. The attrition left a sample of 29 for the one-month follow-up.

Sample:

The participants averaged 19 years of age. The vast majority were Caucasian (95%), while the remaining 5% were Native American. Freshmen and sophomores made up 95% of the sample.

Measures:

The study examined seven student self-reported outcomes. The reported alpha and internal consistency coefficients came from other studies.

• Thin Ideal Internalization (IBSS-R) from the Ideal-Body Stereotype Scale-Revised (alpha = .89).
• Thin Ideal Internalization (SATAQ) from the Sociocultural Attitudes Towards Appearance Questionnaire-Revised (no alphas but showed high correlations with similar measures and with the subscales used to create the measure).
• Body Dissatisfaction (SDBPS) from the Satisfaction and Dissatisfaction with Body Parts Scale (alpha = .94).
• Body Dissatisfaction (BES-R) from the Body Esteem Scale-Revised (alphas ranged from .81-.94).
• Dieting Behavior (DRES) from the Dutch Restrained Eating Scale (alpha = .95).
• Negative Affect (PANAS-X) from the Positive Affect and Negative Affect Scale-Revised (alpha = .95).
• Eating Disorder (EDE-Q) from the Eating Disorder Examination-Questionnaire and focused on the main behavioral (binge eating and purging) aspects of bulimia (alpha = .84).

Analysis:

The study used repeated-measures analyses of variance with condition-by-time interaction terms that included the baseline outcomes. The immediate posttest (n = 75) and one-month follow-up (n = 29) were analyzed separately to maintain the larger posttest sample size.

Intent-to-Treat: The three subjects who did not complete all three sessions were dropped but made up only 4% of the sample.

Outcomes

Implementation Fidelity:

The study did not report quantitative figures. The author noted that "A standardized script was followed for each intervention to ensure that all components of the programs were delivered. All sessions were audio-taped, and used to ensure adherence to the scripts." However, she also suggested that implementation fidelity may have been a problem for the students delivering the program.

Baseline Equivalence:

Using the sample of 75 who completed the programs rather than the randomized sample of 78, the author reported no significant differences between the groups on the five demographic characteristics or six dependent variables (Appendix D, Tables 20 and 21) or on the six predictor variables (Appendix D, Table 22).

Differential Attrition:

Near equal numbers in the two conditions completed the one-month follow-up (15 in the dissonance group and 14 in the psychoeducation group). The three who dropped out during the intervention phase did not differ significantly from the others on the five demographic characteristics or the seven baseline dependent variables (Appendix D, Table 23). There were also no significant differences on the seven baseline dependent variables between the women who completed and did not complete the follow-up (Appendix D, Table 4).

Posttest:

In 14 tests (seven outcomes at posttest and seven at one-month follow-up), none showed significant condition differences (although one difference was marginally significant).

Long-Term:

Not examined.

Study 7

Summary

The study used a randomized controlled trial to examine 78 undergraduate women with high body dissatisfaction who were assigned to the full Body Project intervention group or a partial intervention group that used a subset of Body Project activities. It aimed to determine if a partial version of the program that included only one of the major components would work as well as the full program (implemented with three group-based sessions). There was no treatment-as-usual or assessment-only control group. The participants were assessed at one month follow-up on negative affect and eating disorder symptoms.

The study found found that the full intervention group showed significantly greater decreases than the partial intervention group on measures of:

• Thin-ideal internalization
• Negative affect.

Evaluation Methodology

Design:

Recruitment: The study screened 399 women, ages 18-44, who were taking undergraduate psychology classes at a large southeastern university. Those eligible were ages 18-30 and scored high on measures of thin-ideal internalization and body dissatisfaction but did not screen positive for an eating disorder. Of 125 eligible women, 78 agreed to enroll.

Assignment: The 78 participants were randomly assigned to the full Body Project intervention group or a partial intervention group that used a subset of Body Project activities focused on the dissonance component (i.e., counter-attitudinal advocacy exercises). Both treatment conditions consisted of three weekly group sessions. The study did not report the exact number assigned to each condition.

Assessments/Attrition: Other than the initial screening, which came one to four weeks before the intervention, there was no baseline assessment. Outcome measures were obtained after each of the three sessions and at one month after the program end. Of the 78 randomized participants, 51 (65%) completed all the sessions and the one-month follow-up. 

Sample:

The final sample consisted of women ages 18-28 years of age. Most (70.1%) were Caucasian, with 6.5% African American, 15.6% Hispanic, 5.2% Asian American, and 2.6% other ethnicities.

Measures:

The study examined seven student self-reported outcome measures:

• Thin-Ideal Internalization from the Sociocultural Attitudes Towards Appearance Scale Questionnaire-3 (SATAQ-3, alpha = .90)
• Body Dissatisfaction from the Eating Disorder Inventory Body Dissatisfaction subscale (EDI-BD, alpha = .82)
• Body Dissatisfaction from the Multidimensional Body-Self Relations Questionnaire Physical Appearance Evaluation subscale (MBSRQ-PAE, alpha = .81).
• Body Dissatisfaction from the Body Parts Satisfaction subscale (BPSS, alpha = .87).
• Dieting from the Dutch Eating Behavior Questionnaire Restraint subscale (DEBQ, alpha = .93).
• Bulimic Symptoms from the Eating Disorder Examination Questionnaire (EDE-Q, alpha = .84).
• Negative Affect from the Positive Affect and Negative Affect Scale-Revised (PANAS, alpha = .93).

Analysis:

The analysis used mixed models with repeated measures, full information maximum likelihood estimation, and a time-by-condition interaction term. However, lacking a baseline assessment, the models could not control for baseline outcomes and included time points beginning with the first post-session assessment. Follow-up comparisons of means used the modified Bonferroni procedure to adjust for multiple tests.

Intent-to-Treat: Of the 78 participants, 12 (15%) did not complete the sessions and, in possibly violating the intent-to-treat requirement, the study did not attempt to follow these dropouts and none of the dropouts provided follow-up data. It appears that the FIML analysis used only the 66 participants who completed the program.

Outcomes

Implementation Fidelity:

An analysis of videotapes from a subset of the sessions led the authors to state that "The results from these random checks suggest that the intervention was implemented with fidelity." No figures were presented, however.

Baseline Equivalence:

The only baseline measures came from the screening that took place one to four weeks before the intervention. In six tests, including two outcomes, none showed a significant difference between conditions. However, there were some significant differences in outcomes after the first session for two measures of body dissatisfaction, which the authors suggested (p. 8) may have reflected differences in assignment.

Differential Attrition:

The study noted the condition for only 20 of the 27 dropouts, and of those 20, 65% came from the full intervention group and 35% from the partial intervention group. Otherwise, the authors reported two separate differential attrition analyses, each based on measures obtained after the first session. The first compared intervention dropouts (n = 5) with intervention completers (n = 66) among those who attended the first session. One of seven outcomes showed a significant difference. The second compared those completing (n = 51) and not completing the one-month follow-up (n = 15). None of the outcome measures differed significantly across conditions.

Posttest:

The results in Table 3 indicated a significant time-by-condition coefficient for two of the seven outcomes: The full intervention group declined significantly more than the partial intervention group on thin-ideal internalization (eta-squared = .04) and negative affect (eta-squared = .05). A significant time effect for the other five outcomes indicated improvement in both conditions, which led the authors to conclude that the two program versions were equally effective and that counter-attitudinal advocacy is the crucial component of the full program.

Long-Term:

Not examined.

Study 8

Summary

The study used a randomized controlled trial to examine 113 undergraduate women with body dissatisfaction who were assigned to three conditions: the Body Project intervention, a yoga-based intervention, or a no-treatment control group. The basic outline of the dissonance-based program implemented in the study was the same as the Body Project but included six 45-minute sessions (4.5 hours total) instead of the three 60-minute sessions (3.0 hours total). The additional time was used to present concepts of fattism and self-objectification and to introduce feminist perspectives of the history of thinness and oppression of women. A posttest measuring eating disorders, depression, and anxiety came at the end of the last session.

The study found that the intervention group scored significantly lower at posttest than the control group on measures of:

• Eating disorders
• Anxiety
• Alexithymia (lack of social awareness and attachment).

Evaluation Methodology

Design:

Recruitment: The study recruited 252 female students at a large urban university from the psychology subject pool and responders to fliers advertising for women dissatisfied with their bodies. A total of 113 participants provided baseline data, agreed to participate, and attended at least one session. All but one of the participants in the analysis sample came from the psychology subject pool.

Assignment: Students were randomized to three conditions: the dissonance-based group (n = 30), a yoga intervention group (n = 33), or a control group (n = 30). The authors did not report the randomized sample sizes, and the condition sample sizes excluded dropouts. The dissonance-based intervention group was taught by a single psychology graduate student, which introduced a confound into the design.

Assessments/Attrition: The posttest assessment occurred at the end of the last session. Of the 113 participants, 20 (18%) dropped out before completing the posttest. The analysis sample thus included 93 participants.

Sample:

The sample was 55% white, 25% black, 4% Hispanic, 10% Asian, and 5% other ethnicities. The mean age of the participants was 19.6 years. Their mean body mass index (BMI) was 25.3, and 73% reported no or mild binge eating.

Measures:

The study examined 11 self-reported outcome measures, including scales for eating disorders, binge eating, state and trait anxiety, depression, body satisfaction, eating habits, alexithymia (lack of social awareness and attachment), and body image preoccupation. The measures came from pre-existing instruments and had acceptable reliability.

Analysis:

The analysis used regression models with controls for the baseline outcomes. Imputation of missing posttest data used baseline values of the outcomes.

Intent-to-Treat: The initial sample of 113 excluded an unspecified number of randomized participants who did not attend the first session of the two interventions. Then, of the 113 participants, only those who attended the last session had posttest outcomes, and others were not followed. An intent-to-treat analysis replaced missing posttest scores for those "who attended either the discussion or yoga group, but dropped out before completing the post-intervention assessment." The sample size for the intent-to-treat analysis was not reported. It was thus not clear if the authors did the same for the control group, and it appears that those dropping out before the first session were not included.

Outcomes

Implementation Fidelity:

Not examined.

Baseline Equivalence:

In 11 tests across the three conditions, one measure showed a significant difference. The dissonance group had higher scores on the binge eating scale than the other two conditions. No tests were presented for sociodemographic measures.

Differential Attrition:

Of the 20 dropouts, 17 came from the yoga condition and 3 from the dissonance condition, which suggests differential attrition across conditions. Separate tests for two different types of dropouts indicated no significant differences in attrition, however. First, comparisons of participants who completed the posttest with those who did not participate at all in the intervention found no significant differences "on any of the measures." Second, no significant differences were found on any of the measures between participants who completed both baseline and posttest measures and those who attended several intervention sessions but dropped out and did not complete the posttest.

Posttest:

Of 11 outcomes in Table 2, tests for the sample with imputed data showed four significant condition differences. The dissonance intervention group had significantly lower scores than the control group on two measures of eating disorders (eta-squared = .06 and .08), one measure of trait anxiety (eta-squared = .05), and one measure of alexithymia (eta-squared = .09)

Long-Term:

Not examined.

Study 9

Summary

The study used a randomized controlled trial to examine 107 undergraduate women with body image concerns who were assigned to four conditions: the Body Project, the internet version of Body Project, a video control group, and a brochure control group. The authors referred to an enhanced four-session version of the group-based, clinician-led program. The enhanced version added elements to increase dissonance induction, such as requiring more effort for in-session and home exercises. Measures through a two-year follow-up included negative affect and eating disorder symptoms.

The study found that the group-based intervention students reported significantly lower scores than the video and brochure control students on measures of

• Thin-ideal internalization at posttest
• Body dissatisfaction at posttest, one-year follow-up, and two-year follow-up
• Dieting frequency at posttest, one-year follow-up, and two-year follow-up
• Depressive affect at one-year follow-up
• Eating disorder symptoms at posttest and two-year follow-up
• Body mass index at two-year follow-up.

Evaluation Methodology

Design:

Recruitment: Participants included 107 young women recruited from a local university. The recruitment used e-mail and posters inviting female students with body image concerns to participate in a study evaluating various body acceptance interventions. The sole exclusion criterion was meeting diagnostic criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder. About 90% of those eligible consented to participate.

Assignment: The study randomly assigned participants to four conditions: 1) the Internet version of the Body Project (n = 19), 2) the group-based Body Project intervention (n = 39), 3) an educational video control condition (n = 29), or 4) an educational brochure control condition (n = 20). The extra participants assigned to the group-based intervention resulted from adding to an existing sample from a pilot study.

Assessments/Attrition: The posttest assessment occurred four to six weeks after the baseline assessment, and one- and two-year assessments followed. Only two participants (1.8%) did not complete the posttest (Stice, Rohde, Durant et al., 2012). The attrition rate was 5% at both the one- and two-year follow-ups (Stice et al., 2014).

Sample:

The sample was 7% American Indian/Alaska Native, 12% Asian, 1% African American, 2% Native Hawaiian/Pacific Islander, and 78% European American. Parental education, a proxy for socioeconomic status, was 17% high school graduate or less, 18% some college, 38% college graduate, and 27% graduate or professional degree. The average age was 21.6 years, and the average body mass index was 23.7.

Measures:

The self-reported measures came from interviews or surveys, with female assessors being blinded to the condition of the participants. The six outcome measures came from previously used instruments and had good reliabilities:

• Thin-ideal internalization (alpha = .78)
• Body dissatisfaction (alpha = .91)
• Dieting frequency (alpha = .92)
• Negative affect (alpha = .94)
• Eating disorder symptoms (alpha = .92)
• Body mass index

Analysis:

The analyses used repeated-measures multivariate analysis of variance models with condition-by-time interactions (condition was a two-level between-subjects factor, and time was a two-level within-subjects factor). For the posttest analysis, Stice, Rohde, Durant et al. (2012) imputed missing data using the last-observation-carried-forward approach. For the long-term analysis, Stice, Durant et al. (2014) used maximum likelihood estimation to adjust for missing data.

Intent-to-Treat: The analyses used all randomized participants.

Outcomes

Implementation Fidelity:

For the group intervention, 79% attended all four sessions, 18% attended three sessions, and 3% attended one session. Based on a review of taped sessions, implementation fidelity ratings (on a 1-10 scale) had a mean of 7.8, and only 3% of the components had less than "fair" adherence. Facilitator competence ratings (on a 1-10 scale) had a mean of 7.3, and 12% of items were rated as less than "good/average."

Baseline Equivalence:

Stice, Rohde, Durant et al. (2012) stated that "Participants in the four conditions did not significantly differ on demographic factors or outcome measures at pretest."  

Differential Attrition:

Attrition at all assessments did not exceed 5%.

Posttest:

Stice, Rohde, Durant et al. (2012). The multivariate analysis of variance model confirmed that changes in the outcomes over time differed across all four conditions. Post hoc tests for the five outcomes compared the group-based intervention condition to each of the other three conditions.

Relative to both the video and brochure control conditions, the group-based intervention condition scored significantly lower on all three risk and protective factors (thin-ideal internalization, body dissatisfaction, dieting) and one of two behavioral outcomes (eating disorder symptoms). The mean effect sizes were .72 and .54.

Relative to the Internet intervention condition, the group-based intervention condition did not differ significantly on any of the outcome measures.

Long-Term:

Stice, Durant et al. (2014). The multivariate analysis of variance model confirmed that changes in the outcomes over time differed across all four conditions. Post hoc tests for the six outcomes compared the group-based intervention condition to each of the other three conditions.

At one year, relative to both the video and brochure control conditions, the group-based intervention condition scored significantly lower on 5 of 12 tests (one for a behavioral outcome, depressive affect). The mean effect sizes were .44 and .42. Relative to the Internet intervention condition, the group-based intervention condition did not differ significantly on any of the outcome measures.

At two years, relative to both the video and brochure control conditions, the group-based intervention condition scored significantly lower on 6 of 12 tests (two for behavioral outcomes, eating disorder symptoms, and the body mass index). The mean effect sizes were .40 and .48. Relative to the Internet intervention condition, the group-based intervention condition did not differ significantly on any of the outcome measures.

Study 10

Summary

The study used a randomized controlled trial to examine 977 high school and college women with body image concerns who were assigned to the Body Project intervention or an assessment-only control group. Assessments occurred at baseline and posttest and included a measure of eating disorder symptoms.

This study combined data from three separate trials to increase the power to detect moderation effects and did not examine main effects. The combined data included both three-session and four-session versions of the program, but the authors noted that the two versions contained virtually identical content.

Evaluation Methodology

Design:

Recruitment: Participants came from three separate trials (Studies 1, 4, and 9) and included students from 12 high schools, eight universities, and three states (Texas, Oregon, and Pennsylvania). The recruitment used direct mailings, flyers, and leaflets that invited young women with body image concerns to participate in a study evaluating body acceptance interventions. The study sought to exclude students who met full threshold criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder, but numerous individuals still met the threshold and subthreshold for eating disorders at baseline. The authors referred to a final sample of 977.

Assignment: The study randomized the 977 female participants to either the intervention or an assessment-only control condition (other conditions used in some of the trials were discarded).

Assessments/Attrition: The study referred only to a pretest and posttest without any mention of attrition rates. However, reference to the separate trials indicated that the posttest followed four to six weeks after the intervention start and that attrition fell below 5%.

Sample:

The combined sample had a mean age of 18.6 years and included 6% Asian/Pacific Islanders, 4% African Americans, 10% Hispanics, 1% Native Americans, 74% Caucasians, and 6% who specified other or mixed racial heritage. Parental education, a proxy for socioeconomic status, consisted of 12% high school graduate or less, 19% some college, 36% college graduate, and 33% advanced graduate/professional degree.

Measures:

The self-reported measures came from surveys and blinded interviews and used previously developed and validated instruments. Good reliabilities were reported from these previous studies. Thin-ideal internalization, body dissatisfaction, and eating disorder symptoms served as the three outcome measures.

Analysis:

The analysis tested for moderation only. Regression models controlled for the baseline outcome measures, dummy variables for the trial, and dummy variables for the state of residence.

Intent-to-Treat: Although not explicitly stated, the study appeared to use all available data.

Outcomes

Implementation Fidelity:

Not examined.

Baseline Equivalence:

Of the three outcomes, one differed significantly across conditions. Participants in the intervention group had significantly higher eating disorder symptom scores than those in the control condition. The authors commented that the difference was small and reached significance because of the relatively large sample of 977. Still, all models controlled for baseline eating disorder symptoms.

Differential Attrition:

Attrition appeared to fall below 5%.

Posttest:

About one-third of the moderation tests were statistically significant. The intervention produced (a) significantly stronger reductions in thin-ideal internalization for participants with initial elevations in thin-ideal internalization and a threshold/subthreshold eating disorder at baseline, (b) significantly greater reductions in eating disorder symptoms for participants with an eating disorder at baseline, and (c) significantly stronger reductions in body dissatisfaction for late adolescence/young adulthood than mid-adolescent participants.

Long-Term:

Not examined.

Study 11

Summary

This article reported on two studies, one comparing the clinician-led intervention with the enhanced dissonance induction, a peer-led intervention with the enhanced dissonance induction, and a control group (treated as Study 11 here), and the other comparing a peer-led intervention and a control group (treated as Study 12 below).

The study used a randomized controlled trial to examine 171 undergraduate women with body image concerns who were assigned to the Body Project intervention, a peer-led Body Project intervention, or an educational brochure control group. Assessments occurred through a one-year follow-up and included measures of negative affect and eating disorder symptoms.

The study found that relative to the control group, the intervention group scored significantly lower at both posttest and long-term follow-up on:

• Thin-ideal internalization
• Body dissatisfaction
• Dieting
• Negative affect
• Eating disorder symptoms.

Evaluation Methodology

Design:

Recruitment: The sample included 171 women recruited at two universities in Texas from March 2010 to July 2011. The recruitment used e-mails and posters inviting female undergraduates with body image concerns to participate in a study designed to improve body acceptance. The few individuals who met the criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder at pretest were excluded.

Assignment: The study randomly assigned participants to the clinician-led intervention group (n = 55), a peer-led intervention group (n = 44), or an educational brochure control group (n = 72). Although the authors referred to using a random number table, they did not comment on the wide variation in condition sample sizes.

Assessments/Attrition: Participants completed assessments at pretest, posttest, and one-year follow-up. According to the CONSORT diagram, two participants (1%) did not complete the posttest and seven (4%) did not complete the one-year follow-up. However, the text stated that 11 (6%) did not complete the one-year follow-up.

Sample:

The sample averaged 20.9 years old and was 49% European American, 21% Asian, 20% Hispanic, 4% African American, 2% American Indian/Alaska Native, 2% mixed racial heritage, and 1% Native Hawaiian/Pacific Islander. Average parental education was 16% high school graduate or less, 15% some college, 38% college graduate, and 30% advanced graduate/professional degree.

Measures:

The measures came from student self-reports obtained via surveys and semi-structured interviews. Female assessors who were blinded to the condition of participants did the interviews. The commonly used measures in the study had, according to previous studies, good reliability. The five outcome measures included thin-ideal internalization, body dissatisfaction, dieting, negative affect, and eating disorder symptoms.

One problem occurred for the measure of eating disorder symptoms. A different, less expensive, and shorter measure based fully on self-reports was used for the peer-led intervention, while the more detailed measure based on an interview was used for the other two conditions.

Analysis:

Repeated measures ANOVA models treated condition as a two-level between-subjects factor and time as a two-level within-subjects factor. The time factor included the baseline outcomes in the models. Values for the missing data were imputed using a maximum-likelihood single imputation method.

Intent-to-Treat: With imputation of missing data, the analysis included all participants.

Outcomes

Implementation Fidelity:

Among participants assigned to the clinician-led intervention, 82% attended all four sessions, 15% attended three sessions, and 4% attended one session. Based on a scale from 1-10, the clinician-led intervention had a fidelity rating of 7.44 and a competence rating of 7.29. Among participants assigned to the peer-led intervention, 68% attended all four sessions, 23% attended three sessions, 2% attended two sessions, and 7% attended one session. Attendance was lower for the peer-led groups than the clinician-led groups. Based on a scale from 1-10, the peer-led intervention had a fidelity rating of 6.98 and a competence rating of 6.26. The competence rating was significantly lower than for the clinician-led group.

Baseline Equivalence:

Participants in the three groups did not significantly differ on demographic variables or pretest outcomes with one exception. The peer-led intervention participants had higher eating disorder symptom scores than the other two conditions, a difference likely due to the use of a different measure for the peer-led intervention condition. The authors noted that the focus of the repeated-measures models on change would control for the differences in the eating disorder outcome.

Differential Attrition:

Although all dropouts at the one-year follow-up came from the peer-led intervention condition, the attrition rate was only 4-6%.

Posttest:

An omnibus test confirmed that changes in the outcomes over time differed across all three conditions. At the posttest, the clinician-led intervention group scored significantly lower than the control group on all five outcomes and significantly lower than the peer-led intervention on three of five outcomes (body dissatisfaction, dieting, and negative affect). The peer-led intervention group scored significantly better than the control group on all five outcomes but significantly worse than the clinician-led intervention on three of five outcomes (body dissatisfaction, dieting, and negative affect). Significant effect sizes relative to the control group ranged from .43 to .92.

Long-Term:

At the long-term follow-up, the clinician-led intervention group scored significantly lower than the control group on all five outcomes and significantly lower than the peer-led intervention on one of five outcomes (eating disorder symptoms). The peer-led intervention group scored significantly better than the control group on one of five outcomes (dieting) but significantly worse than the clinician-led intervention on one of five outcomes (eating disorder symptoms, which must be discounted because of the different measures used for the two conditions).

Study 12

Summary

The study came from the same article as in Study 11 but presented results for a different sample and design. It examined the peer-led version of the program relative to a control group and did not include a condition for the clinician-led version of the program. The authors noted as well that the training and script used in this study improved on that of Study 1.

A randomized controlled trial was used to examine 148 undergraduate women with body image concerns who were assigned to the peer-led Body Project intervention or a waitlist control group. A posttest assessment at four weeks included measures of negative affect and eating disorder symptoms.

The study found that the peer-led intervention group scored significantly lower at posttest than the control group on:

• Thin-ideal internalization
• Body dissatisfaction
• Dieting
• Negative affect
• Eating disorder symptoms.

Evaluation Methodology

Design:

Recruitment: The sample included 148 women recruited at two universities in Texas from October 2011 to October 2012. The recruitment used e-mails and posters inviting female undergraduates with body image concerns to participate in a study designed to improve body acceptance. The few individuals who met the criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder at pretest were excluded.

Assignment: Participants were randomly assigned (via random number table) to a peer-led intervention condition (n = 80) with groups led by 17 advanced undergraduates or a waitlist control group (n = 68).

Assessments/Attrition: Participants completed surveys at pretest and four-week posttest, with 94% of participants completing the posttest.

Sample:

The sample was 41% European American, 28% Asian, 17% Hispanic, 7% African American, and 7% mixed racial heritage. Average parental education was 25% high school graduate or less, 14% some college, 32% college graduate, and 29% advanced graduate/professional degree.

Measures:

The measures came from student self-reports obtained via surveys. The commonly used measures in the study had, according to previous studies, good reliability. The five outcome measures included thin-ideal internalization, body dissatisfaction, dieting, negative affect, and eating disorder symptoms.

Analysis: The analysis used repeated measures ANOVA models. Condition-by-time interactions tested for differential changes in outcomes from pretest to posttest between the two conditions. Values for the missing data were imputed using an ML-based single imputation method.

Intent-to-Treat: The imputed data allowed the inclusion of all randomized participants in the analysis.

Outcomes

Implementation Fidelity:

Among participants assigned to peer-led groups, 61% attended all four sessions, 15% attended no sessions, and 24% attended one to three sessions. The average fidelity and competence ratings for peer leaders were 6.62 and 5.92, respectively, on a scale from 1-10.

Baseline Equivalence:

The authors stated that participants in the two conditions did not differ significantly on demographics or any pretest measure.

Differential Attrition:

Not examined, though attrition was only 7%.

Posttest:

The condition-by-time interactions indicated that the peer-led intervention group reported significantly greater reductions than the control group for all five outcomes: thin-ideal internalization, body dissatisfaction, dieting, negative affect, and eating disorder symptoms. Effect sizes ranged from medium to large (d =.40-.91).

Long-Term:

Not examined.

Study 13

Summary

This article contained two separate studies, the second of which examined a peer-led version of the program (treated below as Study 14). This first study (treated here as Study 13) examined the clinician-led version of the program. The sample overlaps with that of Study 5 but added another control group with 29 subjects.

A randomized controlled trial was used to examine 426 undergraduate women with body image concerns who were assigned to the Body Project intervention or to educational brochure or video control groups. Assessments occurred at baseline, posttest, and one-year follow-up and focused on differences in program effects across race and ethnic groups.

The study did not examine main effects.

Evaluation Methodology

Design:

Recruitment: The study recruited participants from October 2009 to October 2011 using e-mails and posters inviting women with body image concerns to participate in a trial designed to improve body acceptance. Participants included 437 young women at seven universities in Oregon, Pennsylvania, and Texas. Research staff excluded students who met the criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder at pretest.

Assignment: The initial randomization assigned the 437 students to the Body Project intervention (n = 203), an educational brochure control condition (n = 205), or a video control condition (n = 29). However, the number of Native American participants (n = 11) was too small to include in analyses focused on differences in program benefits across ethnic groups. Excluding those 11 left 426 participants in four groups: African American, Asian American/Pacific Islander, European American, and Hispanic. Then, the two control groups were combined because they did not differ significantly on the outcomes. After the exclusion and combination, the intervention group and control group had 197 and 229 participants, respectively

Assessments/Attrition: Assessments occurred at baseline, the four-week posttest, and a one-year follow-up. According to the CONSORT diagram, 1% were lost at posttest and 5% were lost at the one-year follow-up (but the text stated that 7% were missing at the one-year follow-up).

Sample:

The sample students averaged 21.6 years of age and were 17% Asian, 6% African American, 65% European American, and 12% Hispanic. Among the parents, 14% were high school graduates or less, 24% had some college, 34% were college graduates, and 29% had advanced graduate/professional degrees.

Measures:

The outcome measures came from student self-reports and interviews by female assessors who were blinded to condition.

• Thin-ideal internalization (alpha = .72)
• Body dissatisfaction (alpha = .82)
• Dieting . (alpha = .92)
• Negative affect (alpha = .91)
• Eating disorder symptoms (alpha = .69)

Analysis:

All models were fit using linear mixed-effects models with repeated measures for the multiple time points (including baseline). Multiple imputation was used to replace missing values.

Intent-to-Treat: The study included all randomized participants except Native Americans.

Outcomes

Implementation Fidelity:

The average session attendance rate ranged from 80-87% across the four ethnic groups.

Baseline Equivalence:

Intervention and control participants did not significantly differ on demographics or baseline outcomes with one exception: The intervention students reported significantly more eating disorder symptoms.

Differential Attrition:

Not tested, but attrition was low.

Posttest and Long-Term:

The analyses examined changes that encompassed both the posttest and long-term assessment but did not estimate main effects. It focused solely on the differences in program effectiveness across race and ethnic groups. There was no significant moderation by race and ethnicity for any of the five outcomes, indicating that the program worked similarly well for minorities as for European Americans. Significance tests were not presented for each group separately, but effect sizes ranged from -.21 to -.98 across the groups and outcomes.

Study 14

Summary

This article contained two separate studies, the second of which examined the peer-led version of the program and is described here. The first study of the clinician-led version is described in Study 13.

A randomized controlled trial was used to examine 192 undergraduate women with body image concerns who were assigned to the peer-based intervention or a waitlist control group. Assessments occurred at baseline and four-week posttest. The analysis examined differences in program effects across racial and ethnic groups.

The study did not examine main effects.

Evaluation Methodology

Design:

Recruitment: The study recruited 192 participants from two universities in Texas from January 2011 to October 2012. The recruitment used e-mails and posters inviting women with body image concerns to participate in a trial designed to improve body acceptance.

Assignment: Participants were randomized to a peer-led Body Project intervention group (n = 124) or a waitlist control group (n = 68). Of the 124 intervention participants, 44 were from Study 1 and 80 from Study 2.

Assessments/Attrition: Data were complete at pretest except for eating disorder symptoms, which was missing for 5% of cases. Another 5% of the participants did not complete the posttest assessment.

Sample:

The sample included 30% Asian Americans, 13% African Americans, 43% Euro Americans, and 22% Hispanics. Average parental education was 18% high school graduate or less, 19% some college, 28% college graduate, and 35% advanced graduate/professional degree.

Measures:

The outcome measures came from student self-report surveys.

• Thin-ideal internalization (alpha = .73)
• Body dissatisfaction (alpha = .82)
• Dieting (alpha = .88)
• Negative affect (alpha = .90)
• Eating disorder symptoms (alpha = .81)

Analysis:

The analyses used linear mixed-effects models with repeated measures for the two time points (baseline and posttest). Multiple imputation was used to replace missing values.

Intent-to-Treat: The study included all randomized participants except Native Americans (n = 3), who were excluded because their numbers were too few for meaningful moderation tests.

Outcomes

Implementation Fidelity:

The average attendance rate was .70 for Asian Americans, .75 for African Americans, .74 for European Americans, and .79 for Hispanics.

Baseline Equivalence:

The intervention condition did not significantly differ from the control condition on demographics or pretest outcomes except for age, for which intervention participants exhibited a significantly lower average. Age was thus included as a covariate in models testing for intervention effects.

Differential Attrition:

Not examined, but attrition for most outcomes was only 6%.

Posttest:

The analyses did not examine main effects and focused solely on the differences in program effectiveness across racial and ethnic groups. There was no significant moderation by race and ethnicity for any of the outcomes, indicating that the peer-led program worked similarly well for minorities as for European Americans.

Long-Term:

Not examined.

Study 15

Summary

The study used a randomized controlled trial to examine 680 women from three universities in Oregon and Texas. Eligible participants were randomly assigned to four conditions: 1) a clinician-led Body Project intervention, 2) a peer-led Body Project intervention, 3) the eBody Project internet intervention, or 4) an educational video control. The participants were followed through a four-year follow-up.

At both posttest and long-term follow-up, the study found that, relative to the controls, participants in both the clinician-led and peer-led versions of the intervention showed significantly greater reductions in:

• Eating disorder risk factors
• Eating disorder symptoms.

Evaluation Methodology

Design:

Recruitment: The sample included 680 women (undergraduate students, graduate students, and university staff) recruited from three universities in Oregon and Texas. The recruitment occurred between March 2013 and April 2015 and used e-mail messages and posters. The women had to report body dissatisfaction for study inclusion, but those who endorsed diagnostic criteria for current anorexia nervosa, bulimia nervosa, or binge eating disorder, were not affiliated with the university, or were already participating in another trial evaluating an intervention were excluded.

Assignment: Eligible participants were randomly assigned to four conditions: 1) a clinician-led Body Project group (n = 173), 2) a peer-led Body Project group (n = 162), 3) the eBody Project internet group (n = 184), or 4) an educational video control group (n = 161) using a random number table. Undergraduate students, graduate students, and staff participants were assigned to the same groups.

Assessments/Attrition: Assessments occurred at baseline, posttest, and six-month follow-up in Stice et al. (2017) and Shaw et al. (2020) and then at one-, two-, three-, and four-year follow-ups in Stice et al. (2020). Attrition was 9% at the posttest, 11% at the six-month follow-up, 10% at the one-year follow-up, 12% at the two-year follow-up, 14% at the three-year follow-up, and 19% at the four-year follow-up.

Sample: The sample was 60% White, 17% Latina, 14% Asian, 5% African American, 3% Native American, and 1% Pacific Islander. Average parental education was 13% high school graduate, 16% some college, 34% college graduate, and 38% graduate or professional degree. The women averaged 22.2 years of age. Reports indicated that 78% of the respondents identified as straight/heterosexual, with the other 22% identifying as either sexual minority (gay/lesbian, bisexual, queer) or heterosexual and another category.

Measures:

The outcome measures came from self-reports and diagnostic interviews done by trained female assessors who were masked to condition. Data on eating disorder symptoms were obtained for each month for the period from baseline to follow-up.

• Thin-ideal internalization (alpha = .75)
• Body dissatisfaction (alpha = .86)
• Dieting (alpha = .91)
• Negative affect (alpha = .94)
• Psychological functioning (alpha = .74)
• Eating disorder symptoms over the past month (alpha = .70)
• Diagnosed eating disorder onset (the first month meeting threshold or subthreshold criteria).

Analysis: Stice et al. (2017) used mixed-effects growth models that controlled for baseline outcomes. The condition main effect showed differences at six months, and a condition-by-time interaction tested whether the reductions in outcomes were significantly stronger at posttest versus six-month follow-up or vice versa. Missing data were imputed. For the eating disorder onset outcome, the analysis used Cox proportional hazard models. Because hazard models accommodate right censoring, missing onset data were not imputed. Models indicated no significant variability attributable to groups in partially clustered models (i.e., the partial clustering refers to having two conditions with groups and two without groups). However, significant variability attributable to site existed for eating disorder symptoms. Thus, all models but one were fit as two-level models in which assessment points were nested within individuals. The exception of eating disorder symptoms contained a Level 3 random effect for site.

Shaw et al. (2020) also examined linear growth models through the six-month follow-up but focused on moderation by sexual orientation and adjusted for multiple tests with the Benjamini-Hochberg correction.

Stice et al. (2020) examined data through the four-year follow-up using mixed-effects growth models. Tests indicated no significant variability attributable to groups or sites. The analysis thus used two-level models in which assessments were nested within individuals. The models adjusted for baseline outcomes and included a condition-by-time interaction to capture intervention effects. Full information maximum likelihood estimation was used with imputation of missing data based on the pretest outcomes, condition, and demographic factors (age, race, and parent education). For the eating disorder onset measure, the analysis used Cox proportional hazard models but did not impute missing data because the models accommodate right censoring.

Missing Data Method: Missing data were handled using full information maximum likelihood (FIML)  and multiple imputation. Specifically, for the 6-month follow-up, Stice et al. (2017) stated: "Missing data were imputed using IVEware (Raghunathan, Solenberger, & Van Hoewyk, 2016) using baseline levels of the outcomes and demographic factors, with imputed data in 20 data sets analyzed separately; model parameters and standard errors were combined following Rubin (1987)" (p. 887). Meanwhile, for the 4-year follow-up, Stice et al. (2020) stated: "We used full information maximum likelihood (ML) estimation to impute missing data…Missing data were imputed using PROC MI using baseline levels of the outcomes, condition, and demographic factors (age, race, and parent education). Based on the recommendations, we imputed 50 data sets" (pp. 485-486).

Intent-to-Treat: The authors followed an intent-to-treat approach, with analyses including all available participants.

Outcomes

Implementation Fidelity: Among participants in clinician-led groups, 47% attended all four sessions, 28% attended three sessions, 6% attended two sessions, 5% attended one session, and 14% attend no sessions. Ratings of the clinician-led groups averaged 75.2 for fidelity and 71.7 for adherence (on a scale from 10 to 100). Among participants in peer-led groups, 45% attended all four sessions, 26% attended three sessions, 9% attended two sessions, 6% attended one session, and 14% attend no sessions. Ratings of the peer-led groups averaged 70.2 for fidelity and 65.1 for competence (on a scale from 10 to 100).

Baseline Equivalence: Stice et al. (2017, 2020) reported that the participants in the four conditions did not significantly differ on race, ethnicity, age, year in school, parental education, or pretest measures of the outcomes. Shaw et al. (2020) added that sexual orientation was not significantly related to condition (p = .08).

Differential Attrition: Stice et al. (2017, 2020) reported that attrition was not associated with condition but was associated with elevated pretest eating disorder symptoms. Shaw et al. (2020) added that missing sexual orientation data was not significantly related to condition. In addition, completion figures listed in the CONSORT Diagram for 6-month follow-up (Stice et al., 2017) show that the attrition meets the What Works Clearinghouse differential attrition standard under the cautious threshold (e.g., at the 6-month follow-up, the attrition rate overall was 11% and the difference in attrition rates between conditions was 0.2%).

For the long-term follow-up, Stice et al. (2020) stated that "Attrition was not associated with condition, demographics, or baseline values of outcome variables, with the exception that participants who dropped out had higher baseline eating disorder symptoms (d = .49)." In addition, completion figures listed in the CONSORT Diagram for 4-year follow-up (Stice et al., 2020) show that the attrition meets the What Works Clearinghouse differential attrition standard under the cautious threshold (e.g., at the 4-year follow-up, the attrition rate overall was 18.6% and the difference in attrition rates between conditions was 2.0%).

In addition, the use of FIML and multiple imputation helps counter concerns over attrition bias.

Posttest: Stice et al. (2017): In comparison to the video control group, the clinician-led intervention group scored significantly lower on all four outcomes (thin-ideal internalization, body dissatisfaction, negative affect, and eating disorder symptoms) at both posttest and the six-month follow-up. Effect sizes ranged from d= -.38 to -.70 at posttest and d = -.20 to -.43 at follow-up. In comparison to the video control group, the peer-led intervention group scored significantly lower on all four outcomes (thin-ideal internalization, body dissatisfaction, negative affect, and eating disorder symptoms) at posttest and on three of four outcomes (all but negative affect) at the six-month follow-up. Effect sizes ranged from -.35 to -.71 at posttest and -.35 to -.52 at follow-up. The peer-led and clinician-led versions did not differ significantly from one another on any of the outcomes at posttest or follow-up.

In comparison to the internet version, the clinician-led intervention group scored significantly lower on three outcomes at posttest (thin-ideal internalization, body dissatisfaction, negative affect) and one outcome at follow-up (thin-ideal internalization). In comparison to the peer-led version, the clinician-led intervention showed no significant differences.

The survival analysis of eating disorder onset did not show significant differences between the clinician-led intervention, the peer-led intervention, and the other two conditions.

Shaw et al. (2020): The analysis found no significant moderation of the outcome levels or outcome changes by sexual orientation (coded as a dichotomy of heterosexual or sexual minority). The results indicated that the clinician-led and peer-led versions worked as well for sexual minority women as for heterosexual women.

Long-Term: Stice et al. (2020) found that the growth rates for the clinician-led intervention showed significantly greater reductions than the control group in four of five outcomes: body dissatisfaction, negative affect, dieting, and eating disorder symptoms (Table 3). Adjusted means in Table 4 showed significantly lower values for the clinician-led intervention group than the control group on five of five outcomes at one year and two years, and one of five outcomes at three years and four years (thin-ideal internalization). Effect sizes (d) ranged from -.16 to -.55 for the significant outcomes. Cox proportional hazard models found no significant difference in the rate of onset between the clinician-led intervention and the control group.

Stice et al. (2020) also found that the growth rates for the peer-led intervention did not differ significantly from the control group or the clinician-led intervention group for any of the five outcomes (Table 3). Adjusted means in Table 4 showed significantly lower values for the peer-led intervention group than the control group on five of five outcomes at one year, four of five outcomes at two years (all but negative affect), two of five outcomes at three years (thin-ideal internalization and body dissatisfaction), and one of five outcomes at four years (thin-ideal internalization). Effect sizes ranged from -.19 to -.59 for the significant outcomes. The adjusted means did not differ for the peer-led and clinician-led groups on any of the outcomes or time points. Cox proportional hazard models showed a significantly lower hazard rate for the onset of eating disorder in the peer-led group than in the control group, but the hazard rate did not differ significantly between the peer-led and clinician-led intervention groups.

Study 16

Summary

The study used a randomized controlled trial to examine 333 undergraduate women with body image concerns who were assigned to the Body Project intervention, an internet Body Project intervention, or an assessment-only control group. Assessments occurred at baseline and posttest and included a measure of eating disorder symptoms.

The study found that, relative to the control group, the intervention group scored significantly higher at posttest on:

• Body esteem or satisfaction subscales (sexual attractiveness, weight concern, and physical condition).

Evaluation Methodology

Design:

Recruitment: The study recruited undergraduate women (ages 18-25) from the psychology participant pool at a large urban university. The study was advertised as a program for women with weight and/or shape concerns but excluded those meeting diagnostic criteria for anorexia nervosa, bulimia nervosa, or binge eating disorder. Although this program was portrayed as a body image program, participants were not required to endorse body image concerns. After exclusions, 343 participants remained.

Assignment: The authors reported randomly assigning 333 participants to the face-to-face Body Project intervention (n = 107), an online version of the intervention (n = 112), or an assessment-only control group (n = 114). The sample sizes exclude 10 randomized participants who completed the posttest but were dropped because they did not attend any sessions.

Assessments/Attrition: Assessments came at pretest and posttest. The CONSORT diagram lists a posttest sample of 199 (58% of the full randomized sample of 343).

Sample:

The final sample was 44% White, 31% Black/African American, 5% Hispanic/Latina, 8% Asian/Asian American, 9% Multiracial, and 3% "Other." Most were first-year students (67%). The students averaged 18.8 years of age.

Measures:

The five outcome measures came from self-reports:

• Eating Disorder Diagnostic Scale (alpha = .79)
• Ideal-Body Stereotype Scale-Revised (alpha = .78)
• Body Esteem Scale with three body satisfaction subscales for sexual attractiveness, weight concern, and physical condition (alphas ranged from .84-.88).

Analysis:

A repeated measures multivariate analysis of variance treated time of assessment (including baseline) as the within-subjects variable and group assignment as the between-subjects variable. For those who received some dose of the program but did not complete the posttest (n = 40), baseline scores were substituted for posttest scores.

Intent-to-Treat: The analysis excluded 10 participants (3%) in either of the intervention groups who completed the posttest but did not attend any sessions. In addition, 26.6% who did not complete the posttest and did not attend any sessions were not included in the imputed sample.

Outcomes

Implementation Fidelity:

Among the face-to-face intervention group, 41.5% attended all sessions, 28% completed all homework assignments. Leaders covered 100% of the main exercises and discussion points included in the manuals.

Baseline Equivalence:

The conditions did not differ significantly with respect to race/ethnicity, year in school, age, BMI, or four of the five outcome measures at baseline. They differed on the Body Esteem Scale Physical Condition, with participants in the control group scoring higher than the other two groups.

Differential Attrition:

Attrition rates did not differ across conditions. Completers and non-completers had significantly different baseline scores on the ideal-body stereotype but not for any of the other four outcomes. The tests did not appear to examine sociodemographic measures.

Posttest:

The combination of outcome variables was significantly influenced by the interaction of time and group in the multivariate analysis of variance. Tests for the five individual outcomes showed three significant differences across conditions. Relative to the assessment only control group, the face-to-face intervention students reported significantly higher scores for the three body-satisfaction subscales (eta-squared = .03-.07) but not for the eating disorder diagnostic scale. The face-to-face intervention did not differ significantly from the online intervention on any of the outcomes.

Long-Term:

Not examined.

Study 17

Summary

The study used a randomized controlled trial to examine 66 young women (ages 13-17) recruited from two local medical clinics. The participants were randomly assigned in groups of eight to the intervention (which used the enhanced dissonance induction version of the program) or control conditions. Assessments occurred at baseline, posttest, and three-month follow-up and included measures of negative affect and eating disorder symptoms.

The study found at both posttest and three-month follow-up that the intervention group reported significantly greater decreases than the control group on:

• Thin-ideal internalization
• Pressure to be thin
• Dieting
• Eating disorder symptoms.

Evaluation Methodology

Design:

Recruitment: The sample included 66 young women recruited from February 2012 to June 2014 at either a small family medicine clinic or a large pediatric clinic. At the smaller family medicine site, all participants were recruited from a direct mailing, whereas at the larger pediatric site, participants were recruited from a medical provider referral or seeing a flier in the exam room. Eligible participants reported body image dissatisfaction, identified as female, were between 13 and 17 years of age, and spoke English or Spanish. Participants who endorsed eating disorder symptoms were not excluded from the trial.

Assignment: The study randomly assigned participants within each clinic to a control group that watched a film called Dying to Be Thin (n = 23) or the Body Project intervention delivered in a medical setting (n = 43). The randomization occurred in blocks of eight with the first eight participants going into the intervention condition, the next eight going into the control condition, and so on. One clinic ran two intervention groups and only one control group, leading to unequal condition sizes. Although each participant had a 50:50 chance of ending up in each condition, randomization was not completely independent for each participant. Therefore, the study referred to itself as a quasi-randomized trial.

Assessments/Attrition: Participants completed assessments at pretest, posttest, and three months after posttest. Of the 66 participants, 77% completed the posttest assessment, and 71% completed the 3-month follow-up assessment.

Sample:

For the 80% of participants who reported their race, the sample was 2% African American, 2% Asian, 15% Hispanic, 72% European American, and 9% "other." About 20% of the sample had a full or partial eating disorder syndrome at baseline.

Measures:

The six outcome measures came from self-reports and showed good reliability:

• Thin-ideal internalization (alpha = .91)
• Perceived pressure to be thin (alpha = .93)
• Body dissatisfaction (alpha = .95)
• Dieting (alpha = .95)
• Negative affect (alpha = .97)
• Eating disorder symptoms (alpha = .86)

Analysis:

The analysis used mixed-effects linear growth models with baseline outcomes and site included as covariates. The coefficients for condition indicated posttest condition mean differences, while the condition-by-time interaction coefficients indicated condition differences in the growth rates from the posttest to the 3-month follow-up assessment. Random effects for the intercept and time parameters accommodated the clustering of participants within the intervention groups (with controls treated as a group of one). Multiple imputation was used to replace missing values. Effect sizes were estimated from t values.

Intent-to-Treat: The multiple imputation allowed the use of all assigned participants.

Outcomes

Implementation Fidelity:

The interventionists received average scores in the 70s and 80s for both program adherence and competence (on a scale with "10" indicating no adherence or competence and "100" indicating perfect adherence or competence).

Baseline Equivalence:

Tests found a significant condition difference at baseline only for recruitment site, and no significant differences on baseline measures existed across the two sites.

Differential Attrition:

The number of completed assessments was not related to study condition, but participants who did not complete all assessments had higher baseline dieting, body dissatisfaction, negative affect, and eating disorder symptoms than those who provided complete data.

Posttest:

At posttest, the conditions differed significantly on four of six outcomes. The intervention group showed greater decreases than the control group for thin-ideal internalization (d = -.71), pressure to be thin (d = -.73), dieting (d = -.82), and eating disorder symptoms (d = -.50). From posttest to the three-month follow-up, the rate of change did not differ significantly across conditions for any of the outcomes, meaning that the posttest differences were maintained through the follow-up.

Study 18

Summary

The study used a randomized controlled trial to examine 47 women (ages 15-33) with eating disorder symptoms or a diagnosis. Participants were randomly assigned to the dissonance-based intervention or an assessment-only control group. For the intervention group, the study used the same structure and format as the dissonance-based Body Project but made some adaptations in content to address objectification and maladaptive social comparison more explicitly. Assessments at baseline, posttest, and two-month follow-up examined eating disorder symptoms and cardiac function.

The study found that the intervention group, relative to the control group, showed significantly greater improvement over time in:

• Self-reported eating disorder symptoms
• Indicators of cardiac risk.

Evaluation Methodology

Design:

Recruitment: Participants were recruited via ads, fliers, and social media from two Midwestern cities and five surrounding suburban and rural communities. Those eligible included nonpregnant women ages 14 to 35 who met diagnostic criteria for an eating disorder or were symptomatic for disordered eating. Of the 82 women responding to the recruitment, 74 completed the online screening, 66 met inclusion criteria, 52 enrolled in the preliminary trial, and 47 participated in the trial.

Assignment: Participants were randomly assigned to the dissonance-based intervention or to an assessment-only control condition, although it was not clear if the 52 enrollees or 47 participants were randomized.

Assessments/Attrition: Assessments occurred at baseline, posttest, and two-month follow-up. Based on an assumed 52 randomized participants, the attrition rates were 6% at baseline, 31% at posttest, and 29% at the two-month follow-up.

Sample:

Participants' ages ranged from 15 to 33 years. Figures on ethnicity were not available. About 47% of the sample had clinical levels and 53% had sub-clinical levels of disordered eating.

Measures:

A measure of eating disorder symptoms came from the self-reported Eating Disorder Examination-Questionnaire. Other studies have shown that the measure has good reliability and validity.

Additional measures came from laboratory assessments of height, weight, and electrocardiogram signals of cardiovascular function.

• Body mass index
• QT interval (high score indicates cardiac risk)
• Mean R wave amplitude (low score indicates cardiac risk)
• Sympathetic tone (low score indicates cardiac risk)
• Vagal tone (high score indicates cardiac risk)

Analysis:

The study used mixed factorial multivariate analysis of variance (MANOVA) and analysis of variance models with condition-by-time interaction terms across baseline, posttest, and 2-month follow-up time points. The interaction terms included the baseline outcomes.

Intent-to-Treat: Mean replacement was used for all missing data. It appears that the mean replacement allowed inclusion of all 47 participants. However, the study referred to 52 enrollees, suggesting that the five who did not participate were dropped and that the analysis might not meet the intent-to-treat criterion.

Outcomes

Implementation Fidelity:

Not examined.

Baseline Equivalence:

The study did not discuss baseline equivalence but showed test results for baseline differences for five of the six outcomes (see Table 3, BD vs. BC, or baseline dissonance versus baseline control). The five tests in the table show four significant condition differences at baseline for cardiac functioning measures.

Differential Attrition:

The study reported the numbers of participants in each condition who were missing, which appear to be similar for intervention and control conditions (2 and 1 at baseline, 8 and 8 at posttest, and 7 and 8 at follow-up). No other information was provided.

Posttest:

Based on the time-by-group interaction terms in Table 2, significant differences in the changes over time emerged for five of six outcomes. Consistent with the hypotheses, eating disorder symptoms decreased, the body-mass index increased, the mean R wave amplitude increased, the sympathetic tone decreased, and the QT interval length decreased more for the intervention group than the control group (partial eta squared effect sizes ranged from .01 - .12). However, tests for condition differences at the last assessment (i.e., the two-month follow-up) found no significant differences (see Table 3, FD vs. FC or follow-up dissonance versus follow-up control).

Long-Term:

Not examined.

Study 19

Summary

The study used a randomized controlled trial to examine 20 undergraduate women who reported body dissatisfaction and past or current sexual activity. Participants were randomly assigned to the dissonance-based intervention or a psycho-educational control group. Assessments at baseline, posttest, and six-month follow-up examined sexual risk behavior.

The study found that, relative to the control group, the intervention group reported a significantly greater decrease in:

• Unexpected sexual activity through the six-month follow-up.

Evaluation Methodology

Design:

Recruitment: The study examined a sample of 20 undergraduate women ages 18-21 who reported body image dissatisfaction and past or current sexual activity. The women were recruited via flyers and e-mails. Those meeting criteria for an eating disorder (past or current) were excluded, as were seniors who likely would no longer be on campus for the follow-up assessment.

Assignment: The study randomized participants into an intervention group (n = 10) or a psycho-educational control group (n = 10). The principal investigator led the single intervention group sessions with an undergraduate research assistant, thus confounding the program with the leaders. The control group received educational materials about the risks associated with eating disorders.

Assessments/Attrition: Assessments occurred at baseline, posttest, and six-month follow-up. All 20 participants completed the first two assessments, and 18 completed the final assessment (90%). The two dropouts were out of the country studying abroad.

Sample:

The sample of women had an average BMI of 23.66 and was 80% White.

Measures:

The self-reported measures included:

• Thin ideal internalization using the Ideal-Body Stereotype Scale-Revised (alpha = .76)
• Body dissatisfaction using the Dissatisfaction With Body Parts Scale (alpha =.79)
• Sexual assertiveness using the Hurlbert Index of Sexual Assertiveness (alpha =.89)
• Sexual Risk Survey for the frequency of sexual risk behaviors over the past six months in the form of an overall score (alpha =.84) and five subscales (e.g., unexpected sexual experience and having sex without a condom).

Analysis:

The analysis used repeated-measures analysis of variance with condition-by-time interaction terms.

Intent-to-Treat: The analysis used all available data.

Outcomes

Implementation Fidelity:

Due to time constraints, the full four-session program was condensed into three sessions.

Baseline Equivalence:

The authors stated that there were no baseline differences in demographic characteristics or scores on self-report measures between the intervention and control groups, though no tests were reported.

Differential Attrition:

Not examined, but likely not a problem given the reason for attrition of going abroad to study.

Posttest:

The tests for condition-by-time interactions found no significant effects on thin-ideal internalization or body dissatisfaction. In the seven tests for sexual risk outcomes, one reached statistical significance: The intervention group showed a greater decrease in the number of unexpected sexual experiences than the control group.

Long-Term:

Not examined.

Study 20

Summary

The study used a randomized controlled trial to examine 141 Brazilian high school students who expressed body image concerns. Participants were randomly assigned to the dissonance-based intervention or an assessment-only control group. Assessments at baseline and posttest examined eating disorder symptoms, depression, and negative affect.

The study found that, relative to the control group, the intervention group reported significantly greater:

• Reductions in body dissatisfaction
• Reductions in sociocultural influence of the media
• Reductions in depressive symptoms
• Reductions in negative affect
• Increases in body appreciation.

Evaluation Methodology

Design:

Recruitment: The sample of adolescent girls attending a technical education high school in Brazil was recruited via flyers between August 2015 and March 2017. The girls expressed body image concerns but did not meet the criteria for eating disorders. A total of 141 participants enrolled in the trial.

Assignment: Using a randomization website, the study randomized the 141 girls into an intervention group (n = 79) or an assessment-only control group (n = 62).

Assessments/Attrition: Assessments occurred at pretest and posttest. About 44% of the sample completed the posttest.

Sample:

The sample averaged 16.25 years of age. Sociodemographic information was not collected.

Measures:

The seven outcome measures came from self-reports. The four primary outcomes included:

• Body dissatisfaction (alpha = .93-.95).
• Sociocultural Attitudes Towards Appearance Questionnaire-3 (alpha = .90-.94).
• Disordered eating attitudes and behaviors (alpha = .71-.79).
• Eating disorders symptoms (alpha = .79-.83).

The three secondary outcomes included:

• Depressive symptoms (alpha = .71-.90).
• Body appreciation (alpha = .90-.91).
• Negative affect (alpha = .93-.95)

Analysis:

The analysis used repeated-measures analysis of variance with group-by-time interactions.

Intent-to-Treat: Only those intervention participants completing all four sessions and control group participants who completed both pretest and posttest were included in the analysis. According to the CONSORT diagram, the other intervention participants could not be contacted, but the loss of all program non-completers raises questions about the use of an ITT sample.

Outcomes

Implementation Fidelity:

The scripted intervention manual was translated into Portuguese, aiming to be as faithful as possible to the original. With regard to attendance, 50.6% of intervention participants attended all four sessions.

Baseline Equivalence:

There were no differences among participants in the two conditions on baseline outcomes, but no information was available for sociodemographic characteristics.

Differential Attrition:

Completion rates differed substantially across conditions - 51% in the intervention condition and 35% in the control condition. The study did not present any other information.

Posttest:

For the four primary outcomes, the group-by-time interactions were significant for two: The intervention group reported greater reductions in body dissatisfaction and sociocultural influence of the media than the control group. For the secondary outcomes, the group-by-time interactions were significant for all three: The intervention group compared to the control group reported greater reductions in depressive symptoms and negative affect, and greater increases in body appreciation. The reported effect sizes were large, ranging from d = .71-.96.

Long-Term:

Not examined.

Study 21

Summary

The study used a randomized controlled trial to examine 84 women who met diagnostic criteria for an eating disorder. Participants were randomly assigned to an eight-session version of the intervention or an active mindfulness control group. Assessments at posttest and six-month follow-up examined negative affect and eating disorder symptoms and diagnoses.

The study found that, relative to the control group, the intervention group had significantly less:

• Self-reported body dissatisfaction at posttest
• Diagnosed eating disorders at the six-month follow-up.

Evaluation Methodology

Design:

Recruitment: The sample of 84 women came from universities and surrounding communities in Eugene, Oregon, and Austin, Texas. The women had to meet diagnostic criteria for an eating disorder but were excluded if they had an extremely low BMI, suicide ideation, or substance misuse, all of which required special treatment.

Assignment: Participants were randomized to the intervention group (n = 39) or a supportive mindfulness group (n = 45) using a random numbers table. Both interventions consisted of eight weekly one-hour group sessions with 4-9 participants and were led by a pair of therapists at each site.

Assessments/Attrition: Assessments occurred at pretest, posttest (two months later), and six months after posttest. Completion rates were 92% at posttest and 89% at follow-up.

Sample:

The sample averaged 25 years of age and was 84% white, 11% Asian American, 4% Native American, and 1% African American. Roughly 11% reported Hispanic ethnicity. Most of their parents (63%) had graduated from college.

Measures:

The two primary outcomes came from a diagnostic interview, with the assessment done by research assistants who were not informed of the participants' condition. The posttest and follow-up assessments measured the eating disorder symptoms composite (alpha = .73), and the follow-up assessment measured the remission of the threshold or sub-threshold eating disorder that was required for enrollment.

The secondary outcome measures came from self-reports:

• Thin ideal internalization (alpha = .93)
• Dissonance regarding affirming the thin ideal (alpha = .75)
• Body dissatisfaction (alpha = .83)
• Negative affect (alpha = .93)
• Functional impairment with family members, peers, and at school and work (alpha = .78)

Analysis: The analysis used mixed-effects analysis of covariance models with adjustment for the pretest outcome values. Age was included to adjust for a significant baseline difference between conditions. Nested models accounted for group variability, where participants in the conditions were clustered in treatment groups, but not if the group-level variance was estimated to be zero. The reported significance tests were one-tailed.

Intent-to-Treat: The analysis used multiple imputation to include all participants.

Outcomes

Implementation Fidelity:

Intervention participants attended an average of 5.1 of the eight sessions. Only 8% attended all eight sessions but 62% attended five to seven sessions. They completed an average of 70% of the home exercises.

Baseline Equivalence:

Participants in the two conditions did not differ significantly on demographics, pretest values of the outcomes, use of psychiatric medications, or ancillary treatment, with one exception: The average intervention age of 22.6 for the Body Project intervention fell well below the age of 26.6 for the mindfulness intervention. Although not significantly different, the percentage with a college degree was substantially higher in the mindfulness group (42%) than in the Body Project intervention group (28%).

Differential Attrition:

The authors stated that, based on the non-significant Little's test (p = .076), "the missing completely at random assumption (MCAR) remained tenable." They also stated that "Attrition was not related to condition, demographics, or pretest outcomes."

Posttest:

One of the two primary outcomes differed significantly at the six-month follow-up: 77% of Body Project participants and 60% of the mindfulness participants no longer met the criteria for a threshold or subthreshold eating disorder (p = .049).

For the 10 secondary outcomes (five outcomes at both posttest and follow-up), there was one significant effect (p < .05). The Body Project intervention group reported lower body dissatisfaction at posttest than the mindfulness group. The authors noted that, although not significant, all the comparisons favored the Body Project intervention group.

Long-Term:

Not examined.

Study 22

Summary

The study used a quasi-experimental design that examined 127 undergraduate men at two universities in the United Kingdom. The participants were non-randomly assigned to an adapted Body Project intervention for men or an assessment-only control group. This adapted version focused on the male appearance ideal and was shorter than the Body Project program tested with females. Measures of body dissatisfaction and bulimic symptoms were assessed at baseline, posttest, and three-month follow-up.

The study found no significant intervention effects on changes in the outcomes but found that, relative to the control group, the intervention group reported significantly lower:

• Muscularity dissatisfaction at posttest only
• Internalization at posttest only.

Evaluation Methodology

Design:

Recruitment: The study examined 127 undergraduate men recruited from two universities in the United Kingdom from September 2013 to January 2015.

Assignment: The quasi-experimental design selected men who were enrolled in psychology single major undergraduate degrees for the intervention condition (n = 74) and men who were enrolled in psychology dual major degrees (e.g., psychology and criminology) for the assessment-only control condition (n = 53). The lead researcher served as the facilitator for the intervention groups, which confounded the program with the single leader.

Assessments/Attrition: Assessments occurred at baseline, posttest, and three-month follow-up. Of the 127 assigned participants, 87% completed the baseline assessment, 61% completed the posttest, and 54% completed the three-month follow-up.

Sample:

Of those who completed the baseline assessment, the majority were young (18-45 years, with an average age of 20.8 years) and heterosexual (91%). Most participants identified as British (96%) and white (87%).

Measures:

The seven outcome measures came from student self-reports.

• Body fat dissatisfaction (e.g., "I feel fat") as measured by the Male Body Attitudes Scale-Revised (alpha = .89).
• Muscularity dissatisfaction as measured by the Drive for Muscularity Scale (alpha = .93).
• Body appreciation (i.e., positive body image) as measured by the Body Appreciation Scale (alpha = .89).
• The drive for masculinity behaviors (e.g., lifting weights to build muscle) as measured by the Muscularity Enhancing Behaviors Subscale (alpha = .89).
• Bulimic pathology (e.g., binge eating and purging) as measured by the diagnostic items of the Eating Disorders Examination Questionnaire (alpha = .77).
• Appearance comparison to others as measured by the Physical Appearance Comparison Scale (alpha = .70).
• Internalization of the male appearance ideal as measured by the subscale of the Sociocultural Attitudes Towards Appearance Questionnaire 3-Male (alpha = .86).

Analysis:

The analysis used mixed repeated-measures models in which time was nested within individuals and condition-by-time interactions tested for program effects. Covariates included baseline outcomes and demographic measures. Significance tests for mean differences at posttest and follow-up adopted a p-value of .025 to correct for the family-wise error rate. Multiple imputation used all available variables (drop-out status, condition, demographic, and outcome variables) in the imputation model.

Intent-to-Treat: The multiple imputation included all participants with baseline data (n = 110). A per-protocol analysis used participants who had complete data and, for the intervention group, attended both sessions (n = 69).

Outcomes

Implementation Fidelity:

About 80% of the intervention participants attended the first session and 53% attended both sessions. On average, the facilitator delivered 78.5% of the content in each session.

Baseline Equivalence:

Tests found no significant differences between conditions for 11 demographic and outcome measures.

Differential Attrition:

Attrition was high and differed across conditions (36% for the control group and 53% for the intervention group). The Little's MCAR test was significant (p = .009), meaning that the missing data were not missing completely at random. Tests showed that dropping out at both posttest and three-month follow-up was significantly related to lower baseline internalization.

Posttest:

For the ITT sample, the condition-by-time coefficients in Table 4 showed no significant program effects for the seven outcomes. The per-protocol analysis showed the same. Pairwise comparisons in Table 3 for the ITT sample revealed two significant differences in 14 tests: muscularity dissatisfaction at posttest (d = -.36) and internalization at posttest (d = -.37).

Long-Term:

Not examined.

Study 23

Summary

The study described their study of 29 female college athletes as "not fully randomized" but assigned the women randomly to three conditions: a three-session cognitive-dissonance program that adapted the Body Project program for female athletes that served as the intervention group (n = 12), an alternative Healthy Weight intervention group (n = 7), and a waitlist control group (n = 10). A three-week posttest assessed negative affect and disordered eating.

The study found no significant program effects.

Evaluation Methodology

Design:

Recruitment: The sample included 29 female athletes from a Division I university located in the southwest United States. A list of university athletes was used to identify and recruit female student-athletes who fell in the bottom half on a self-reported body satisfaction score. Although 41 met the criterion, two with a previously diagnosed eating disorder were excluded. Of the 39 remaining, only 29 were included in the study because of scheduling problems and previous contact with the study personnel.

Assignment: The authors stated that "we could not do a fully randomized assignment" but also stated that "29 of the 39 identified and eligible athletes were randomly assigned to condition: Healthy Weight (n = 7), Cognitive-dissonance (n = 12), and [waitlist] Control (n = 10)." The inconsistency might stem from the treatment of the 10 non-randomized participants, but the authors provided no clarification.

Assessments/Attrition: The baseline assessment occurred three months before the intervention began. The study referred to a Time 2 assessment that presumably came at the end of the three-session program. Although the study made no mention of the presence or absence of attrition, the statement that "All individuals attended all three sessions" suggests that all participants completed the posttest.

Sample:

The sample averaged 19.32 years of age. Regarding race/ethnicity, 82% were Caucasian, 4% African American, 7% Hispanic, and 7% Native American. The athletes' actual and ideal body mass index (BMI) were 23.51 and 21.58, respectively.

Measures:

The ten outcome measures came from self-reports. It appears that the alpha values relied on figures from other studies.

• Thin-ideal internalization from the Beliefs About Attractiveness Scale-Revised, which had two subscales, one for being thin and attractive and one for being fit and in shape (alpha = .85 for both).
• Body satisfaction from the Body Parts Satisfaction Scale-Revised (alpha = .90)
• Body dissatisfaction from the Body Shape Questionnaire-Revised (alpha = .96).
• Negative affect from the Positive Affect and Negative Affect Scale-Revised, with four subscales for sad/depressed, anxious, shame/guilt, and stress items (no alpha reported).
• Disordered eating from the Bulimia Test-Revised for bulimia (alpha = .98) and from the Dutch Restrained Eating Scale for current dieting behaviors (alpha = .95).

Analysis:

The analysis used repeated-measures ANOVA models with group-by-time interaction terms.

Intent-to-Treat: The analysis used 29 of the 39 eligible participants, but it remains unclear whether the exclusions came after randomization.

Outcomes

Implementation Fidelity:

Not examined.

Baseline Equivalence:

The conditions did not differ significantly on actual BMI, ideal BMI, age of first period, or average length of period. There was, however, a significant difference on age, with athletes in the cognitive-dissonance and control groups being older than the athletes in the Healthy Weight group. For the outcomes, the three groups did not differ significantly with respect to bulimic behaviors, depression/sadness, anxiety, stress, shame/guilt, the importance of being thin and attractive, or body satisfaction. The groups, however, were significantly different on the importance of being physically fit and in shape, with control athletes scoring lower than the other two conditions.

Differential Attrition:

It is likely, though not certain, that there was no attrition.

Posttest:

The time-by-group interaction did not reach statistical significance for any of the 10 outcomes. The authors noted, however, that post-hoc exploratory analyses suggested that the cognitive-dissonance intervention provided "limited positive effects, particularly with respect to decreases in sadness/depression, decreases in internalization of the importance of being physically fit and in shape, and increases in body satisfaction."

Long-Term:

Not examined.

Study 24

Summary

Of the three studies described in Vanderkruik (2019) and Vanderkruik et al. (2020), only the second one included a control group and is described below. The two articles largely presented the same information except for some additional details on participant flow and missing data in Vanderkruik et al. (2020). The studies examined only the peer-led version of the program.

The study used a randomized controlled trial to examine 81 female high school students who were assigned to the peer-led intervention group or an assessment-only control group. Assessments at the end of the program (four weeks after baseline) measured depression, anxiety, and negative affect.

The study found at posttest that relative to the control group, the peer-led intervention group reported significantly improved:

• Body dissatisfaction
• Dietary restraint
• Thin-ideal internalization
• Body surveillance
• Loneliness
• Self-compassion.

Evaluation Methodology

Design:

Recruitment: The sample of 81 participants, all females in grades 9-12, came from two public high schools in Boulder, Colorado. Flyers distributed around the high schools, sent via e-mail, and given to classes and student groups were used to recruit women with an interest in improving body satisfaction. The intervention was implemented over two semesters, fall 2018 and spring 2019.

Assignment: The study randomly assigned participants to either the Body Project intervention condition (n = 43) or an assessment-only control condition (n = 38) with a block randomization technique (i.e., all participants were assigned to one of the eight group sessions based on schedule availability and half of each group was randomized to the intervention).

Assessments/Attrition: The posttest assessment occurred four weeks after baseline. Of the 81 randomized participants, 63 (78%) completed the posttest. Dropouts included seven who left after allocation and 11 who completed the baseline assessment but not the posttest assessment.

Sample:

The sample ranged in age from 14-18 years. Most identified as White (77%), followed by Asian/Asian American (28%), Hispanic/Latino (8%), African American/Black (1.4%), and more than one race/ethnicity category (15%).

Measures:

The 12 outcome measures came from self-reports:

• Body dissatisfaction from the Body Parts Scale (alpha = .92).
• Thin-ideal internalization from the Thin Ideal Internalization Scale (alpha = .92).
• Dietary restraint from the Dutch Restrained Eating Scale (alpha = .95).
• Negative affect from the Positive Affect and Negative Affect Scale-Revised (alpha =.96).
• Depressive symptom severity from the Patient Health Questionnaire (alpha = .85).
• Anxious symptoms from the Generalized Anxiety Disorder Scale (alpha = .91).
• Self-compassion from the Self-Compassion Scale (alpha = .92).
• Self-efficacy from the General Self-Efficacy Scale (alpha = .91).
• Self-esteem from the Rosenberg Self-Esteem Scale (alpha =.91).
• Body surveillance from the Body Surveillance subscale of the Objectified Body Consciousness Scale (alpha = .72).
• Loneliness from the UCLA Loneliness Scale (alpha = .90).
• Peer influence on eating and body concerns from the Inventory of Peer Influence on Eating Concerns (alpha = .96).

Analysis:

A series of analyses of covariance tested for differences between groups in mean changes between baseline and postintervention when controlling for baseline scores and semester enrolled in the Body Project workshop. As a sensitivity check for missing data, the study reported effect sizes using imputation, which were similar to those using complete cases.

Intent-to-Treat: The CONSORT diagram lists five intervention participants that were excluded because they "Did not receive allocated intervention (dropped out)," which suggests that non-attendees were not followed. The study was not clear if the multiple imputation included these randomized participants.

Outcomes

Implementation Fidelity:

Mean adherence across all groups was 9.25 and mean facilitator competence was 9.51 on a 10-point scale. Of the intervention participants, 75.8% completed all four sessions, 21.6% completed three sessions, and 2.7% completed one session.

Baseline Equivalence:

The authors reported finding no significant baseline differences between the two conditions for any of the key outcome or demographic variables.

Differential Attrition:

The authors reported finding no significant differences between participants who completed the follow-up assessment and those who did not. They also reported no significant interactions with randomization group and completer status on baseline scores.

Posttest:

Six of the 12 tests reached statistical significance. The intervention group scored significantly lower than the control group for body dissatisfaction, dietary restraint, thin-ideal internalization, body surveillance, and loneliness and scored significantly higher on self-compassion. Effect sizes ranged widely from .55 (medium) to 1.38 (large). The intervention did not significantly improve scores for negative affect, depression, or anxiety.

Long-Term:

Not examined.

Study 25

Summary

The study evaluated a virtually implemented version of the peer-led program sessions via Google Hangouts rather than the original face-to-face version. The use of a booster contact at 12 months consisting of a message and brief assignment also differed from the original version. Undergraduates in Psychology served as peer leaders for the internet sessions.

It used a randomized controlled trial to examine 443 Swedish female high school students who were assigned to the peer-led intervention group, an expressive writing alternative intervention group, or a waitlist control group. Assessments for all three conditions at six months were followed by assessments for the intervention and expressive writing groups through 24 months. The incidence of a diagnosed eating disorder served as the primary outcome.

The study found that, relative to the expressive writing group or the control group, the peer-led intervention group had reduced:

• diagnosed eating disorders (six months and 24 months)
• clinical impairment (six months)
• body part dissatisfaction (six months)
• body shape dissatisfaction (six months)
• thin-ideal internalization (six months and 24 months)

Evaluation Methodology

Design:

Recruitment: The sample consisted of young Swedish women ages 15-20 with body image concerns (a subjective sense of body dissatisfaction) but without a current diagnosis of eating disorders, concurrent psychological treatment, or severe depression, suicidality, or other serious conditions. The recruitment included ads placed in high schools, project ambassadors visiting high schools, and Instagram and Facebook advertisements. A total of 1,678 individuals declared an interest to participate, of which 1,242 were assessed for eligibility and 443 were eligible and consented.

Assignment: The 443 participants were randomly assigned using a list obtained from the Research Randomizer (www.randomizer.org) to one of three arms: (a) Body Project delivered through virtual groups (n = 149), (b) an expressive writing alternative intervention (n = 148), or (c) a waitlist control condition (n = 146).

Assessments/Attrition: Assessments occurred at baseline, postintervention, and at six-month follow-ups for all participants and at 12-, 18-, and 24-month follow-ups for the two active conditions but not the waitlisted control group. With a 12-month booster session, the 24-month follow-up meets the long-term requirement. Completion rates for the full sample (n = 443) were 70% at posttest and 65% at the six-month follow-up; completion rates for the truncated sample (n = 297) were 56% at 12 months, 55% at 18 months, and 51% at 24 months.

Sample:

The sample had a mean age of 17 years. About 91% of the participants were students, and about 51% lived in the four largest cities.

Measures:

The primary outcome was the clinical diagnosis of an eating disorder onset measured at six, 12, and 24 months through a phone interview using diagnostic items of the Eating Disorders Examination for anorexia nervosa, bulimia nervosa, binge eating disorders, and other specified feeding or eating disorders. The clinical interviews were done by trained research assistants who were blind to condition. The items were standardized and summed into a composite score (alpha = .82).

The other seven secondary outcome measures came from participant self-reports.

• Eating disorder symptoms from the Eating Disorder Diagnostic Scale (alpha = .88).
• Negative affect from the Positive and Negative Affect Schedule-Revised (alpha = .87).
• Restraint in eating, one of four subscales from the Eating Disorders Examination Questionnaire (alpha = .86).
• Impact of eating disorders on functioning in everyday life from the Clinical Impairment Assessment (alpha = .93).
• Body part dissatisfaction from the Body Parts Dissatisfaction Scale, a count of body parts checked because of the desire for them to be smaller or thinner.
• Body shape dissatisfaction with a focus on feeling fat from the Body Shape Questionnaire brief version (alpha = .91).
• Internalization of the thin ideal from the Ideal Body Stereotype Scale-Revised (alpha = .87).

Analysis:

Analysis of the incidence of eating disorders used the proportional hazard model. Analysis of continuous outcomes used mixed models with full information maximum likelihood estimation, repeated measurements nested within individuals, and an autoregressive covariance type for the repeated measures. The models did not adjust for clustering by intervention group, noting that the intraclass correlations were small (mostly around .02 or .04). Significance tests adjusted for multiple comparisons at different time points with the Sidak correction.

Intent-to-Treat: The FIML estimation allowed the inclusion of all randomized participants.

Outcomes

Implementation Fidelity:

Adherence ratings ranged from 4.7 to 8.9 with a mean of 7.6 (i.e., between good and very good) on a 10-point scale. Competence ratings of the group leaders ranged from 5.6 to 7.2 with a mean of 6.5 (i.e., slightly above average). About 72% of participants attended at least half the sessions.

Baseline Equivalence:

The authors stated (p. 648) that there "were no statistically or clinically significant differences between the three conditions on any of the demographic or outcome variables at baseline." Tables 2 and 3 present tests indicating the equivalence of the outcomes at baseline.

Differential Attrition:

The authors stated (p. 654) that "detailed dropout analysis (for the entire sample, and on each condition separately) did not reveal any statistically or clinically meaningful differences between the drop-out group and the completers." However, the CONSORT diagram reveals one substantial condition difference in completion rates (i.e., 48% versus 61% at 18 months). Attrition analyses submitted upon request by Blueprints did not show attrition bias of condition by baseline outcomes or demographic characteristics.

Posttest:

Results for all three conditions covered the period from baseline through six months (Table 2). Five of eight outcomes showed both significant time-by-condition interactions and significant condition differences at six months: the primary outcome of an eating disorder diagnosis, clinical impairment, body part dissatisfaction, body shape dissatisfaction, and thin-ideal internalization. At the six-month follow-up effect sizes varied from .30-.58.

Long-Term:

The long-term results included only two of the three conditions, as the waitlist control group had received the program soon after posttest. The primary outcome, incidence of eating disorders, was measured through 24 months after baseline. Hazard models found that the intervention group had significantly lower risk of the onset of an eating disorder diagnosis than the expressive writing group (hazard ratio = .26).

In Table 3, two of eight outcomes showed both significant time-by-condition interactions and significant condition differences at 24 months: the primary outcome of an eating disorder diagnosis and thin-ideal internalization. Effects sizes across the outcomes and various time points ranged from .26-.65.