Food, Mood, and Attitude: Reducing Risk for Eating Disordersin College Women

Debra L. FrankoNortheastern University and Harvard Eating Disorders Center

Laurie B. MintzUniversity of MissouriColumbia

Mona VillapianoMassachusetts Eating Disorders Association and Inflexxion, Inc.

Traci Craig Green, Dana Mainelli, Lesley Folensbee,and Stephen F. Butler

Inflexxion, Inc.

M. Meghan Davidson and Emily HamiltonUniversity of MissouriColumbia

Debbie Little and Maureen KearnsNortheastern University

Simon H. BudmanInflexxion, Inc.

Food, Mood, and Attitude (FMA) is a CD-ROM prevention program developed to decrease risk foreating disorders in college women. Female 1st-year students (N 240) were randomly assigned to theintervention (FMA) or control group. Equal numbers of students at risk and of low risk for developingan eating disorder were assigned to each condition. Participants in the FMA condition improved on allmeasures relative to controls. Significant 3-way interactions (Time Condition Risk Status) werefound on measures of internalization of sociocultural attitudes about thinness, shape concerns, and weightconcerns, indicating that at-risk participants in the intervention group improved to a greater extent thandid low-risk participants. At follow-up, significantly fewer women in the FMA group reported overeatingand excessive exercise relative to controls.

Keywords: eating disorders, prevention, risk, college women, risk factors

Excessive dieting, unhealthy weight loss practices, and disor-dered eating behaviors occur frequently among college women andcan adversely influence physiological, psychological, and behav-ioral functioning (Hill, 2002). Only a small percentage of collegewomen (1%3%) have diagnosable eating disorders, although

many (estimated at 10%30%) appear to be at risk for developingan eating disorder over the course of their college years. Mintz,OHalloran, Mulholland, and Schneider (1997) found in a sampleof 1st-year college women that 4% had eating disorders and 19%had some risk factors (e.g., chronic dieting, use of appetite controlpills) for the development of diagnosable eating disorders. In astudy of 1st-year college women, 15% of those deemed to be atrisk (10% of the sample) in the fall semester moved into theprobable bulimia category by the spring semester (Drewnowski,Yee, Kurth, & Krahn, 1994).

Attention has thus turned to the prevention of risk factors foreating disorders (Franko & Orosan-Weine, 1998). On the basis ofa meta-analysis of 38 eating disorder prevention studies conductedin various populations (e.g., adolescents, college students), Sticeand Shaw (2004) reported that 61% of prevention programs re-sulted in significant reductions in at least one risk factor, such asbody dissatisfaction, and one third of programs led to eitherdecreases in eating pathology or prevention of increases seen incontrols. Nearly half (18 of 38) of these prevention studies wereconducted with college students. By and large, the approach ofthese programs (e.g., Martz & Bazzini, 1999; Mutterperl & Sand-erson, 2002; Nicolino, Martz, & Curtin, 2001) were either univer-sal (i.e., prevention programs in which the goal is to strengthenprotective factors in symptom-free individuals; see Mrazek &Haggerty, 1994) or selective (i.e., prevention programs in which

Debra L. Franko, Department of Counseling and Applied EducationalPsychology, Northeastern University and Harvard Eating Disorders Center;Laurie B. Mintz, M. Meghan Davidson, and Emily Hamilton, Departmentof Educational, School, and Counseling Psychology, University of Mis-souriColumbia; Mona Villapiano, Massachusetts Eating Disorders As-sociation and Inflexxion, Inc., Newton, Massachusetts; Traci Craig Green,Dana Mainelli, Lesley Folensbee, Stephen F. Butler, and Simon H. Bud-man, Inflexxion, Inc., Newton, Massachusetts; Debbie Little and MaureenKearns, Department of Counseling and Applied Educational Psychology,Northeastern University.

This research was supported, in part, by Small Business InnovationResearch Grant R44DK53583, awarded to Simon H. Budman, and wasfunded jointly by the National Institute of Diabetes & Digestive & KidneyDiseases, the National Institute of Child Health and Human Development,and the National Institute of Mental Health. We thank Frederick L. New-man and Ann Marie Rakowski for their contributions to the study.

Correspondence concerning this article should be addressed to Debra L.Franko, Department of Counseling and Applied Educational Psychology,Northeastern University, 203 Lake Hall, Boston, MA 02115-5000. E-mail:d.franko@neu.edu

Health Psychology Copyright 2005 by the American Psychological Association2005, Vol. 24, No. 6, 567578 0278-6133/05/$12.00 DOI: 10.1037/0278-6133.24.6.567

567

the goal is to modify risk factors; e.g., Bearman, Stice, & Chase,2003; Celio et al., 2000; Franko, 1998; Stice & Ragan, 2002; Stice,Trost, & Chase, 2003). Across college-specific programs, thosethat were effective tended to be interactive and consisted of mul-tiple sessions (Bearman et al., 2003; Kaminski & McNamara,1996; Stice & Ragan, 2002; Winzelberg et al., 1998, 2000) ratherthan being a one-shot deal (Mann et al., 1997; Martz & Bazzini,1999; Mutterperl & Sanderson, 2002; Nicolino et al., 2001). More-over, college-specific programs that were selective in nature hadlarger effect sizes than universal programs. Overall, the effectiveprograms resulted in some reduction in risk factors, suggesting thateating disorder prevention programs designed for college studentsmay be useful, given substantial prevalence rates in this age group(Drewnowski et al., 1994; Heatherton, Mahamedi, Striepe, Field,& Keel, 1997).

A series of studies conducted by a Stanford research groupoffers promise in the use of multimedia to reduce risk factors foreating disorders in college women (Celio et al., 2000; Winzelberget al., 2000; Winzelberg et al., 1998; Zabinski, Celio, Wilfley, &Taylor, 2003). Winzelberg and colleagues (1998, 2000) investi-gated an 8-week Internet-based program (Student Bodies) de-signed to reduce risk factors and found greater decreases in bodydissatisfaction in the intervention group than in the control condi-tion. Celio et al. (2000) examined the effects of the same programrelative to an eating disorders course and a waiting list controlgroup. Decreases in body dissatisfaction were found for the inter-vention group at posttest (though not at follow-up), and dietingwas reduced at both time points. In a study using the same programwith high-risk individuals, Zabinski et al. (2001) reported that bothparticipants using the program and those in a control group im-proved over time (i.e., no significant difference between high-riskparticipants in the intervention group relative to the control group).Nevertheless, Zabinski et al. (2001) also reported effect sizes thatsuggested some positive impact of the intervention.

The findings and conclusions from previous studies were centralin the design and testing of Food, Mood, and Attitude (FMA), a2-hr CD-ROM program developed to reduce risk factors for eatingdisorders. First, as was done by the Stanford group, our interven-tion used multimedia technology. Such technology is considered arecent innovation in health education and prevention and is ap-pealing because it is accessible to large groups, cost-effective, andless labor intensive than traditional leader-led programs (Budman,2000). Moreover, multimedia technology has the distinct benefit ofproviding a multisensory experience and actively engaging partic-ipants. Indeed, this active engagement of participants (i.e., inter-active nature) is in line with conclusions reached by Stice andShaw (2004). Also, consistent with Stice and Shaws findings,FMA (a) was designed especially for college women, (b) wasrelatively short in duration, and (c) consisted of multiple compo-nents. Finally, the shorter time span of the program is in line withsuggestions from the Stanford researchers (Winzelberg et al.,1998, 2000).

FMA also was designed for both low-risk and at-risk groups(both universal and selective prevention) as a logical next step onthe basis of findings from the Stanford group, which included ahigh-risk group only (Zabinski et al., 2001). We specifically de-veloped and tested FMA for suitability with participants deemed tobe at risk for developing eating disorders, as well as those at lowrisk who might benefit from the information provided by increas-ing their knowledge about risk factors. In addition, on the basis of

Zabinski et al.s (2001) self-critique of how they defined partici-pants as high-risk (i.e., having high body-image concerns), wesought to both define risk more specifically and empirically ex-amine our categorization of risk.

The theoretical conceptualization of eating disorder risk thatunderlies the development of FMA is the dual pathway modelproposed by Stice, Nemeroff, and Shaw (1996). Although severalother theories exist (Smolak, Levine, & Gralen, 1993; Striegel-Moore, Silberstein, & Rodin, 1986; Vohs, Bardone, Joiner,Abramson, & Heatherton, 1999), the model by Stice et al. (1996)has been the one most scrutinized and tested in the empiricalliterature. This model conceptualizes risk of eating disorder pa-thology in the following way. Sociocultural pressure to be thinleads to internalization of the thin ideal. Because this ideal islargely unattainable, body dissatisfaction results, which in turnleads to dieting and negative affect. Both dieting and negativeaffect increase the risk of one developing eating disorder behav-iors, particularly bulimic symptoms such as binge eating andcompensatory behaviors. These five factors (perceived pressure tobe thin, thin-ideal internalization, body dissatisfaction, dieting, andnegative affect) have been found to predict bulimic symptom onsetin a number of studies (Stice & Agras, 1999; Stice, Killen, Hay-ward, & Taylor, 1998; Stice, Presnell, & Spangler, 2002) andreceived support for increasing risk for eating disorder behaviorsin a recent meta-analysis (Stice, 2002).

In addition to the focus on the dual-pathway model, FMA alsoincorporated relevant aspects of interpersonal/relational theory aswell as cognitivebehavioral theory, both of which have beenwidely used by eating disorder researchers and clinicians (Stein etal., 2001). Interpersonal/relational theory acknowledges the role ofinterpersonal needs and concerns in the acquisition of high-riskbehaviors. It has been found that efforts to be accepted by onespeers and to feel that one is attractive to others are powerfulmotivators with regard to eating and dieting (e.g., Wilfley et al.,1993). Paxton, Schutz, Wertheim, and Muir (1999) found thatadolescent girls who had friends who dieted were much morelikely to engage in eating pathology, and Stice, Maxfield, & Wells(2003) recently reported that exposure to social pressure to be thinincreased body dissatisfaction. In addition, theoretically relevantcognitivebehavioral constructs, such as cognitive restructuringand cognitive distortions, were included in the program design andactivities on the basis of a large body of literature supporting theuse of cognitivebehavioral therapy in treating eating disorders(Stein et al., 2001).

We also considered recent conceptualizations of preventiontheory in the design of FMA. Specifically, three theoretical con-ceptualizations were used in the development of FMA (Levine &Piran, 2001). One, the disease-specific social cognition theory(Perry, 1999), which focuses on eliminating risk factors seen asspecific to a disorder, informed the overall content of FMA.Programs that follow this model use various agents of socialinfluence (e.g., peers) to transmit information, teach skills, andincrease knowledge. Two, the nonspecific vulnerability stressormodel (Cowen, 2000) was used in developing program content thatwas not specific to eating disorders but focused on more generalfactors such as self-esteem and coping skills that are known to beimportant for prevention of eating disorders. Finally, given therelatively recent introduction of prevention theory specific to eat-ing disorders, we also drew from harm reduction, a conceptualapproach used to prevent other intractable behaviors, including the

568 FRANKO ET AL.

successful prevention of alcohol misuse on college campuses(Marlatt, 1995). Specifically, this approach provides informationand objective feedback about the dysfunctional behavior and itsconsequences, places responsibility for change on the respondent,provides suggestions about ways to make changes and build self-efficacy, and teaches skills for specific behavior change (Miller etal., 1995). In short, in developing the content of FMA, we trans-lated theoretical ideas (e.g., dual-pathway model, interpersonaltheory, cognitivebehavioral theory, and prevention theories suchas harm reduction) into the program components.

Hypotheses

The current study examined the efficacy of FMA with 1st-yearcollege women on measures of (a) knowledge; (b) internalizationand awareness of the sociocultural ideal; and (c) weight concerns,shape concerns, and restrained eating. For the latter two measures(Items b and c), we hypothesized a significant three-way interac-tion of Time Condition (FMA vs. control) Screening Status(at risk vs. low risk), such that at-risk women in the FMA condi-tion would show the most positive change. We predicted thatlow-risk women in the FMA condition would not evidence anyiatrogenic effects with respect to Items b and c but would report anincrease in knowledge over time equal to that of the high-riskwomen in the FMA condition.

Method

Design and Screening ParticipantsThis study involved two phases: (a) a screening phase to identify

participants for the efficacy study on the basis of inclusion and exclusioncriteria and (b) a randomized controlled clinical trial comparing the FMAprogram to an attention control condition. The study was approved by theInstitutional Review Boards of Northeastern University, University ofMissouriColumbia, and Inflexxion, Inc. Screening participants included827 first-year female students at the two university sites.

Screening MeasuresDemographic and follow-up contact questionnaire. All participants

completed a short demographic questionnaire and a form requesting futurecontact information.

The Questionnaire for Eating Disorder Diagnoses (Q-EDD; Mintz et al.,1997). The Q-EDD is a reliable and valid self-report measure that can beused as both a diagnostic tool and a screening tool to assess those at riskfor developing eating disorders on the basis of current behaviors. Forclarity, an understanding of the scoring of the Q-EDD is necessary. Scoringfor the Q-EDD yields categorical labels. At the most general level are thecategories of eating disordered (diagnosis from the Diagnostic and Statis-tical Manual of Mental Disorders [4th ed.; DSMIV; American PsychiatricAssociation, 1994]: anorexia nervosa [AN], bulimia nervosa [BN], andfour types of eating disorders not otherwise specified [EDNOS]) andnon-eating disordered (non-ED; no DSMIV diagnosis). The non-eatingdisordered category comprises two categories: asymptomatic (no symp-tomatic behaviors and/or behavioral risk factors) and symptomatic (nodiagnosable disorder, including no EDNOS disorder, but some transient ormild symptomatic behaviors and/or behavioral risk factors such as chronicdieting).

Evidence for the Q-EDD as a diagnostic tool comes from studiescomparing Q-EDD diagnoses and those yielded by clinical interview.Mintz et al. (1997) reported predictive validity statistics for the ED versusnon-ED diagnostic differentiation: false-negative rate .03, false-positiverate .02, sensitivity .97, specificity .98, positive predictive power.94, and negative predictive power .99. In addition, accuracy rates

ranged from 69% to 97% for the differentiation between different EDdiagnoses, including between AN, BN, and subthreshold (EDNOS) ver-sions of these disorders. Such findings supported the use of the Q-EDD toscreen out women with a diagnosable ED for the current prevention study.

Evidence for the Q-EDD as a screening tool for those at risk for EDs butnot possessing diagnosable EDs comes from validity studies of the Q-EDDand from findings explicitly examining the Q-EDD symptomatic group.First, Mintz et al. (1997) reported a high overall accuracy rate of 90% (k .82), indicating a high agreement between Q-EDD categorization andclinician assessment that a woman was at risk for an eating disorder, asopposed to having either a diagnosable disorder or no behavioral symptomsor risk factors. Second, because the Q-EDD categorizes symptomaticwomen mainly on the basis of symptomatic behaviors and/or behavioralrisk factors, studies have investigated whether these women also possessattitudinal and cognitive risk factors, as would be expected. Tylka andSubich (1999) reported that the symptomatic women (as defined by theQ-EDD) differed as would be expected from DSMIV eating-disorderedand asymptomatic women on a number of cognitive and attitudinal vari-ables specified in risk models such as Stice et al.s (1996), includinginternalization of the thin ideal, body dissatisfaction, and perceived pres-sure to be thin. Such findings support the use of the Q-EDD to categorizeat-risk and low-risk women for the current prevention study. Nevertheless,as a precaution, we also tested differences in risk factors between ourQ-EDD-defined at-risk and low-risk groups to verify expected groupdifferences (see Results).

Screening ProcedureTrained research assistants (RAs) used advertisements to recruit women

around campus, in residence halls, and in student centers in the fall of 2001.The flyers read Participate in a study to evaluate computer and videoprograms for women on health, nutrition, and womens issues and pro-vided contact information to sign up for the study. Recruitment aimed toinclude at least 25% minority students, so recruitment efforts focused onensuring an adequate pool of eligible minority women. Volunteers whoagreed to participate gave informed consent, including a statement that theymay be contacted to participate in the second phase of the study. They thencompleted the demographic and follow-up contact questionnaire and theQ-EDD. Participants were paid $5 for participation in the screening phaseof the project.

Screening measures were scored for inclusion and exclusion criteria todetermine eligibility for the efficacy study. Inclusion criteria were (a)female, (b) first semester of university, (c) between 18 and 22 years of age,(d) willing to sign informed consent and undergo initial and follow-upassessments, and (e) identification by the screening measure as being eitherat risk or of low risk for developing eating disorders. At risk were thoseparticipants categorized by the Q-EDD as symptomatic (no diagnosabledisorder, including no EDNOS disorder, but some transient or mild symp-tomatic behaviors and/or behavioral risk factors), and low risk was cate-gorized by the Q-EDD as those who were asymptomatic (no symptomaticbehaviors and/or behavioral risk factors). Exclusion criteria included (a)being currently in treatment with a diagnosis of any type of eating disorderor (b) identification by the Q-EDD as meeting criteria for an eatingdisorder diagnosis. If a student met either exclusion criteria, she was senta letter informing her that she was not chosen for the study that includedreferrals for counseling services.

Efficacy Study Participants and MeasuresThe efficacy study was conducted with 240 first-year female students

from two universities (northeastern and midwestern United States), whowere chosen from the screening participants (see Efficacy Study Proce-dure). Participants were equally divided between the two campuses. Table1 summarizes the characteristics of the study sample. Except as noted, thefollowing measures were administered at baseline, postintervention, and3-month follow-up.

569REDUCING RISK FOR EATING DISORDERS

Knowledge test. We developed a questionnaire to test knowledge ex-pected to be gained after exposure to FMA. An initial pool of questionsbased on information provided in FMA was given to 59 first-year collegewomen. Any item answered correctly by less than 20% of the sample wassubmitted to eating disorder experts, who rated them for importance andcontent validity. The best items composed the final version, consisting of20 multiple-choice items.

Sociocultural Attitudes Toward Appearance Questionnaire (SATAQ;Heinberg, Thompson, & Stormer, 1995). This 14-item questionnaire in-cludes two subscales: an Awareness subscale, measuring recognition of thesocietal influence on attitudes toward appearance, and an Internalizationsubscale, measuring acceptance or endorsement of those societal standards.Internalization of the thin-ideal standard has been shown across multiplestudies to be a risk factor for eating disorders (Streigel-Moore & Cachelin,2001) and is one of the five risk factors in the guiding model of risk thatwe used in the development of FMA (Stice et al., 1996). Acceptableinternal consistency has been demonstrated for female undergraduates (.71for Awareness and .88 for Internalization; Heinberg et al., 1995). Evidencefor convergent validity has been demonstrated by significant correlationsbetween both subscales and various measures of body image and eatingdisturbances (Heinberg et al., 1995).

Eating Disorder Examination Questionnaire (EDE-Q; Fairburn & Beg-lin, 1994; Fairburn & Cooper, 1993). This questionnaire has 36 ques-tions and four subscales (Shape Concerns, Weight Concerns, Restraint, andEating Concerns), which assess cognitions, attitudes, and behaviors asso-ciated with eating disorder symptomatology and which are commonly usedin efficacy studies of eating disorder prevention, including in studiesevaluating other multimedia interventions (e.g., Zabinski et al., 2001).Respondents indicate whether the item has been true over the previous 28days. The current study used the Shape Concerns, Weight Concerns, andRestraint subscales. The Eating Concern subscale was not included, asFairburn and Beglin (1994) do not deem this a key subscale. A global scoreof the three subscales is also reported, following the scoring suggested byFairburn and Beglin (1994). In addition, the EDE-Q behavioral items thatreflect the symptoms of AN and BN were examined, including episodes ofbinge eating, episodes of overeating, use of compensatory behaviors (vom-iting, laxatives, diuretics), and excessive exercise. Respondents answeredthese items using a yes/no format. We decided against creating a compositescore of these behavioral items on the basis of their poor intercorrelationswhich, at baseline, ranged from a mere .16 to .26. The EDE-Q has goodpsychometric properties (Cronbachs alphas ranged from .67 to .90; Luce& Crowther, 1999) and corresponds well with the EDE-Interview (Fairburn

& Beglin, 1994). Because of the time frame of the EDE-Q (past 28 days),the EDE-Q was administered only at baseline and the 3-month follow-up.

Program satisfaction. Participants also answered a brief satisfactionquestionnaire, developed by the research team, after completion of theprogram. This 28-item questionnaire asked participants to respond on a7-point Likert scale (1 strongly disagree, 7 strongly agree) to a seriesof questions that centered on program design and content, relevance,usefulness, and overall satisfaction.

Efficacy Study ProcedureOf the 827 students who completed the screening measure, 110 (13.3%)

women were screened out of the possible study sample (see Figure 1). Ofthese, 107 (12.9%) women had scores on the Q-EDD suggesting thepresence of an eating disorder, and an additional 3 students indicated theywere in treatment for an eating disorder but did not currently score aseating disordered on the Q-EDD. We stratified the 717 eligible women intoat risk (n 288) and low risk (n 429) groups, and we then divided theminto the following groups on the basis of Q-EDD categorization andethnicity: low-risk nonminority, at-risk nonminority, low-risk minority,and at-risk minority. We then randomly selected those to be called forparticipation in the study, so that 75% were from the two nonminoritygroups, and 25% were from the two ethnic minority groups. After partic-ipants were selected across the four groups, we randomized them tocondition (FMA or control) using a computerized urn randomization pro-gram (Wei, 1978); this program took into account ethnicity, age, and risk statusin assigning condition. Minimal participant attrition resulted in a sample sizeof 231 for all completed assessments (i.e., 97.9% response rate).

Participants were contacted, and an appointment was set up. For Session 1,participants completed the baseline assessment battery (knowledge test,SATAQ, EDE-Q) and then participated in the intervention or control condi-tion. The RA who administered the intervention (or control) was blind to therisk status of the participant. For intervention participants, the RA oriented theparticipants to a laptop computer for use of the FMA program, and participantsproceeded with the program independently for 1 hr. Participants in the controlcondition underwent the same procedure but instead viewed two videos (30min each) at each of two sessions, as a control for time and attention thatparticipants had in the intervention condition. The videos were general videoson womens and/or gender issues (e.g., menopause, sex differences in infancy),and participants were not led to believe that these videos would improve theirbody image or eating habits. The RA remained near the lab throughout thesession to affirm that the participant completed FMA (or the videos), and

Table 1Study Population Characteristics by Randomly Assigned Condition

Characteristic

FMA (n 118) Control (n 118) All (N 236)

n % M SD n % M SD n % M SD

RaceWhite 85 72.0 88 74.6 173 73.3Asian 8 6.8 8 6.8 16 6.8Black 15 12.7 12 10.2 27 11.4Latino/Hispanic 5 4.2 2 1.7 7 3.0Others 5 4.2 8 6.8 13 5.5

Age 18.2 0.4 18.2 0.4 18.2 0.4Baseline body mass index 23.7 3.9 23.2 3.6 23.4 3.8Family income

Low (less than $25,000) 10 8.9 4 3.5 14 6.2Middle ($25,000$74,999) 46 41.1 49 42.6 95 41.9High ($75,000 or more) 56 50.0 62 53.9 118 52.0

Athletic involvementInvolved 7 5.9 7 5.9 14 5.9Not involved 111 94.1 108 94.1 219 94.1

Note. Values vary slightly because of missing data. FMA Food, Mood, and Attitude program.

570 FRANKO ET AL.

participants were paid $20 after completion of the first session. The secondsession was scheduled for 12 weeks later, during which participants com-pleted the 2nd hour of FMA or watched an additional two 30-min videos. Afterthe second session, in which they completed FMA or watched two videos, allparticipants were administered the postintervention assessment battery andwere paid $40. One week before the 3-month follow-up date, participants weremailed a packet containing the study measures and were paid $65 after thepacket was returned. The study concluded in summer 2002.

InterventionDevelopment of FMA. FMA was developed by use of a two-step

process. First, the following 2-hr focus groups were conducted: five with

44 undergraduate women from four universities; two with 9 undergraduatemen from two universities; four with 34 staff members from four universityhealth and student services, and one with 10 experts in the field of eatingdisorders. In the focus groups, participants were asked to discuss typicaleating patterns and diet-related concerns of college women as well as theirunderstanding of what constitutes a problematic eating habit. Partici-pants were also presented with ideas for various components of FMA toget their opinions on design and structure. Finally, group memberswrote down 10 questions that they thought would be important for theprogram to answer. After these data were collected, the results werediscussed, and relevant literature was reviewed in weekly meetings thatincluded all coauthors; program content and design was developed overthe course of a year.

Figure 1. Flowchart based on Consolidation Standards of Reporting Trials (CONSORT) guidelines (Moher,Schulz, & Altman, 2001). Q-EDD Questionnaire for Eating Disorder Diagnoses.

571REDUCING RISK FOR EATING DISORDERS

As noted earlier, the design of FMA was based on the dual-pathwaymodel (Stice et al., 1996), cognitivebehavioral theory, interpersonal/relational theory, and other theories of prevention. For each of the five riskfactors outlined in Stice et al.s (1996) model, specific program compo-nents were designed to provide both information and an interactive expe-rience centered on learning about and reflecting on that risk factor. Inaddition, program components that addressed interpersonal issues (e.g.,family, peers) and cognitive behavior strategies (e.g., recognizing errors inthinking) were included throughout the program. Finally, relevant preven-tion theory topics were incorporated throughout the program, includingreducing risk factors, increasing coping skills, and examination of conse-quences for harm reduction.

Description of FMA. At the start of the program, the user is asked toserve as a peer counselor to assist the virtual university in addressing theproblem of disordered eating on campus. The peer counselor (the researchparticipant) then meets Jen, who is thin and has some medical issuesrelated to her body weight. She also is introduced to Kate, an athlete whoovereats and overexercises, and Naomi, who is depressed and lonely andbinges. It is important to note that the user does not see what these threewomen look like until the very end of the program, so that the possibilitythat the user would be influenced by their appearance is eliminated. Eachof the three young women portrayed in the program has kept a scrapbookof information that covers various aspects of her life at college. Eachscrapbook entry provides psychoeducational information and interactivetools to educate and develop participants skills in each of the five theo-retically based risk factor areas and theoretically relevant topics (seeAppendix).

In using FMA, participants must explore each aspect of all three scrap-books (i.e., one for each woman portrayed in the program). The program-ming of the CD-ROM is such that a participant cannot skip content; shemust complete each module before she is allowed to go on to the next. Thepresence of the RA in the room ensured that the participant was in factusing FMA for the full session, as the RA would occasionally check on theprogress of the participant. Each scrapbook contains multiple modules, andin each module, didactic information is presented first, either in text,voice-over, or video. Then, an interactive tool appears, which is designedto get the participant involved in actively learning and thinking about thecontent in that particular module. To take but one example (out of 52 thatoccur in the program), one module (which addresses media images and thethin-ideal internalization) first presents information on the ideal body shapeespoused by the media for women over the past 5 decades. The psycho-educational point is that what is viewed as beautiful has changed over timeand is a construction by society of what is most appealing with regard towomens body shape and size. The participant is then provided fivesilhouettes of a woman who characterizes the ideal body shape for a givendecade and is asked to match the shape to the decade (e.g., MarilynMonroes silhouette for the 1950s, Twiggys for the 1960s). The programprovides feedback as to whether the choice is correct, as well as anexplanation of what was viewed as the beauty ideal of that time. Forexample, when the user clicks on the Marilyn Monroe silhouette, a pictureof her appears and the voice-over states,

At five and a half feet and 140 pounds, Marilyn Monroe was consid-ered the epitome of beauty in the 1950s. She is closest to the size ofthe average American woman of 2001. Monroe was one of the firstwomen to be considered a sex symbol. Women tried to imitate herglamour and sex appeal.

Similarly, when the user clicks on Twiggys silhouette, her picture appearsaccompanied by the following information,

At five foot seven and 91 pounds, women made themselves sick tryingto reach the impossible ideal of the famous fashion model Twiggy.Women all over the world cut their hair in Twiggys boyish style andimitated her wardrobe, which required women to be extremely thin inorder to wear those types of clothes.

Finally, the user is given information that can assist in combating (i.e.,decreasing internalization of) the current thin ideal:

Notice these women from various decades looked very different, yetall were considered beautiful. Societys idea of beauty constantlychanges over time, and women go to great lengths to look like theideal no matter what the consequences. There are habits that somewomen have that make them feel bad about themselves. Click here toget some tips on how to avoid them: Dont compare other womensbodies to your own; dont criticize or envy other women; wearclothing that fits comfortably and well; get rid of too-small clothesthat make you feel bad; use cosmetics and fragrances you love; get thesleep you need; eat a variety of healthy foods; read a variety ofmagazinesnews, travel, or hobbies; be aware of unhealthy bodymessages in fashion magazines.

At the end of the program, the participant is asked to provide feedback andtreatment suggestions (via a multiple-choice format) to each of the threehypothetical students in the program.

Data Analysis

We conducted preliminary analyses to assess the effects of potentialmoderators of outcomeincluding campus, age, ethnicity, marital status,socioeconomic status, athletic experience, and academic major (or ex-pected major)using a conservative Type I error of 0.10 to identifypossible covariates. Preliminary analyses to verify that our at-risk andlow-risk groups differed as expected were also conducted.

To examine our hypothesis predicting a significant three-way interactionof Time Condition Screening Status, we used a repeated measuresanalysis of variance (ANOVA) in an intent-to-treat design. The initialanalysis conducted was an overall repeated measures ANOVA with twofactors (condition and risk status at screening) at two levels (FMA vs.control and at risk vs. low risk of developing an eating disorder). Whereour hypothesis was confirmed (i.e., a significant three-way interaction wasdetected), we then conducted separate repeated measures ANOVAs foreach risk group. In all analyses, baseline levels of outcome measures werecontrolled.

Tests of significance were two-sided and performed at the .05 level.As an estimate of effect size in the repeated measures analysis, we presentthe partial eta squared (partial 2), which describes the proportion ofvariance accounted for in the sample by a given factor for all significantinteractions.

In addition to the analyses we conducted to examine our main hypoth-esis, we used exploratory multiple logistic regression analyses to examinegroup differences in the occurrence of problematic eating behaviors (di-chotomously coded as present or not present), controlling for reportedbaseline behaviors. Because not all women reported problematic eatingbehaviors at baseline, this remained an exploratory analysis to verify thatbehaviors did not increase with the intervention. Note, because of thepaucity of data on purging behaviors (e.g., vomiting, laxatives, diuretics),these categories were combined for analyses. However, these behaviorswere infrequent, creating wide confidence intervals. Odds ratios (OR), 95%confidence intervals (95% CI) and p values express these multivariaterelationships.

Results

Baseline Characteristics

Demographic characteristics of the efficacy study participantscan be found in Table 1. Bivariate analyses revealed no differencesin the study samples from the two university sites. All primaryoutcome baseline measurements were similar between the exper-imental and control groups.

572 FRANKO ET AL.

Comparing the sociodemographics of at-risk participants withthat of low-risk study participants, two otherwise indistinguishablegroups emerged. The at-risk and low-risk women were of similarage, body mass index, race, and income. They participated inathletics in an equivalent fashion and were represented equally atboth universities. Their baseline knowledge of disordered eatingalso was comparable. However, all other baseline outcome mea-sures were significantly different from one another ( p .005).At-risk women were differentiated from their low-risk peers bytheir greater recognition and internalization of societal influenceon attitudes toward their appearance, more restrained eating ten-dencies, greater concerns about their weight and shape, and morefrequent disordered eating behaviors, including excessive eating,exercising, and purging. The observed differences justify our risk-specific study design and analyses.

Hypothesis 1a: KnowledgeTables 2 and 3 display the means and standard deviations for the

intervention and control groups by risk status at each assessment.Repeated measures ANOVA revealed statistically significant gainsin knowledge for the FMA group over time relative to the controlgroup, F(1, 227) 22.27, p .001, partial 2 .09. The two-wayinteraction of knowledge by screening status was also significant,F(1, 227) 4.40, p .037; linear, partial 2 .02), suggestingthat at-risk women (regardless of condition) increased their knowl-edge more over the study period than did low-risk women. Nothree-way interaction (i.e., Time Condition Risk Status) wasobserved, because of lack of differences at Baseline Risk Status.

Hypothesis 1b: Awareness and Internalization of theSociocultural Ideal

On both subscales of the SATAQ (Awareness and Internaliza-tion), the FMA group showed greater improvements than did thecontrol group, F(1, 223) 4.67, p .032, partial 2 .02 forAwareness; F(1, 218) 10.02, p .002, partial 2 .04 forInternalization. On the Internalization subscale, a significant three-

way interaction (Time Condition Risk Status) revealed thatat-risk women who used FMA experienced the most change, F(1,218) 4.54, p .034, partial 2 .02, whereas at-risk women inthe control group experienced no significant change on the Inter-nalization subscale. In contrast, the three-way interaction of theAwareness Subscale Condition Risk Status was notsignificant.

Hypothesis 1c: Weight Concerns, Shape Concerns, andRestraint

Looking at the change in EDE-Q subscales between the twoobservations from baseline to follow up, significant three-wayinteractions were found for the Shape Concerns, F(1, 227) 5.01,p .03, and Weight Concerns subscales, F(1, 227) 6.64, p .01, again indicating that the at-risk FMA group improved relativeto the at-risk control group. A general trend to decrease restrainedeating behaviors, as evidenced by scores on the EDE-Q Restraintsubscale, was observed over the study period. Nevertheless, re-ductions were similar in both conditions and both risk groups (i.e.,no two- or three-way interactions were found). For the EDE globalscore, a significant three-way interaction was detected, F(1,227) 6.09, p .01, suggesting that the at-risk FMA groupreduced the overall severity of the eating disorder psychopathol-ogy in comparison with their at-risk control group counterparts.

A general pattern of the significant three-way interactions canbe detected in Figures 2A, 2B, and 2C. These figures visuallydepict our initial observations that the at-risk and low-risk womenrepresent very different groups with respect to these variables. Inaddition, women in the low-risk groups did not evidence signifi-cant change on these variables. Rather, the change that does occuris evident in the FMA at-risk group only.

Exploring the Significant Three-Way Interactions by RiskStatus

Where significant three-way interactions in our repeated mea-sures ANOVA were detected (i.e., SATAQ Internalization,

Table 2Means and Standard Deviations for the Knowledge and Attitudes Measures, by Condition and Risk Status at Screening

Outcome

FMA group (n 116) Control group (n 115)

BaselinePost-

intervention Follow-up BaselinePost-

intervention Follow-up

M SD M SD M SD M SD M SD M SD

KnowledgeAll 7.8 2.3 10.3 2.6 9.7 2.7* 7.5 2.3 7.8 2.6 8.0 2.7At risk (n 112) 7.8 2.7 10.6 2.8 10.3 3.3 7.3 2.3 7.9 2.6 8.1 2.3Low risk (n 119) 7.7 2.1 10.0 2.6 9.3 2.5 7.6 2.2 7.6 2.5 7.9 2.6

SATAQ: AwarenessAll 20.5 3.8 20.4 3.8 21.3 3.5* 21.2 3.8 20.8 3.8 20.9 3.5At risk 21.7 3.7 21.5 3.8 22.1 3.6 21.4 3.5 21.1 3.9 21.2 3.6Low risk 19.5 3.9 19.4 3.3 20.6 3.2 21.0 3.6 20.4 4.0 20.6 3.5

SATAQ: InternalizationAll 26.0 7.8 25.0 8.6 24.7 8.8* 25.4 8.6 25.3 8.6 26.4 8.4At risk 28.9 7.2 27.9 8.4 26.1 8.7 27.3 8.2 27.4 7.2 28.5 8.0Low risk 23.7 7.7 22.8 8.0 23.6 8.8 23.4 8.8 23.2 9.1 24.2 8.4

Note. FMA Food, Mood, and Attitude program; SATAQ Sociocultural Attitudes Toward Appearance Questionnaire.* p .05 (two-way interaction, Time Condition). p .05 (three-way interaction, Time Condition Screening Status).

573REDUCING RISK FOR EATING DISORDERS

EDE-Q Shape Concerns, EDE-Q Weight Concerns, EDE-Qglobal), we conducted separate repeated measures ANOVAs forthe two risk groups. Table 4 presents these outcomes in the twoseparate analyses (one for at-risk women and one for low-riskwomen).

For the at-risk women, FMA produced significant change on theInternalization subscale, F(1, 105) 13.25, p .001. As Figure2A depicts, and the partial 2 effect sizes (see Table 4) underscore,the women at risk for disordered eating who used FMA experi-enced a strong decrease in the internalization of negative socialand cultural attitudes toward appearances, whereas at-risk womenin the control group experienced no change in their internalizationof these negative attitudes. Likewise, attitudes captured by theEDE-Q (Shape Concerns and Weight Concerns) also decreasedamong the at-risk women who used FMA. Shape Concerns amongFMA participants decreased more over time than among theircounterparts in the control group, F(1, 110) 7.05, p .01, as didWeight Concerns among the same FMA at-risk women, F(1,110) 5.70, p .02.

Turning to the separate analyses for the low-risk women, as canbe seen in Table 4, no condition-related differences in scores wereevident for the SATAQ Internalization subscale (i.e., no deterio-ration of scores or increased internalization occurred, regardless ofcondition). Likewise, neither the EDE-Q Shape Concerns andWeight Concerns subscales nor the global score increased overtime for low-risk women in the FMA or control conditions. Inaddition, all of the effect sizes that tested for iatrogenic effects (asestimated by partial eta squared values) were equal to or less than.01.

Exploratory Analysis of Disordered Eating BehaviorsIn an exploratory analysis, we examined several behavioral

items from the diagnostic items on the EDE-Q as a measure of

disordered eating behaviors. The following analyses compared theintervention group with the control group, without regard to riskstatus, because there were too few women in each risk group whoreported these behaviors (a natural consequence of screening outwomen with diagnosable eating disorders). A multiple logisticregression analysis taking their baseline behavior into accountindicated that women in the experimental group were more likelyto decrease their use of excessive purging methods (i.e., any use ofvomiting, laxative, diuretics; OR .27, 95% CI 0.080.87, p.03) at follow-up, compared with the controls. This effect of FMAwas also detected in the decline in reported overeating at follow-up(OR .33, 95% CI 0.140.79, p .01). Finally, a trend ofdecreasing excessive exercise behaviors favored FMA participa-tion (OR .60, 95% CI 0.331.10, p .098).

Program Evaluation

The satisfaction survey conducted among students using FMArevealed extremely high ratings for overall satisfaction: 97% ofstudents were very satisfied to extremely satisfied with theprogram. Most students found FMA to be an effective resource foreducation on disordered eating (90%); 82% found FMA useful forgetting information if they were concerned about a friends dietingbehaviors; and 89% believed that other college women would findFMA useful. Compared with the typical ways of educating stu-dents about eating disorders, 82% found the FMA program moreeffective.

Discussion

Consistent with our hypothesis, both at-risk and low-risk par-ticipants in the FMA group increased their knowledge relative toparticipants in the control group. Further, as predicted, significantthree-way interactions were found for measures of internalization

Table 3Means and Standard Deviations on the Eating Disorder Examination Questionnaire (EDE-Q)Subscales, by Condition and Risk Status at Screening

Outcome

FMA group (n 116) Control group (n 115)

Baseline Follow-up Baseline Follow-up

M SD M SD M SD M SD

EDE-Q: Shape ConcernsAll 2.7 1.5 2.3* 1.5 2.7 1.5 2.6 1.5At risk (n 112) 3.2 1.4 2.6 1.3 3.3 1.6 3.1 1.6Low risk (n 119) 2.1 1.4 2.1 1.4 2.2 1.3 2.1 1.1

EDE-Q: Weight ConcernsAll 2.4 1.4 2.1* 1.4 2.4 1.6 2.2 1.5At risk 2.9 1.3 2.3 1.3 2.9 1.6 2.7 1.5Low risk 1.8 1.3 1.9 1.4 1.9 1.5 1.7 1.2

EDE-Q: RestraintAll 1.9 1.4 1.6 1.4 1.8 1.4 1.7 1.4At risk 2.4 1.4 1.8 1.4 2.4 1.3 2.2 1.6Low risk 1.4 1.2 1.3 1.3 1.3 1.2 1.2 1.2

EDE-Q: Global scoreAll 2.3 1.4 2.3 1.3 2.3 1.4 2.6 1.3At risk 2.8 1.8 2.6 1.8 2.9 1.7 3.1 1.7Low risk 1.8 1.7 2.1 1.7 1.8 1.8 2.0 1.7

Note. FMA Food, Mood, and Attitude program.* p .05 (two-way interaction, Time Condition). p .05 (three-way interaction, Time Condition Screening Status).

574 FRANKO ET AL.

Figure 2. Graphical displays of significant three-way interactions from the overall analysis: SocioculturalAttitudes Toward Appearance Questionnaire (SATAQ) Internalization subscale (A), Eating Disorder Examina-tion Questionnaire (EDE-Q) Shape Concerns subscale (B), and EDE-Q Weight Concerns subscale (C).

575REDUCING RISK FOR EATING DISORDERS

of the sociocultural ideal, shape concerns, and weight concerns.For the measures awareness of sociocultural ideals and restraint,the interaction of Time Condition Risk Status was notsignificant. Instead, we found that relative to the control condition,FMA positively affected awareness in 1st-year college womenwithout regard to risk status. Although restrained eating behaviorswere unexpectedly unaffected by FMA, exploratory analyses ofreported overeating, excessive exercise, and purging behaviors atfollow-up suggested that FMA had a positive effect with regard tothese behaviors.

Results suggest that FMA may be safe and effective for bothat-risk and low-risk college women. Programs that address bothgroups were recently recommended by Abascal, Brown, Winzel-berg, Dev, and Taylor (2004). More specifically, for the at-riskgroup, internalization of sociocultural attitudes about thinness, aswell as shape concerns, weight concerns, and the global score ofthe EDE were positively affected by the intervention. In otherwords, these risk factors for the development for eating disorderswere minimized in an at-risk group exposed to FMA. In addition,as expected, women of low risk exposed to the FMA program,already presumably evidencing positive attitudes and behaviors,did not change on these variables; however, theyand their at-riskcounterpartsdid acquire greater knowledge about risk factorsthan did women in the control group. This result emphasizes thatFMA can serve as a safe, educative prevention tool for women atlow risk of developing eating disorders without inducing iatro-genic effects. The combination of both universal and selectiveprevention in a single program suggests that FMA can be usedacross a wide range of risk profiles and has applicability in avariety of settings.

Our findings regarding increased knowledge and greater aware-ness in the FMA group, relative to control participants, are worthfurther discussion. In this efficacy study, both at-risk and low-riskparticipants in the intervention group increased their knowledgeover the course of the study, suggesting that regardless of riskstatus, participants learned about risk factors for eating disorderswhen using FMA. Similarly, regardless of risk status, awareness ofsociocultural ideals about thinness increased in the FMA condi-tion. These findings can be considered positive in terms of thepreventative effect of FMA. Specifically, research has shown thataugmenting knowledge and awareness can increase readiness tochange for a variety of health-related behaviors (McConnaughy,

Prochaska, & Velicer, 1983; Molaison, 2002), and thus, suchincreased knowledge could benefit both at-risk college womenwho have not yet pursued treatment (Geller, 2002) as well aslow-risk women who simply want to know more about eatingdisorder risk.

Our research appears to be the first prevention study to findpositive intervention effects without the benefit of a leader orfacilitator. Previous work either has involved facilitation (Franko,1998; Neumark-Sztainer, Butler, & Palti, 1995; Paxton, 1993;Steiner-Adair et al., 2002) or has used a facilitated multimediaformat (Winzelberg et al., 1998, 2000). As a stand-alone program,FMA is less resource intensive and potentially more cost-effective.FMA also may be of interest to students who may not be ready toseek help (e.g., go for counseling, join a chat room) because theprogram can be completed independently and anonymously. In thisway, FMA might be a useful tool for counseling centers or campushealth services or even as part of an orientation tool kit, in whichstudents could use the program in a confidential manner withoutassistance. Furthermore, because the program is not Internet based,privacy issues are less problematic (i.e., the data are not availableon an accessible server).

There are both strengths and limitations to this study. Strengthsinclude the relatively large sample, low attrition rate, the ease ofuse and short completion time of the program, and the clearlydefined risk categories. An additional strength was that partici-pants came from geographically diverse groups, with no site dif-ferences found on any of the baseline or outcome measures,suggesting that the program has some generalizability across dif-ferent areas of the United States. In addition, our 27% minoritysample is one of the largest in prevention studies to date. We alsohad a high recruitment and completion rate, likely related to thecompensation received by participants. Finally, we used an atten-tion control group to control for the effect of participation. Nev-ertheless, this attention control group had its limits in that it did notallow us to rule out demand characteristics or expectancy effects asalternative explanations for the intervention effects. Indeed, be-cause only the intervention participants received any content abouteating disorders, they may have experienced implicit pressure toreport decreases in eating disorder attitudes and behaviors that thecontrol group did not experience. Additional limitations includethe self-report nature of the data and the relatively short follow-upperiod. Also, the relatively higher number of women screened outby the Q-EDD as having an ED than would be expected on thebasis of epidemiological studies may have been a function ofselection bias; those with eating disorders may have been morelikely to choose to participate in the screening phase of the studyon the basis of its advertised content area (i.e., womens healthissues).

Future studies are needed to determine whether gains can besustained over a longer period and whether FMA is superior whencompared with other prevention programs. Additional researchwill be useful in determining which components of FMA weremost likely to be related to outcome and whether FMA preventsthe onset of clinical eating disorders over time. It would be ofinterest to combine this program with some face-to-face interac-tion, either with professionals or with peers, or to add boostersessions to see whether the obtained program effects might beenhanced. Finally, the efficacy of this program for specific popu-lations (e.g., athletes, advanced college students) remains to beexamined.

Table 4Estimates of Effect for Food, Mood, and Attitude by Risk StatusGroup for Outcomes

Group F p Partial 2

At risk (n 112)SATAQ: Internalization 13.25 .001 .11EDE-Q: Shape Concerns 7.05 .01 .06EDE-Q: Weight Concerns 5.70 .02 .05EDE-Q: Global score 8.24 .005 .07

Low risk (n 119)SATAQ: Internalization 0.57 .45 .005EDE-Q: Shape Concerns 0.09 .76 .001EDE-Q: Weight Concerns 1.44 .23 .012EDE-Q: Global score 0.26 .61 .002

Note. SATAQ Sociocultural Attitudes Toward Appearance Question-naire; EDE-Q Eating Disorder Examination Questionnaire.

576 FRANKO ET AL.

In conclusion, FMA appears to be a brief and effective multi-media program for both at-risk and low-risk college women. FMAis likely to have wide applicability on university campuses in theefforts to stem the tide of increasing numbers of eating disorders inthe college population.

ReferencesAbascal, L., Brown, J. B., Winzelberg, A. J., Dev, P., & Taylor, C. B.

(2004). Combining universal and targeted prevention for school-basedeating disorder programs. International Journal of Eating Disorders, 35,19.

American Psychiatric Association. (1994). Diagnostic and Statistical Man-ual of Mental Disorders (4th ed.). Washington, DC: Author.

Bearman, S. K., Stice, E., & Chase, A. (2003). Evaluation of an interven-tion targeting both depressive and bulimic pathology: A randomizedprevention trial. Behavior Therapy, 34, 277293.

Budman, S. H. (2000). Behavioral health care dot-com and beyond:Computer-mediated communications in mental health and substanceabuse treatment. American Psychologist, 55, 12901300.

Celio, A. A., Winzelberg, A. J., Wilfley, D. E., Eppstein, D., Springer,E. A., Dev, P., et al. (2000). Reducing risk factors for eating disorders:Comparison of an Internet and classroom delivered psychoeducationalprogram. Journal of Consulting and Clinical Psychology, 68, 650657.

Cowen, E. L. (2000). Psychological wellness: Some hopes for the future.In D. Cicchetti, J. Rappaport, I. Sandler, & R. P. Weissberg (Eds.), Thepromotion of wellness in children and adolescents (pp. 477503). Wash-ington, DC: Child Welfare League of America.

Drewnowski, A., Yee, D. K., Kurth, C. L., & Krahn, D. D. (1994). Eatingpathology and DSMIIIR: A continuum of bulimic behaviors. Ameri-can Journal of Psychiatry, 151, 12171219.

Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders:Interview or self-report questionnaire. International Journal of EatingDisorders, 16, 363370.

Fairburn, C. G., & Cooper, Z. (1993). The Eating Disorder Examination(12th ed.). In C. Fairburn & T. Wilson (Eds.), Binge eating: Nature,assessment, and treatment (pp. 317332). New York: Guilford Press.

Franko, D. L. (1998). Secondary prevention in college students at risk foreating disorders. Eating Disorders: The Journal of Treatment and Pre-vention, 6, 2940.

Franko, D. L., & Orosan-Weine, P. (1998). The prevention of eatingdisorders: Empirical, methodological and conceptual considerations.Clinical Psychology: Science and Practice, 5, 459477.

Geller, J. (2002). Estimating readiness for change in anorexia nervosa:Comparing clients, clinicians, and research assessors. InternationalJournal of Eating Disorders, 31, 251260.

Heatherton, T. F., Mahamedi, F., Striepe, M., Field, A. E., & Keel, P. K.(1997). A 10-year longitudinal study of body weight, dieting, and eatingdisorder symptoms. Journal of Abnormal Psychology, 106, 117125.

Heinberg, L. J., Thompson, J. K., & Stormer, S. (1995). Development andvalidation of the Sociocultural Attitudes Toward Appearance Question-naire. International Journal of Eating Disorders, 17, 8189.

Hill, A. J. (2002). Prevalence and demographics of dieting. In C. G.Fairburn & K. D. Brownell (Eds.), Eating disorders and obesity: Acomprehensive handbook (pp. 8084). New York: Guilford Press.

Kaminski, P. L., & McNamara, K. (1996). A treatment for college womenat risk for bulimia: A controlled evaluation. Journal of Counseling andDevelopment, 74, 288294.

Levine, M. P., & Piran, N. (2001). The prevention of eating disorders:Toward a participatory ecology of knowledge, action, and advocacy. InR. H. Striegel-Moore & L. Smolak (Eds.), Eating disorders: Innovativedirections in research and practice (pp. 233254). Washington, DC:American Psychological Association.

Luce, K. H., & Crowther, J. H. (1999). The reliability of the EatingDisorder ExaminationSelf-Report Questionnaire Version (EDE-Q).International Journal of Eating Disorders, 25, 349351.

Mann, T., Nolen-Hoeksema, S., Huang, K., Burgard, D., Wright, A., &Hanson, K. (1997). Are two interventions worse than none? Joint,primary, and secondary prevention of eating disorders in college fe-males. Health Psychology, 16, 215225.

Marlatt, G. A. (1995). Preventing alcohol abuse in college students: Aharm-reduction approach. In G. M. Boyd, J. Howard, & R. A. Zucker(Eds.), Alcohol problems among adolescents: Current directions inprevention research (pp. 147172). Northvale, NJ: Erlbaum.

Martz, D., & Bazzini, D. (1999). Eating disorders prevention programmingmay be failing: Evaluation of two one-shot programs. Journal of CollegeStudent Development, 40, 3242.

McConnaughy, E. A., Prochaska, J. O., & Velicer, W. F. (1983). Stages ofchange in psychotherapy: Measurement and sample profile. Psychother-apy, 20, 368375.

Miller, W. R., Brown, J. M., Simpson, T. L., Handmaker, N. S., Bien,T. H., Luckie, L. F., et al. (1995). What works: A methodologicalanalysis of the alcohol treatment outcome literature. In R. K. Hester &W. R. Miller (Eds.), Handbook of alcoholism treatment approaches:Effective alternatives. (2nd ed., pp. 1244). Boston: Allyn & Bacon.

Mintz, L. B., OHalloran, S. M., Mulholland, A. M., & Schneider, P. A.(1997). Questionnaire for eating disorder diagnoses: Reliability andvalidity of operationalizing DSMIV criteria into a self-report format.Journal of Counseling Psychology, 44, 6379.

Molaison, E. F. (2002). Stages of change in clinical nutrition practice.Nutrition Clinical Care, 5, 251257.

Moher, D., Schulz, K. F., & Altman, D. (2001). The CONSORT statement:Revised recommendations for improving the quality of reports ofparallel-group randomized trials. Journal of the American Medical As-sociation, 285, 19871991.

Mrazek, P. J., & Haggerty, R. J. (Eds.). (1994). Reducing risks for mentaldisorders: Frontiers for prevention intervention research (Report of theInstitute of Medicines Committee on Prevention of Mental Disorders).Washington, DC: National Academy Press.

Mutterperl, J. A., & Sanderson, C. A. (2002). Mind over matter: Internal-ization of the thinness norm as a moderator of responsiveness to normmisperception education in college women. Health Psychology, 21,519523.

Neumark-Sztainer, D., Butler, R., & Palti, H. (1995). Eating disturbancesamong adolescent girls: Evaluation of a school-based primary preven-tion program. Journal of Nutrition Education, 27, 2430.

Nicolino, J. C., Martz, D. M., & Curtin, L. (2001). Evaluation of acognitive-behavioral therapy intervention to improve body image anddecrease dieting in college women. Eating Behaviors, 2, 353362.

Paxton, S. J. (1993). A prevention program for disturbed eating and bodydissatisfaction in adolescent girls: A one year follow-up. Health Edu-cation and Research, 8, 4351.

Paxton, S. J., Schutz, H. K., Wertheim, E. H., & Muir, S. L. (1999).Friendship clique and peer influences on body image concerns, dietaryrestraint, extreme weight-loss behaviors, and binge eating in adolescentgirls. Journal of Abnormal Psychology, 108, 255266.

Perry, C. L. (1999). Creating health behavior change: How to developcommunity-wide programs for youth. Thousand Oaks, CA: Sage.

Smolak, L., Levine, M. P., & Gralen, S. (1993). The impact of puberty anddating on eating problems among adolescent girls. Journal of Youth andAdolescence, 22, 355368.

Stein, R. I., Saelens, B. E., Zoler-Dounchis, J., Lewczyk, C. M., Swenson,A. K., & Wilfley, D. E. (2001). Treatment of eating disorders in women.The Counseling Psychologist, 29, 695732.

Steiner-Adair, C., Sjostrom, L. A., Franko, D. L., Pai, S., Tucker, R.,Becker, A. E., et al. (2002). Primary prevention of risk factors for eatingdisorders in adolescent girls: Learning from practice. International Jour-nal of Eating Disorders, 32, 401411.

Stice, E. (2002). Risk and maintenance factors for eating pathology: Ameta-analytic review. Psychological Bulletin, 128, 825848.

Stice, E., & Agras, W. S. (1999). Subtyping bulimic women along dietary

577REDUCING RISK FOR EATING DISORDERS

restraint and negative affect dimensions. Journal of Consulting andClinical Psychology, 67, 460469.

Stice, E., Killen, J. D., Hayward, C., & Taylor, C. B. (1998). Age of onsetfor binge eating and purging during late adolescence: A 4-year survivalanalysis. Journal of Abnormal Psychology, 197, 671675.

Stice, E., Maxfield, J., & Wells, T. (2003). Adverse effects of socialpressure to be thin on young women: An experimental investigation ofthe effects of fat talk. International Journal of Eating Disorders, 34,108117.

Stice, E., Nemeroff, C., & Shaw, H. (1996). Test of the dual pathwaymodel of bulimia nervosa: Evidence for dietary restraint and affectregulation mechanisms. Journal of Social and Clinical Psychology, 15,340363.

Stice, E., Presnell, K., & Spangler, D. (2002). Risk factors for binge eatingonset in adolescent girls: A 2-year prospective investigation. HealthPsychology, 21, 131138.

Stice, E., & Ragan, J. (2002). A controlled evaluation of an eating distur-bance psychoeducational intervention. International Journal of EatingDisorders, 31, 159171.

Stice, E., & Shaw, H. (2004). Eating disorder prevention programs: Ameta-analytic review. Psychological Bulletin, 130, 206227.

Stice, E., Trost, A., & Chase, A. (2003). Healthy weight control anddissonance-based eating disorder prevention programs: Results from acontrolled trial. International Journal of Eating Disorders, 33, 1021.

Streigel-Moore, R. H., & Cachelin, F. M. (2001). Etiology of eatingdisorders in women. The Counseling Psychologist, 29, 635661.

Striegel-Moore, R. H., Silberstein, L. R., & Rodin, J. (1986). Toward anunderstanding of risk factors for bulimia. American Psychologist, 41,246263.

Tylka, T. L., & Subich, L. M. (1999). Exploring the construct validity of

the eating disorder continuum. Journal of Counseling Psychology, 46,268276.

Vohs, K., Bardone, A., Joiner, T., Abramson, L., & Heatherton, T. (1999).Perfectionism, perceived weight status, and self-esteem interact to pre-dict bulimic symptoms: A model of bulimic symptom development.Journal of Abnormal Psychology, 108, 695700.

Wei, L. J. (1978). An application of an urn model to the design ofsequential controlled trials. Journal of American Statistical Association,73, 559563.

Wilfley, D. E., Agras, W. S., Telch, C. F., Rossiter, E. M., Schneider, J. A.,Cole, A. G., et al. (1993). Group cognitivebehavioral therapy andgroup interpersonal psychotherapy for the nonpurging bulimic individ-ual: A controlled comparison. Journal of Consulting and Clinical Psy-chology, 61, 296305.

Winzelberg, A. J., Eppstein, D., Eldredge, K. L., Wilfley, D., Dasmaha-patra, R., Dev, P., et al. (2000). Effectiveness of an Internet-basedprogram for reducing risk factors for eating disorders. Journal of Con-sulting and Clinical Psychology, 68, 346350.

Winzelberg, A. J., Taylor, C. B., Altman, T. M., Eldredge, K. L., Dev, P.,& Constantinou, P. S. (1998). Evaluation of a computer-mediated eatingdisorder intervention program. International Journal of Eating Disor-ders, 24, 339349.

Zabinski, M. F., Celio, A. A., Wilfley, D. E., & Taylor, C. B. (2003).Prevention of eating disorders and obesity via the Internet. CognitiveBehaviour Therapy, 32, 137150.

Zabinski, M. F., Pung, M. A., Wilfley, D. E., Eppstein, D. L., Winzel-berg, A. J., Celio, A. A., et al. (2001). Reducing risk factors for eatingdisorders: Targeting at-risk women with a computerized psychoedu-cational program. International Journal of Eating Disorders, 29,401 408.

Appendix

Components of Food, Mood, and Attitude CD

Jen Kate Naomi

Time with boyfriend vs. friendsa Social interaction at mealsa Eating alone in cafeteriaaGet skinny before spring breaka Alcohol and foodb Missing friends from homeaEating aloneb Aggressive vs. assertive with peersa Not fitting inaFeeling guilty about foodb Perfectionism quiza Friends who are like youaUnder/overeating with stressa Cognitive distortionsa A body different from mediaaEating in secretb Ways of thinkinga Talking/lunch with MomaPerfectionisma Parent pressure to do wella Disappointed about partaMom critical about weighta,b Stress related to roommatesb What do guys like?aEmotional eatinga Traumatic/stressful eventsb Dos and donts about bodyaFood and mood diarya Different types of identityb Expectations of familyaLearning eating triggersa Ballet dancerdeath due to eating

disorderbStressed in new situationsa,b

Identity uncertaintyb Exercise if hurtbFitting inaDealing with stressb Choose mealsget nutrient infoa

Recognizing emotionsb Stress meterpast 12 monthsa,bResponses to stressa,bTalking directly to othersa

Food traya Personality typeType AbBone healthb Body is different than idealaMedia images 18901999a Power foodsaAltering bodiesa Freshman 15a

Complications of eating disorderbDieting/set pointa

a Addresses one of five risk factors (pressure to be thin, thin ideal internalization, body dissatisfaction, dieting,negative affect, or interpersonal/cognitivebehavioral theories. b Addresses theories of prevention (e.g., dis-ease specific, nonspecific stress, harm reduction).

578 FRANKO ET AL.