A Randomized Clinical Trial of an Identity Intervention Program for Women with Eating Disorders

Disturbances in the self-concept are broadly recognized as a psychosocial risk factor in the etiology of the eating disorders and as such, have been identified an important focus in the eating disorder psychotherapies (for examples see Fairburn, Cooper & Shafran, 2003; Wonderlich et al, 2008). Yet despite the centrality of self-concept disturbances in theoretical models of the disorders, theoretically derived and empirically supported interventions that target changes in the self-concept as the mechanism of eating disorder symptom change and health recovery are lacking. In this paper, we report results of a randomized trial of a new psychotherapy for the eating disorders of anorexia nervosa (AN) and bulimia nervosa (BN), referred to as the identity intervention program (IIP). IIP is based on an empirically supported model that focuses on impairments in organizational properties of the total collection of self-cognitions, referred to as self-schemas, as a key determinant of the maladaptive attitudes and behaviors that characterize the disorders. The IIP seeks to build new positive self-schemas that are distinct and separate from other conceptions of the self in memory as the means to promote recovery and health. The trial examined whether IIP in combination with nutritional counseling would be more effective than supportive psychotherapy and nutritional counseling in reducing the attitudinal and behavioral symptoms that characterize the eating disorders and improving functional health and psychological well-being.

Self-Schema Disturbance and the Eating Disorders: Theoretical Background and Empirical Support

The proposition that the eating disorders stem from fundamental disturbances in the process of identity development and the resultant collection of self-conceptions or identities is a basic tenet in a diverse array of eating disorder (ED) theories ranging from early psychoanalytic, cultural, and feminist approaches (Bruch, 1982; Bjorck, Clinton, Sohlberg, & Norring, 2007; Lamoureux & Bottorff, 2005; Tan, Hope, & Stewart, 2003; Vitousek & Ewald, 1993). Although theories differ with respect to the causes of the identity disturbances – ranging from genetically-determined temperaments (Strober, 1991); disturbances in parent-child relationship due to stressors, mental disorder or other illness(Wechselblatt, Gurnick & Simon, 2000); childhood abuse (Polivy & Herman, 2002) and gender-based cultural pressures that interfere with self-development (Malson, 1999), all posit that the relative absence of positive identities contributes to the compensatory focus on body weight as the primary source of self-definition.

Our work on disturbances in identity development associated with the eating disorders is based on the schema model. Self -schemas are highly elaborated and chronically accessible representations about the self that function to motivate and regulate behavior (Banting, Dimmock & Lay, 2009; Holloway, Wildrip & Ickes, 2009; Kendzierski); 1990). Results have shown that positively valenced self-schemas predict effective and reliable behaviors necessary for successful performance in the domain where as negative self-schemas are associated with anxiety, behavioral inhibition, low levels of involvement in the domain, and contextually dependent evaluations of the self (Feiring, Cleland, & Simon, 2010; Lips, 1995;Ledoux, Winterowd, Richardson & Clark, 2010).

In addition to the number of valenced self-schemas, the organization of self-schemas in memory also influences emotional and behavioral regulation (Dozois, Bieling, Patelis-Siotis, Hoar, Chudzik, McCabe, Westra, 2009; Chandra & Shadel, 2007). People vary in the extent to which self-schemas are stored in memory as separate units or linked together to form a tightly interrelated network. Highly interrelated schemas derive their meaning from each other, and therefore, function in an “all-or-nothing” fashion such that activation of one self-schema leads to activation of the related schemas in the network (Linville, 1987). A highly interrelated collection of self-schemas, despite the possibility of including many identities, reflects a high level of overlap in content. Therefore, a highly interrelated collection of self-schemas provides a more limited collection of functionally distinct schemas to guide information processing and regulate emotions and behaviors. Individuals with high interrelatedness among their self-schemas are more likely to: 1) respond to challenging social feedback with less effective coping strategies, 2) react to stressors with decreases in mood and self-esteem, and 3) experience physical illness in response to stress than persons who have many independent self-schemas available in memory (Brown & McConnell, 2009; Linville, 1985, McConnell, Strain, Brown, & Rydell, 2009).

Results of our previous work have shown that a self-concept comprised of a highly interrelated collection of few positive and many negative self-schemas contributes to the onset and persistence of eating disorder symptoms. In the first study, women with AN and BN had fewer positive and more negative self-schemas available in memory and higher interrelatedness among their self-schemas compared to healthy controls but no differences were found between the two ED groups. Further, the number of positive self-schemas negatively predicted eating disordered attitudes and behaviors and these effects were mediated through the fat self-schema. A high level of interrelatedness among the self-schemas had a direct influence on eating disordered behaviors and attitudes(Stein & Corte, 2007). These results were replicated other studies including a 12-month longitudinal study of college freshmen women that showed that the number of positive and negative self-schemas predicted fat self-schema scores (both measured at baseline)predicted increases in disordered eating behaviors 6 and 12 months later (Stein & Corte, 2008; Stein, Corte & Ronis, 2010).

Results of other recent studies converge to support our findings that negative self-schemas predict risk for eating disorders. In young female dieters with no current or past Axis I disorders, enhanced processing of negative self-beliefs previously shown to be associated with eating disorders predicted scores on the Eating Attitudes Test both cross-sectionally (Pringle, Harmer & Cooper, 2010a) and 12 months later (Pringle, Harmer, & Cooper, 2010b). Not surprisingly, enhanced processing of body weight and shape related self-beliefs also predicted scores on the Eating Attitudes Test, but this finding was much stronger for dieters who also showed enhanced processing of negative self-beliefs (Pringle Harmer & Cooper, 2010a).

Based on the eating disorder theoretical perspective and research findings described above, the Identity Intervention Program (IIP) was developed with a focus on modifying the array of self-schemas that comprise the self-concept as the primary means to promote recovery from the eating disorders.

Identity Intervention Program

The aim of the IIP is to alter the array of self-schemas that comprise the self-concept by fostering the development of new distinct and separate positive self-schemas. Since positive self-schemas function to motivate and direct goal directed behaviors (Banting, et al., 2009; Kendzierski, 1990), an increase in positive self-schemas should serve to increase involvement and investment in more diverse behavioral domains and decrease the singular maladaptive focus on body weight and shape. The Identity Intervention Program (IIP) is a 20-week individual and group psychotherapy program that focuses on identification of a personally meaningful and feasible future-oriented possible self-goal, and elaboration of that goal into an enduring positive self-schema that is distinct and separate from other self-schemas. The program was designed to translate theoretically based principles of self-schema organization and development (Kendzierski & Morganstein, 2009; Klein, Sherman & Loftus, 1996; Norman & Aron, 2003), into a cognitive-behavioral intervention program. Conceptually the IIP program is similar to the cognitive restructuring included in extended CBT (see Fairburn, 2008; Waller, Cordery, Corstorphine, Hinrichsen, Lawson, Mountford & Russell, 2007) in that the goal of the intervention is to increase the range of domains of behavioral involvement and positive experiences. IIP differs from the extended CBT in that the IIP is a systematic program that specifically focuses on the development of new cognitive structures (e. g. , possible selves and self-schemas) that serve as the cognitive source of emotional commitment to a behavioral domain (Cross & Markus, 1994; Strachan & Brawley, 2008) and have been shown to lead to reliable and consistent behavioral involvement (Allom & Mullan, 2011; Froming, Nasby & McManus, 1998; Sheeran & Abraham, 2003). Whereas CBT focuses on revising the underlying depressogenic cognitions including attributional bias and “persuading the individual to undertake behaviors” (Waller et al., 2007 page 249), IIP strives to foster development of new cognitive structures. IIP builds on the intervention program outline by Oyserman, Terry and Bybee (2002) in providing supportive interpersonal contexts and focused activities that facilitate development of conception of the self in the future. In addition, the treatment supports development of specific plans and strategies to achieve the identified future self-goal. Treatment sessions include a sequence of activities and task that provided the individual with opportunities to form a detailed vision for the future and to outline and practice activities needed to achieve the targeted future possible self. The content of the program is organized in six units that are briefly outlined in Table 1. Each unit has a specific focus and outcome: a) identification of a small collection of personally meaningful and feasible desired possible selves, b) prioritization of the desired possible selves as short and long-term goals, c) identification, elaboration, implementation and evaluation of strategies to achieve the desired possible self selected as the goal, d) identification of barriers that interfere with achieving the possible self goal, e) utilization of active problem solving strategies to overcome the barriers, and f) review and consolidation of the steps needed to achieve a new possible self goal and plans for termination. Each unit includes specific cognitive and behavioral intervention strategies to facilitate achievement of the unit goal. However, the therapist retains flexibility in specific strategies that can be used in a session. The unit foci and outcomes are sequential but movement back and forth across the units is allowed when issues emerge related to an earlier section. For example, a selected possible self-goal may be experienced as less meaningful or less feasible than anticipated once efforts to define implementation strategies are explored. In that case, moving back to the unit on possible self prioritizing and selecting would be warranted and acceptable. Progression through units is individually paced based on the participant's successful achievement of the unit outcome. The group component of the therapy was included to provide social context to enhance commitment and elaboration of the possible self-goal and related self-schema. One-hour group sessions were held weekly and followed the format and focus of the individual therapy program.

Table 1.

Unit 1 Orientation. Goals: 1) to create an understanding of the intervention program; 2) to create conceptions of the self in the future that reflect one’s personal interests, values, and talents. This unit consists of four modules that include establishing mutual expectation of patient and therapist roles in treatment, collecting background history, providing an overview of the rationale and methods of the intervention and fostering development of future-oriented hoped-for possible selves. Exercises in the last module include activities to encourage imagining about the future and recognizing personal interests and talents.

Unit 2: Selecting a Possible Self. Goal: to select a hoped-for possible self to work on during treatment (e.g., referred to as the target possible self). Six modules are designed to stimulate the development of personalized conceptions of selves one would like to be in the future and to identify one possible self that will serve as the focus for the duration of program. Activities include generating list of possible future selves, diary writing about the self one would like to be in 3 years, creating a visual arrangement of possible selves relative to “your center or what is most you” and “challenges you to grow but does not overwhelm you”.

Unit 3: Envisioning. Goal: to elaborate the target possible self. Four modules are designed to foster cognitive elaboration of the target possible. Activities encourage imagining both the steps needed to achieve the possible self goal and the self at the desired end-state. They including brainstorming, writing a story, creating a collage about the self I will be and “trying on the possible self for a day”.

Unit 4: Planning and Doing. Goals: 1) to break down the target possible self into smaller, more intermediate steps or goals and plan behaviors needed to achieve goals. Four modules seek to identify the personal strengths available and resources needed to achieve intermediate goals. Specific plans are formulated each week for activities needed to be completed to achieve the goal and mental rehearsal exercises are used to facilitate completion of the action plan.

Unit 5: Reflecting and Evaluating. Goal: to identify barriers that interferred with implementation of behavioral plan over the last and develop plans for overcoming barriers. Activities include systematic review of strength and barriers; evaluation of strategies and planning alterative approaches.

Unit 6: Taking Stock and Saying Goodbye. Goals: 1) to frame identity development as a life long process; 2) to support termination from therapy. Activities include a review of the treatment units and activities that were most helpful in promoting successful achievement of the target possible self and a review of other desired possible selves identified in the process that could be the focus of future efforts.

We conducted a randomized clinical trial to test the efficacy of the IIP compared to supportive psychotherapy in promoting recovery in women with an eating disorder. In keeping with recent work in which mental health is conceptualized as more than the absence of mental illness (Keyes, 2005. Westerhof & Keyes, 2010), in this study, the effects of the IIP program on psychological well-being and functional health are also examined. Supportive psychotherapy (SPI) (Walsh & Wilson, 1997) was selected as the control treatment for several reasons: 1) it is a manualized treatment for the eating disorders, 2) like IIP it does not include a focus on nutritional knowledge and distorted thinking about body weight and shape, hence we could standardize nutritional counseling across the two treatment conditions, and 3) it has been used as control condition in several eating disorder randomized treatment trials.

Standardized nutritional counseling based on the Health Eating: A Meal Planning System (Neuropsychiatric Research Institute, 1998) was delivered to both groups (e. g. IIP and SPI). The program was presented over 20 weeks and included 45-minute session with master’s prepared dietitians who were trained with the program. In keeping with the program, participants met weekly with the dietician for the first 8 weeks and bi-weekly thereafter.

This first trial of IIP was conducted with women with diagnosed AN or BN at threshold or subthreshold levels. The decision to include women with both disorders in this trial was based on recent literature that conceptualizes AN and BN as a single disorder (Fairburn, Cooper, Shafran, 2003; Wade, Bergin, Martin, Gillespie, & Fairburn, 2006) and findings from our earlier study that supports the influence of self-schema disturbances in symptoms of both disorders (Stein & Corte, 2007).

In this first trial of the IIP program, the following research hypothesis was tested: 1) Compared to women in the SPI group, women in the IIP group will have a greater decrease in ED symptoms and a greater increase in psychological health and well-being. These effects will be mediated through an increase in the number of positive self-schemas available in memory and a decrease in the overall level of interrelatedness among the self-schemas.

Method

Design

We conducted a pre-test post-test randomized experimental control group design, with post-tests one month post-intervention, 6 months post-intervention, and 12 months post-intervention. Data collection and interventions occurred during the period of 2002 and 2006. The 6 and 12 month post-intervention follow-ups are not addressed in this report. Data from this study have been published in a manuscript that addresses correlates of attrition in a randomized clinical trial for treatment of the eating disorders (Stein, Wing, Lewis & Raghunathan, 2011). There is no overlap with RTC results reported in this manuscript.

Participants and Recruitment

Participants were women between the ages of 18 and 35 years with diagnosed threshold or subthreshold anorexia nervosa or bulimia nervosa. They were recruited through provider referrals, community-based and Internet advertisements in the period between 2003 and 2005. The Midwestern urban communities in which recruitment occurred were in close proximity to a large state university and other smaller four-year and community colleges. Potential participants who contacted our research office were asked to complete a brief phone-screening interview that focused on eating disorder symptoms and current psychotropic medication use and psychotherapy treatment involvement. Women who appeared eligible were invited to complete a two-step eligibility assessment that included the Structured Clinical Interview for DSM-IV (First et al, 1997) and a physical health assessment. An information session was scheduled prior to the screening interview if desired by the potential participant. Study procedures including the random assignment to treatment were discussed with the potential participant before the informed consent process. Informed consent was either signed at the orientation meeting or prior to the start of the SCID interview.

To be eligible to participate in the study, women had to meet the following criteria: 1) 18 and 35 years of age and not pregnant, 2) full or subthreshold criteria for anorexia nervosa or bulimia nervosa, 3) no prescribed psychotropic medication for at least 2 weeks prior to screening, and, 4) involved in no other form of psychotherapy. Subthreshold AN and BN were defined based on DSM-IV EDNOS criteria. Subthreshold AN criteria included: 1) all criteria for AN were met except for amenorrhea for 3 consecutive months or 2) all criteria for AN were met except that despite significant weight loss, weight was not ≤ 85% of ideal, and 3) endorsement of two items developed by Strober related to levels of distress and dysfunction associated with weight concerns and eating (Strober, Freeman, Lampert, Diamond & Kaye, 2000). Subthreshold BN criteria included: 1) all criteria for BN were met except binge and compensatory behavior cycles occurred at an average rate of 1 time weekly for 3 months or 2) all criteria for BN were met except purging behaviors were in response to the ingestion of a small amount of food. Women were excluded from participation if they met any of the following criteria: 1) a level of symptom severity that required inpatient treatment, 2) suicidality, 3) any lifetime history of schizophrenia, other DSM-IV Axis I psychotic disorders or mental retardation, or 3) concurrent DSM-IV Axis I disorder at threshold level.

Participants went through a two-step assessment phase to determine eligibility to participate. The first phase consisted of the informed consent process followed by measurement of height and weight, completion of the Beck Depression Inventory as an initial suicide screen and the SCID. Three experienced clinicians who were trained in the administration of the SCID completed the diagnostic interviews. All SCID interviews and diagnoses were reviewed with the first author to confirm eligibility to participate. Those who met eligibility criteria based on the SCID returned for the second phase that included a physical assessment, blood laboratory studies and an EKG.

Random Assignment to Treatment Group

Women who met eligibility criteria based on SCID and physical assessment, were scheduled for baseline data collection that consisted of a battery of measures and a 21-day ecological momentary assessment (EMA). After completion of all baseline measures, an envelope with the random assignment was opened and the participant was scheduled for first treatment appointment. Randomization of the entire sample was accomplished prior to the recruitment of any participants. A stratified randomization method was employed, with strata defined by level and type of illness: (1) full-threshold AN (2) full-threshold BN (3) sub-threshold AN, and, (4) sub-threshold BN. Dieticians who delivered nutritional counseling and the research assistants who collected post intervention data were blind to the treatment status and had no other role on the project.

Sample Size

Sample size determination was based on power analysis using Cohen’s (1988) methods for a t-test as implemented in PASS software (Hintze, 1996) with refinements for Analysis of Covariance (Taylor & Innocenti, 1993), and estimates of standard deviations and pre-post correlations based on the preliminary studies. The sample size was selected to provide at least 80% power and 5% type-I error rate for detecting effects of the treatment on key dependent variables. It was calculated that the sample size of 75 per group provided 86% power to detect a medium-sized effect (i.e., a difference in means equal to .5 SD) using a t-test. This assured that practically important effects will be detected. With estimated attrition at each phase of the trial as 22%, the total of 246 referrals to the study were estimated.

Mediating Variables

Self-Schema Number, Valence and Interrelatedness

Zajonc's card-sorting task (1960) was used to measure the number of valenced self-schemas and interrelatedness.

Task 1: Participants were given a stack of 52 blank index cards labeled A through ZZ and asked to write down all descriptors that are important to how they think about themselves. Next, participants rated each descriptor according to: 1) degree of self-descriptiveness, 2) degree of importance to one’s self-description, 3) valence, positive, negative or neutral. In keeping with previous research on self-schemas by Markus and others (Kendzierski, 1990; Kendzierski & Whitaker, 1997; Markus, 1977; Markus, Hammill, & Sentis, 1987) descriptors that were rated highly self-descriptive (8–11 on an 11-point scale) and highly important (8–11) were identified as self-schemas. The number of positive (negative, neutral) self-schemas was computed by totaling the number of self-descriptors that met the criteria for a self-schema and were rated positive (negative, neutral).

Task 2: Participants were then asked to focus on each self-descriptor separately and to identify all other descriptors that would change if the targeted descriptor was "changed, absent or untrue of you. " Responses were used to form a dependency matrix such that when descriptor A_j causes a change in descriptor A_i, a value of 1 is assigned. The total dependence of an element was calculated by summing the row entries, and the total dependency of the schema was calculated by summing the dependencies across all characteristics. To compare interrelatedness across self-concepts that include varying numbers of descriptors, the measure of interrelatedness is normalized by dividing the total dependency by the total number of possible dependencies of the structure.

Validity of the Markus self-rating to identify self-schemas has been supported by a large collection of studies demonstrating predicted information processing and behavioral differences between schematic and aschematic participants (Kendzierski, 1990; Kendzierski & Whitaker, 1997; Markus, 1977; Markus, Hamill & Sentis, 1987; Markus, Smith & Moreland, 1985; Stein & Corte, 2007). Stein & Corte, 2008 used intra class correlation coefficients (ICC) to show test-retest reliability across a 12-month interval for the number of self-schemas, ICC = .44, p<.001; number of positive schemas, ICC = 0.53, p<0.001; number of negative schemas, ICC .51, p<0.001; and interrelatedness, ICC = 65, p<0.001.

Primary Outcome Measures: Eating Disorder Symptoms

Ecological Momentary Assessment (EMA) of Eating Disordered Behaviors

A computerized menu-driven interview was used to prospectively measure ED behaviors for 22-day intervals pre and post-intervention. Participants were asked to carry with them at all times during their waking hours, a PalmPilot computer and were instructed to record all ED behaviors at the time they occur (event-triggered). The first screen of the computerized questionnaire was comprised of a list of ED behaviors including binge eating, vomiting, laxative, diuretic and diet pill use, and exercising to lose weight. The second screen focused on the first behavior endorsed and consisted of specific questions about the enacted behavior. Responses were automatically recorded with a date and time stamp that could not be erased by the participant. In addition to the event triggered recording, a single time triggered recording was also used. At the end of each day, participants were asked to respond to questions about restricting and fasting over the last 24 hours. Questions for the EMA interview were based on items from the EDE (Fairburn & Cooper, 1993) that were rewritten to focus on the behavioral episode just enacted.

Participants were instructed on definitions of target behaviors, use of the PalmPilot, and procedures for recording ED behaviors in a 1-hour individual orientation session. Participants were also given a manual that provided a step-by-step description of the recording and the project phone number in the event that questions or problems with the PDA arose. Participants met with project staff weekly during the EMA period for data transfer and backup.

Participant compliance with the PDA recordings was assessed by computing the percentage 21-day period with at least one recording. Since participants were asked to complete a recording at the end of each day to indicate the presence or absence of restricting or fasting behavior, if compliant with the EMA instructions each participant should have had at least one recording for each of the 21 days. Across the 63 participants completing pre-intervention PDA recordings, the mean number of days of recordings was 19.43 (range: 11–21). For the immediate post-intervention data collection, 36 women completed the EMA protocol; the mean number of days of recordings was 19.06 (range = (7 – 21 days).

EMA has been shown to be a valid and reliable approach to measurement of eating disorder behaviors in samples of women with anorexia nervosa (Burd, et al., 2009) and bulimia nervosa (Hilbert & Tuschen-Caffier, 2007; Stein & Corte, 2003). Evidence to support the construct, predictive and criterion related validity has been shown (Smyth, Wonderlich, Sliwinski, Crosby, Engel, Mitchell & Calogero, 2009; Tasca et al. 2009;). Further, evidence showing non-reactivity of eating behaviors supports the reliability of EMA (Stein, et al., 2003).

The number of episodes of self-induced vomiting, laxative, diuretic and diet pill use was summed to create an average weekly number of purging behaviors score. Similarly, the number of days of fasting and restricting was used to compute an average weekly restricting score. The number of binge episodes and excessive exercise episodes (≥1.5 hours of exercise for purpose of weight control) were averaged to compute a weekly binge episode score and a weekly exercise score, respectively.

Eating Disorder Inventory

(Garner, 1991). This widely used self-report questionnaire was used to measure current eating attitudes and behaviors. It consists of 64 items to generate 8 subscale scores, which are computed by summing scores across all items in the subscale. Criterion-related and concurrent validity of the subscales have been shown with clinical and nonclinical samples (Garner, 1991). In our study, we used 3 of the 8 subscales: Drive for Thinness, Body Dissatisfaction and Bulimia. Test-retest reliability has also been shown over 3-week intervals in normal and symptomatic college samples (Body Dissatisfaction r=0.96 –0.97, Drive for Thinness r=0.85 to 0.93; Bulimia: r=0.90 to 0.92) (Garner, Olmstead & Polivy, 1983; Wear & Pratz, 1987). In our sample, alpha coefficients from baseline data were Body Dissatisfaction alpha=0.89, Drive for Thinness alpha=0.84 and Bulimia alpha=0.88.

Secondary Outcome Measures: Psychological Well-Being and Functional Health

Psychological Well-Being scale

(Ryff, 1989) is comprised of 6 theoretically derived dimensions of psychological well-being including: Self-Acceptance, Positive Relations with Others, Autonomy, Environmental Mastery, Purpose in Life and Personal Growth to measure eudemonic well-being. Each dimension is measured with a 14-item subscale. Psychometric studies provide evidence to support the validity and reliability of the measure. The scales were positively correlated with other measures of psychological health including life satisfaction, affect balance and negatively correlated with measures of depression. Confirmatory factor analysis based on data from a large national probability sample of adults support the six-scale structure (Ryff & Keyes, 1995). Internal consistency for the scales was high (alphas 0.86 to 0 .93) and test-retest reliability for a 6-week interval ranged from 0.81 to 0.88 (Ryff, 1989; Schmutte & Ryff, 1997). The measure has been shown to be sensitive to changes in well-being precipitated by discreet life events (Kling, Ryff, & Essex, 1997; Ryff & Singer, 1996). In our sample, alpha coefficients based on baseline data were high: Self-Acceptance alpha=0.92, Positive Relations with Others alpha=0.93, Autonomy alpha 0.86, Environmental Mastery alpha=0.87, Purpose in Life alpha 0.91, and Personal Growth alpha=0.82.

SF-36 Health Survey

(Ware, Snow, Kosinski, & Gandek, 1993). The SF-36 was used to assess changes in functional health. The measure taps four different dimensions of functional health including behavioral functioning, perceived well-being, social and role disability and perceived general health with 8 subscales: physical functioning, bodily pain, role limitation due to physical problems, role limitations due to emotional health problems, social functioning, emotional well-being, energy/fatigue and general health perceptions. It has been used extensively to measure treatment outcomes with medical and psychiatric populations and includes dimensions included in recent conceptualizations of mental health. Validity of the measure has been established using factor analysis, known group and criterion-related approaches (McHorney, Ware, & Raczek, 1993; McHorney, Ware, Rogers, Raczek, & Lu, 1992; Ware & Sherbourne, 1992). In a clinical sample of 969 adults, alpha coefficients for the 8 scales ranged from .77 to .98 providing evidence to support the internal consistency of the scales. Adequate test-retest reliability has been shown with nonclinical and clinical samples (Ware et al., 1993). In our sample at baseline, alpha coefficients ranged from 0.77 to 0.90.

Treatment

The psychotherapy component of treatment consisted of weekly one-hour individual psychotherapy session and one hour weekly group psychotherapy for both the IIP and SPI conditions. Treatment was 20 weeks in duration and participants were encouraged to make up missed sessions. One therapist treated participants in both conditions (individual and group sessions). The therapist was a female masters prepared psychiatric nurse with extensive experience in delivering individual psychotherapy, including treatment for the eating disorders.

Therapist training

The therapist received training in IIP and SPI prior to the start of the study by the first author. The therapist was required to study the treatment manuals and received didactic training on self-schema theory and the foundations of the intervention strategies. Once in the intervention phase of the study, the therapist attended biweekly supervision sessions with the first author where case presentations and intervention protocols were discussed.

Treatment Fidelity

To establish fidelity to the treatment assignment, transcriptions of a randomly selected set of sessions were coded for the number of condition-specific intervention behaviors used by the therapist. All psychotherapy sessions were audiotaped. For each participant, the 20 potential sessions were grouped into four units of five sessions and one session from each unit was randomly selected for fidelity assessment. Tapes of the selected sessions were transcribed and coded using the Possibilities Project Psychotherapy Coding Questionnaire (PPPCQ) (Stein, Sargent & Raphael, 2007) developed for this study. The PPPCQ consists of 98 items comprising 3 scales, IIP scale (58 items, alpha=0.79), the SPI scale (18 items, alpha=0.84) and the common factors scale (22 items, alpha=0.45). Items for the IIP scale (e. g. , “the therapist worked with the patient to explore possible selves”, “the therapist worked with the patient to identify personal strengths that were helpful in achieving the possible self goal”) and SPI scales (“The therapist encourages consideration of the link between past experience and present eating disorder symptoms”, “The therapist assisted the patient to identify her own ideas, feelings and opinions”) were based on their respective manuals. A total of 143 tapes were coded. Results showed that for participants randomized to the IIP condition, the therapist used on average 12% of the total IIP behavioral interventions in a single therapy session compared to less than 0.01% in a single session for participants in the SPI condition, t(78.4) = 12.63, p<.0001. In contrast, the therapist used on average 35% of the SPI behavioral interventions in a single session for the SPI group compared to only 4% in the IIP group, t(118)=−15.23, p<.0001. Finally no significant differences were found between the two groups in the average number of common behavioral interventions used per session, IIP =23% and SPI 25%, t(141)=<1. The difference between the IIP and SPI in the mean number of interventions per session is most likely due to the fact that the IIP interventions were written at a molar level to address a single primary goal per session. In contrast, the SPI interventions referred to specific behavioral strategies used to foster exploration and communication during a therapy session (Stein et al, 2007).

Statistical Method

Longitudinal analyses were performed to analyze intervention effects for eating disorder symptoms (EMA data and EDI) and indicators of psychological and functional health, on an intent-to-treat (or as randomized) basis using all available data from the 69 randomized participants. To account for correlation among observations from the same subject, linear mixed-effects models were used to estimate the regression relationship over time and to test whether the regression relationship differed across the two treatment groups. From this model, appropriate--tests were used to assess the statistical significance of intervention effects.

To examine relationships between changes in positive self-schemas and level of interrelatedness with changes in the eating disorder behaviors of interest, a linear regression model was used with change in behavior as the dependent variable and change in the number of positive schemas or level of interrelatedness as the independent variable of interest, while testing whether to incorporate and adjust for baseline values. Regression model diagnostics for assumptions of normality and variance homogeneity were investigated and model significance was tested by F-test followed by the t-tests to assess the statistical significance of the regression coefficients measuring the associations with changes in behavior and the change in the number of positive schemas or level of interrelatedness. Several regression models were fit and we found that the covariates for baseline values were not statistically significant and therefore the final regression models were simple linear regression model between the change in behavior and change in the number of positive schemas or level of interrelatedness. We also conducted a mediation analysis following Baron and Kenny (1986)’s steps to test the self-schema/level of interrelatedness as a mediator and found that it was not statistically significant and therefore the results were not reported.

All analyses were conducted in SAS, version 9.2, using a 5% level of significance and two-sided tests. Analyses and results were not adjusted for multiple comparisons.

Results

Recruitment and Retention

Although pilot testing was completed to establish feasibility of the trial, referrals were lower than expected and the target sample size was not reached. As shown in Figure 1, a total of 149 women completed the SCID and physical health screening. Forty-six women (30.9%) were ineligible to participate and 5 women (3.4%) failed to respond to calls after screening was completed and eligibility to participate was established. Twenty-nine women (19.5%) dropped out of the study during the pre-intervention data collection prior to random assignment. A total of 69 completed baseline data collection and were randomly assigned to a treatment condition (IIP n=34 and SPI n=35). A total of 41 women (59.4%) completed the one-month post-intervention data collection (IIP n=20 and SPI n=21). No differences were found between dropout and treatment completers in demographic characteristics, baseline eating disorder symptoms or self-schema properties (see Stein, et al., 2011 for a more detailed description of attrition over time).

Figure 1.

^aIIP=Identity Intervention Program

^bSPI=Supportive Psychotherapy Intervention Program

^cInvestigator removed

Random Assignment

Preliminary data analyses showed that at baseline the two groups did not differ in age [IIP: M=23.5 (4.0); SPI: M=24.4 (4.1)], years of education [IIP: M=13.7 (5.5); SPI: M=14.1] or race [IIP: 85.3% (n=29) White; SPI: 67.6% (n=23) White] (all p-values>.10). The groups also did not differ according to eating disorder diagnosis, lifetime prevalence of major depressive disorder, and lifetime prevalence other than major depressive disorder were similar for the two treatment groups (see Table 2).

Table 2.

Diagnoses, self schema properties, eating disorder symptoms and psychological and functional health at baseline by treatment group

	IIP (N = 34)		SPI (N = 35)		P-Value
	N	%	N	%
Diagnosis					0.88
AN	1/34	2.9	3/35	8.6
AN Sub -threshold	1/34	2.9	1/35	2.9
BN	23/34	67.6	22/35	62.9
BN Sub -threshold	9/34	26.5	9/35	25.7
Lifetime Prevalence of Major Depressive Disorder	9/32	28.1	15/34	44.1	0.21
Lifetime Prevalence other than Major Depressive Disorder	10/34	29.4	7/35	20.0	0.41
Self-Schema Properties	M (SD)		M (SD)
# of Positive Schemas	8.47 (5.05)		10.1 (6.16)		0.23
Interrelatedness	0.18 (0.13)		0.20 (0.13)		0.68
Eating Disorder Symptoms
Binge	2.23 (2.57)		1.92(1.83)		0.58
Purging	4.65 (4.33)		6.13 (5.56)		0.24
Fasting & Restricting	6.00 (2.89)		4.11 (3.05)		0.01 *
Exercise	0.64 (1.23)		0.47(0.83)		0.52
EDI -Drive for Thinness	15.85 (4.45)		12.46 (5.72)		0.01 *
EDI -Bulimia	11.68(6.50)		10.43 (5.86)		0.41
EDI -Body Dissatisfaction	15.97 (8.11)		15.11 (6.86)		0.64
Psychological Well -Being
Autonomy	54.88 (12.89)		54.69 (12.85)		0.95
Environmental Mastery	52.26 (12.36)		52.26 (11.48)		1.00
Personal Growth	67.26 (10.22)		67.77 (8.2)		0.82
Positive Relations	60.12 (14.01)		59.23 (15.45)		0.80
Self-Acceptance	52.42 (14.56)		52.15 (14.97)		0.94
Purpose in Life	59.76 (13.61)		59.14 (12.471)		0.86
Functional Health
Physical Functioning	95.30 (10.45)		94.29 (11.12)		0.70
Role Limitations	84.85 (30.58)		90.71 (21.93)		0.37
Due to Physical Health
Bodily Pain	81.52 (18.98)		82.57 (20.84)		0.83
General Health	65.76 (20.32)		66.88 (19.22)		0.82
Vitality	47.88 (17.55)		45.71 (19.22)		0.63
Social Functioning	68.56 (20.99)		73.21(23.12)		0.39
Emotional Well -Being	51.52 (40.90)		67.61 (40.01)		0.11
Mental Health	62.79 (15.95)		66.06 (16.36)		0.41

Note. Mean and standard deviations in bracket are reported for continuous variables. Frequency and percentages are reported for categorical variables.

^aMean number of behaviors including binge eating, purging (vomiting, laxative, diet pill, diuretic use) and exercise ≥ 1.5hrs /daily for weight control per week.

^bMean number of days (per week) reported restricting or fasting to control weight.

^*P-value less than 0.05

Also shown in Table 2 are baseline self-schema properties, eating disorder symptom levels, psychological well-being and functional health scores by group. At baseline, women randomized to the IIP group had significantly higher mean number of fasting/restricting days per week and EDI drive for thinness scores compared to women randomized to the SPI group. The two groups were comparable on all other measures. At baseline, a total of 6 women had threshold or sub threshold levels of AN (4 women had AN, and 2 women had sub threshold AN). One woman had severe symptoms, three had moderate symptoms and two had mild symptoms. The mean BMI of women with AN type disorder was 17.54 (0.44) and five of the six women had amenorrhea.

The IIP and SPI conditions included 20 individual and 20 group sessions. The number of individual sessions were not significantly different across groups [(Individual: IIP – M=15.8 (6.3); SPI – M=13.1 (7.2), t(64)=1.65, p>0.10. Further, the number of group sessions completed by participants was lower than the individual sessions, but there were no between group differences [IIP: M=10.5 (7.0); SPI – M=7.5 (7.2), t(d64)=1.71, p=0.09.

Effects of Treatment on Self-Schemas

Based on the conceptual model, we predicted that the IIP intervention would lead to a greater increase in positive self-schemas and a greater decrease in interrelatedness among the self-schemas compared to the SPI. As seen in Table 3, the IIP group had an estimated mean increase of 2.86 positive schemas with a 95 % confidence interval (CI) ranging from as few as an increase of 0.90 to as many as 4.82 positive self-schemas (p = 0.005). For the SPI group, the change from baseline to one-month post-intervention was not significant; they had an estimated mean increase of 1.46 (CI = −0.45, 3.38; p = 0.132). The between group difference in mean changes was not found to be significant (p = 0.315).

Table 3.

Effects of interventions on self-schema properties, eating disorder symptoms and indicators of psychological and functional health by treatment groups. Effect size(ES) is calculated by the mean difference (Δ) divided by its SE.

	IIP	SPI	Δ
	Estimate (SE )	Estimate (SE )	Estimate (SE )	ES
Self-Schema Properties
# of Positive Schemas	2.86 (1.0) *	1.46 (1.0)	1.39 (1.4)	0.99
Interrelatedness	0.024 (0.03)	0.022 (0.03)	0.001 (0.04)	0.03
Eating Disorder Symptoms
Drive for Thinness	−11.2 (1.2) *	−6.53 (1.2) *	−4.69 (1.6) *	2.93
Bulimia	−9.97 (1.1) *	−9.14 (1.1) *	−0.84 (1.6)	0.53
Body Dissatisfaction	−5.89 (0.9) *	−4.23 (0.8) *	−1.66 (1.2)	1.38
Binge	−1.72 (0.36)*	−1.50 (0.37)	−0.22 (0.51)	0.43
Purging	−3.47 (0.83)*	−4.73 (0.85)*	1.26 (1.19)	1.06
Fasting & Restricting	−4.60 (0.73)*	−2.75 (0.74)*	−1.85 (1.04)*	1.78
Exercise	−0.11 (0.24)	−0.15 (0.25)	0.04 (0.34)	0.12
Psychological Well-Being
Autonomy	5.92 (1.3) *	2.01 (1.2)	3.91 (1.8) *	2.17
Environmental Mastery	9.92 (1.6) *	5.98 (1.5) *	3.94 (2.2)	1.79
Personal Growth	6.15 (1.3) *	5.12 (1.2) *	1.03 (1.8)	0.57
Positive Relations	7.61 (1.7) *	3.27 (1.6) *	4.33 (2.4)	1.8
Self-Acceptance	13.6 (2.2) *	10.5 (2.2) *	3.09 (3.2)	0.97
Purpose in Life	8.95 (1.8) *	6.57 (1.8) *	2.38 (2.5)	0.95
Functional Health
Physical Functioning	1.57 (2.4)	5.22 (2.3) *	−3.65 (3.3)	1.11
Role Limitations Due to Physical Health	2.73 (6.8)	6.83 (6.6)	−4.09 (9.5)
Bodily Pain	3.34 (5.5)	0.69 (5.3)	2.66 (7.2)	0.28
General Health	10.6 (3.0) *	13.2 (2.9)*	−2.60 (4.2)	0.62
Vitality	14.7 (4.7) *	16.9 (4.5) *	−2.20 (6.4)	0.34
Social Functioning	17.1 (5.0) *	14.3 (4.8) *	2.84 (6.9)	0.41
Emotional Well-Being	21.9 (10.3) *	5.73 (9.9)	16.2 (14.3)	1.13
Mental Health	11.5 (4.2) *	4.75 (4.0)	6.74 (5.8)	1.16

Note. Mixed regression model estimates and standard errors in the bracket are reported.

^*P-value less than 0.05

The expected decrease in self-schema interrelatedness was not found. Both the IIP group and the SPI group experienced no change from baseline to one-month post-intervention, and consequently, no group difference was found (p = 0.979). Estimated changes in means by group and their difference can be found in Table 3.

Effects of Treatment on Eating Disorder Symptoms

We also posited that the IIP treatment would lead to greater decreases in eating disorder symptoms compared to the SPI treatment. Results of these analyses are shown in Table 3. Both the IIP and SPI groups showed a significant decrease in the weekly mean number of binge eating and purging episodes (EMA) between baseline and one-month post treatment, but between group differences were not significant. The IIP group reported a significantly greater decrease in the mean number of restricting days per week compared to the SPI group. The mean number of weekly excessive exercise episodes did not change between pre and post intervention for either group.

For women with AN, three of the six completed through the post-intervention follow-up and their mean BMI at that point was 20.8 (4.47). All three remained amenorrheic.

For the EDI drive for thinness scale, women in the IIP group reported a significantly greater decrease in drive for thinness compared to those in the SPI group (p = 0.005). For the EDI bulimia and body dissatisfaction scales, both groups experienced significant decreases between baseline and post-intervention, but the group effect was not significant.

Effects of Treatment on Psychological and Functional Health

It was hypothesized that the IIP would contribute to greater improvements in psychological and functional health and well-being compared to the SPI treatment. Estimated changes in means by group and their differences can be found in Table 3.

For psychological well-being, we found that the increase in the autonomy score (Ryff) was significantly greater for the IIP post intervention compared to the SPI group (p = 0.028). All other well-being scale scores increased between pre and post intervention but the difference between the IIP and the SPI groups was not significant.

For functional health (SF-36) increases were found in the IIP group for all the scores and these increases were found to be statistically significant with the exceptions of Physical Functioning, Role Limitations Due to Physical Health and Bodily Pain. Similar increases were see in the SPI group, but in addition to Role Limitations Due to Physical Health and Bodily Pain, Emotional Well-Being and Mental Health were also not significant. None of the differences between the IIP and SPI groups were found to be significant.

Effects of Positive Self-Schema Change on ED Symptoms, Psychological and Functional Health

Because the predicted group differences were generally not found, we focused on increase in positive self-schemas, regardless of treatment condition as a determinant of change in ED symptoms, well-being and functional health. A key hypothesis motivating this study was that changes in self-schema properties were the mechanism underlying behavioral and psychological change. More specifically, it was hypothesized that increases in the number of positive self-schemas and decreases in the level of interrelatedness among self-schemas were hypothesized to be associated with a decrease in eating disorder behaviors and an increase in psychological and functional health in women with AN and BN. As shown in Table 4, no significant relationships were found between an increase in the number of positive self-schemas and eating disorder behaviors. Only a non-significant trend was found between positive self-schema increase and drive for thinness with suggesting that the dissatisfaction greater the increase in number of positive self-schemas, the greater the decrease in drive for thinness between pre and post-intervention.

Table 4.

Effects of changes in positive self-schema properties and level of interrelatedness on eating disorder symptoms and indicators of psychological and functional health

	Associations with Positive Self-Schema			Associations with Interrelatedness
	Cor	β (SE)	p-value	Cor	β (SE)	p-value
Eating Disorder Symptoms
Binge	−0.13	−0.05(0.06)	0.46	−0.03	−0.45(2.45)	0.86
Purging	−0.11	−0.08(0.13)	0.52	−0.09	−2.62(4.95)	0.60
Fasting & Restricting	0.05	0.03(0.09)	0.77	−0.03	−0.65(3.66)	0.86
Exercise	−0.22	−0.05(0.04)	0.19	−0.04	−0.30(1.46)	0.84
EDI -Drive for Thinness	−0.26	−0.29(0.17)	0.11	−0.06	−2.26(6.65)	0.74
EDI -Bulimia	−0.08	−0.08(0.16)	0.63	−0.20	−7.14(5.84)	0.23
EDI -Body Dissatisfaction	−0.02	−0.02(0.15)	0.89	0.29	10.07 (5.56)	0.08
Psychological Well-Being
Autonomy	0.31	0.43(0.21)	0.05	0.20	10.63 (8.34)	0.21
Environmental Mastery	0.28	0.46(0.25)	0.07	0.05	3.06(9.77)	0.76
Personal Growth	0.26	0.34(0.20)	0.10	0.05	2.38(7.80)	0.76
Positive Relations With Others	0.14	0.24(0.28)	0.39	0.04	2.86(10.79)	0.79
Self-Acceptance	0.40	0.92(0.33)	0.01	0.05	4.70(13.94)	0.74
Purpose in Life	0.35	0.57(0.25)	0.02	0.04	2.30(10.06)	0.82
Functional Health
Physical Functioning	0.08	0.15(0.30)	0.62	0.02	1.72(11.75)	0.88
Role Limitations Due to Physical Health	0.04	0.20(0.90)	0.82	−0.09	−17.88 (34.92)	0.61
Bodily Pain	0.14	0.60(0.69)	0.39	−0.01	−0.97(26.96)	0.97
General Medical Health	0.33	0.96(0.45)	0.04	−0.16	−17.50 (18.07)	0.34
Vitality (Energy/Fatigue)	0.33	1.21(0.57)	0.04	0.01	1.10(23.39)	0.96
Social Functioning	0.30	1.23(0.66)	0.07	−0.17	−26.95 (26.24)	0.31
Role Limitations Due to Emotional Problems	0.39	1.38(0.53)	0.01	0.12	15.71 (21.96)	0.48
Emotional Well -Being / Mental Health	0.24	2.11(1.40)	0.14	−0.22	−72.37 (54.56)	0.19

Note. Pearson correlation, parameter estimate from the linear regression model and the associated p-value are reported. Bolded are the p-values< 10%.

An increase of in the number of positive self-schema between baseline and post-intervention was found to be associated with a significant increase in all the psychological well-being scales scores including Autonomy, Self-Acceptance and Purpose in Life and a trend in the same direction was found for the Environmental Mastery and Personal Growth scale scores. Only the Positive Relations with Others was not significantly related to baseline to post-intervention increases in positive self-schemas.

For the functional health scores, increases in the number of positive self-schemas were associated with statistically significant increases in General Medical Health, Vitality, and Emotional Well-Being / Mental Health and a non-significant trend was found for social functioning.

The results of the interrelatedness models yielded no statistically significant results.

Discussion

The results of this study suggest that a cognitive behavioral intervention that focuses on increasing the number of positive self-schemas may be central to improving emotional health and well-being in women with the eating disorders of AN and BN. Although both the IIP and SPI approaches were equally effective in reducing eating disorder symptoms, the results provide preliminary evidence to suggest that the IIP program was more effective in fostering the development of positive domains of self-definition. Further, results show that an increase in the number of positive self-schemas, regardless of the treatment type, was predictive of an increase in psychological well-being and emotional aspects of functional health. Contrary to our prediction, however, no significant change in the level of interrelatedness among the self-schemas was detected for either the IIP or SPI groups. Finally, no association was found between the amount of interrelatedness change and the eating disorder symptoms, psychological well-being or functional health outcomes.

The level of eating disorder behaviors and attitudes at baseline in our sample was similar to the severity of illness of reported for other eating disorder trials. At post-intervention both groups had significant improvements in binge eating and purging behaviors and related ED attitudes of body dissatisfaction and bulimia. In addition, the IIP group had a greater decrease in number of restricting/fasting days and body dissatisfaction scores compared to the control group. However, contrary to our prediction, the changes in eating disorder behaviors and attitudes were generally unrelated to changes in the number of positive self-schemas that occurred during the treatment phase. No ED symptom change was predicted by the positive self-schema change suggesting that overall, self-schema change is not an explanatory mechanism underlying symptom remediation.

Previous studies have shown that the effect of the number of positive self-schemas on eating disorder behaviors and attitudes is mediated through a stable and elaborated conception of the self as fat (i.e., a fat self-schema) (Stein & Corte, 2008; 2007). Although it is not possible to determine from this study, a plausible explanation for the observed pattern of findings is that the nutritional counseling component of the intervention programs modified aspects of the established fat self-schema. In addition to generalized declarative knowledge and incident specific episodic knowledge, self-schemas also include procedural knowledge related to the behavioral domain in the form of strategies and routines (Kendzierski, 1988). Nutritional interventions including factual knowledge about energy and nutrient needs and strategies for healthy eating behavioral patterns may alter procedural knowledge included in the fat self-schema. The change in the fat self-schema knowledge may, in turn, contribute to an overall decrease in eating disorder behaviors and attitudes.

The fact that a broad range of eating disorder behaviors were measured in this study provides an opportunity to determine whether reductions in presenting patterns of eating disorder behaviors are associated with increases in other types of weight control behaviors. Results of many ED clinical trials, particularly those focused on BN, tend to measure a narrow complement of disordered eating behaviors, leaving open the question of whether substitutions rather than elimination of disordered eating behaviors occur. Results of this study show that not only did rates of purging behaviors decrease but days of restricting and fasting also decreased after treatment. Interestingly, exercise was not a commonly used method of weight control and neither intervention led to change in this infrequent behavior.

Perhaps most significant, results of this study suggest that an increase in positive self-schemas during the intervention period was predictive of improvements in mental health at post-intervention. To date, eating disorder treatment studies have tended to focus on the remediation of behavioral and attitudinal symptoms but generally do not include indicators of improved health. These studies imply that the absence of symptoms is commensurate with recovery to a state of health. Yet there is an emerging consensus within the field of psychiatry that mental health is more than the absence of mental disorder symptoms. A recent approach, referred to as the two continua model, is based on the argument that mental illness and health are two distinct but related phenomena that follow different developmental trajectories across adulthood (Westerhof & Keyes, 2010). In this model, mental health is defined as a state of well-being that includes: 1) emotional well-being defined as subjective feelings of life satisfaction and happiness, 2) psychological well-being defined in terms of individual striving and functional health, the realization of one’s potentials, and 3) social well-being defined as social engagement and societal embeddedness.

Results of this study provide evidence that interventions that contribute to the development of positive self-schemas lead to improvements in the dimension of mental health. At baseline, women in both groups scored below age and gender specific norms both on the measure of psychological well-being and the emotional dimensions of functional health. Results of this study show that regardless of the treatment condition, a greater increase in positive self-schemas between the pre and post-intervention period predicted increases in measures related to psychological well-being and some dimensions functional health. For psychological well-being, an increase in positive self-schemas predicted increases in the autonomy, environmental mastery, personal growth, self-acceptance and purpose in life at post-intervention. In fact, for women in the IIP group, the post-intervention scores on these subscales were sufficiently increased to make them approximately equivalent to the means found in a healthy community-based sample of women at midlife (Ryff, 2002). Only, the positive relations with others subscale score change was not associated with changes in the number of positive self-schemas.

An increase in the number of positive self-schemas was also predictive of increases in general health, vitality and mental health measured by the SF-36. The general health subscale of the SF-36 taps a comprehensive view of one’s health including current health, resistance to illness and health outlook. The vitality and mental health subscales were designed to measure subjective-well being, and therefore are consistent with Westerhof’s first dimension of mental health. At baseline, the study groups’ means for these three scales were at or lower than the twenty-fifth percentile based on national norms for women 25 to 35 years of age. However, at post-intervention follow-up, scores for both groups were improved post-intervention to at least the 50^th percentile and these increases in scores were predicted by the increase in number of positive self-schemas.

It is important to note that self-schema theory suggests that positive self-schemas include behavioral routine, strategies and goals that serve to increase commitment to the domain and motivate effective goal-directed behaviors (Kendzierski, 1988). Additionally, studies have shown that positive self-schemas in a domain are predictive of positive affective states and optimism about the future, which are elements of subjective well-being as defined by Westerhof. Consistent with the self-schema theory, results of this study showed that an increase in positive self-schemas was predictive of an increase in purpose and meaning in life, a sense of autonomy, self-acceptance and mastery. In addition, an increase in positive self-schemas predicted increases in vitality and mental health, two subscales of the SF36 that tap subjective well-being as defined by Westerhof.

It is noteworthy that both the treatment and control groups experienced significant increases in social functioning, as measured by the positive relations scale of the Well-being measure and the social functioning subscale of the SF-36. But only social functioning was related to the increase in the number of positive self-schemas and the association did not reach the standard level of significance. Although only speculative, it is possible that the non-specific elements of psychotherapy stemming from the therapeutic relationship contributed to the improvements in the social domain. Given that social well-being is defined as an important component of mental health, continued exploration of the therapeutic active ingredient that contributes to improved social well-being is an important focus for future research.

The decision to use a transdiagnostic approach to the eating disorders in this RCT was based several factors. First, several theories of the etiology of AN and BN are similar in their focus on low self-esteem and self-concept disturbances as an important vulnerability contributing to the disorders (Fairburn, Cooper & Shafran, 2003; Vitousek & Ewald, 1993). In addition support for the identity impairment model predicting eating disorder symptoms as a function of self-schema properties has been shown for both populations (Stein & Corte, 2007). Although women with AN were included in this initial trial, the number of women with AN who were actually enrolled and completed the trial was very small. In addition, the severity of illness and study retention for this subset of the sample were low. Therefore, it is not possible to determine the utility of this approach for this population. Additional research is needed to determine whether increasing the number of positive self-schemas will similarly lead to improvements in subjective and psychological well-being in populations of women with anorexia nervosa.

Despite the fact that this study showed promising results related to the identity intervention program as a means to promote mental health in women with an eating disorder, several limitations must be noted. The first and most obvious limitation of this study is the small sample size and the high rate of attrition. Although the results of this trial are promising, additional higher powered studies are required before firm conclusions regarding the efficacy of the intervention can be made. Another important limitation of the study is that the sample included only women who had no other concurrent DSM-IV axis I disorder. Since the eating disorders are associated with high levels of concurrent depression, anxiety and substance use disorders, the generalizability of the study results to a full range of women with diagnosable eating disorders in limited. A final limitation of the study is that adherence to the group component of the psychotherapy programs was relatively low, and therefore, the extent to which this dimension contributed to improvements in well-being remains unknown. The theoretical perspective underlying this aspect of the treatment program is that social context plays an important role in commitment and development of possible self-goals and contributes to the cognitive elaboration of these new structures. The lower level of adherence to this aspect of the treatment program leaves the question of its importance in positive self-schema development in this population unknown.

Clearly additional research is needed to determine the utility of the identity intervention program in the promotion of recovery and health in women with anorexia and bulimia nervosa. Yet the results of this initial trial are promising and highlight an approach to eating disorder treatment that goes beyond alleviation of presenting symptoms to the promotion of mental health in wellness in this vulnerable population.