Predictors and moderators of outcome for severe and enduring anorexia nervosa

Behaviour Research and Therapy 56 (2014) 91e98

Contents lists available at ScienceDirect

Behaviour Research and Therapy journal homepage: www.elsevier.com/locate/brat

Predictors and moderators of outcome for severe and enduring anorexia nervosa Daniel Le Grange a, *, Ellen E. Fitzsimmons-Craft a, Ross D. Crosby b, Phillipa Hay c, Hubert Lacey d, Bryony Bamford d, Colleen Stiles-Shields a, e, Stephen Touyz f a

The University of Chicago, Chicago, IL, USA Neuropsychiatric Research Institute and the University of North Dakota School of Medicine and Health Sciences, Fargo, ND, USA University of Western Sydney, Sydney, Australia d St. George’s, University of London, London, UK e Northwestern University, Chicago, IL, USA f University of Sydney, Sydney, Australia b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 30 October 2013 Received in revised form 17 March 2014 Accepted 19 March 2014 Available online 29 March 2014

Few of the limited randomized controlled trials (RCTs) for adults with anorexia nervosa (AN) have explored predictors and moderators of outcome. This study aimed to identify predictors and moderators of outcome at end of treatment (EOT) and 6- and 12-month follow-up for adults with AN (N ¼ 63). All participants met criteria for severe and enduring AN (duration of illness  7 years) and participated in an RCT of cognitive-behavioral therapy (CBT-AN) and specialist supportive clinical management (SSCM). General linear models were utilized and included all available outcome data at all time points. Outcome was assessed across three domains: eating disorder quality of life (EDQOL), mental health (MCS), and depressive symptoms (BDI). Predictors of better outcome included: lower age, shorter duration of illness, having AN-R, being employed, not taking psychotropic medication, and better social adjustment. Four moderators of treatment outcome emerged: eating disorder psychopathology (EDE Global), depression (BDI), age, and AN subtype. Participants with higher baseline scores on these measures, older age, or binge eating/purging subtype benefited more from CBT-AN than SSCM. Older patients with more severe eating-related psychopathology and depression have better outcomes in a behaviorally targeted treatment such as CBT-AN rather than a supportive treatment such as SSCM. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Predictors Moderators Severe and enduring anorexia nervosa CBT-AN SSCM

Introduction Few randomized controlled trials (RCTs) examining different psychosocial treatments for adults with anorexia nervosa (AN) have been conducted (e.g., Dare, Eisler, Russell, Treasure, & Dodge, 2001; Lock, Agras, Fitzpatrick, et al., 2013; McIntosh et al., 2005; Pike, Walsh, Vitousek, Wilson, & Bauer, 2003; Russell, Szmukler, Dare, & Eisler, 1987). Most of these studies are compromised through lack of statistical power, and findings are generally inconclusive. The most recent published RCT for this patient population compared the relative efficacy of cognitive-behavioral therapy (CBT-AN) to specialist supportive clinical management (SSCM) in 63 women with severe and enduring anorexia nervosa (SE-AN).

* Corresponding author. The University of Chicago, Department of Psychiatry and Behavioral Neuroscience, 5841 S. Maryland Ave., MC3077, Chicago, IL 60637, USA. Tel.: þ1 773 702 9277; fax: þ1 773 702 9929. E-mail address: [email protected] (D. Le Grange). http://dx.doi.org/10.1016/j.brat.2014.03.006 0005-7967/Ó 2014 Elsevier Ltd. All rights reserved.

While this study was also compromised due to a modest sample size, satisfactory retention in treatment and follow-up was achieved (85% of patients remained in treatment and follow-up). This study demonstrated that patients with SE-AN could make significant and meaningful improvements with both therapies. Both CBTAN and SSCM contributed to improvements over time in several outcome domains: health-related quality of life, body weight, depression, and motivation to change (Touyz et al., 2013). Kraemer, Wilson, Fairburn, and Agras (2002) remind us that while the evaluation of the relative efficacy of two or more treatments in an RCT is helpful, our understanding for whom a specific treatment may be best suited for (moderators of outcome), or the mechanisms through which a treatment might achieve its aims (mediators on outcome), has significant clinical relevance. Few, if any, of the published RCTs for adults with AN have examined moderators and mediators on outcome. Some of the RCTs for adults with AN have examined predictors of treatment outcome. In a study of 33 females with AN treated with CBT or nutritional counseling, Pike et al. (2003) found no significant effect of

92

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

medication status on outcome for those treated with nutritional counseling. They did find a medication effect for CBT in that seven out of eight patients who met criteria for a good outcome were receiving medication compared to four of ten who did not meet criteria for good outcome. McIntosh et al. (2005) randomized 56 females with AN to CBT, SSCM or interpersonal therapy (IPT). These authors found that differences in outcome among the treatment groups were not explained by any difference among treatment groups at baseline. Outside of RCTs, moderators on treatment outcome have also received some attention. Lockwood, Serpell, and Waller (2012) examined moderators of weight gain in the early stages of treatment for 40 females with AN receiving CBT. They found that neither age nor body mass index (BMI) at the start of therapy predicted degree of weight change during the first 10 sessions of CBT. However, participants with elevated anxiety or phobic anxiety were slower to gain weight in the first 10 sessions (or even lost weight), and more severe levels of dietary restraint and shape concern were associated with lower levels of weight change from sessions 6e10. In a study of 218 adults with either BN or BED receiving CBT, Castellini et al. (2011), showed that eating psychopathology, psychiatric comorbidity, impulsivity and emotional eating differ in their association with both objective and subjective binge eating across BN and BED patients. Most recently, Lock, Agras, Le Grange, et al. (2013) demonstrated that for adults with AN, the most efficient predictor of weight recovery at followup (BMI > 19 kg/m2) was weight gain to greater than 85.8% of expected body weight at the end of treatment. In addition, the most efficient predictor of psychological recovery was achievement of a low score on the Eating Disorder Examination (EDE) Weight Concern subscale (<1.8). Exploring mediators and moderators of outcome for adolescents with AN has been equally limited (Eisler et al., 2000; Le Grange, Eisler, Dare, & Hodes, 1992; Lock, Agras, Bryson, & Kraemer, 2005; Lock, Couturier, Bryson, & Agras, 2006). In the largest such study to date, Le Grange et al. (2012), were able to identify at least two moderators at end-of-treatment: eating-related obsessionality (YaleeBrowneCornell Eating Disorder Total Scale) and eating disorder specific psychopathology (EDE Global). In an RCT of familybased treatment (FBT) and adolescent focused therapy (AFT), participants with higher baseline scores on these measures benefited more from FBT than AFT. No mediators of treatment outcome were identified. Taken together, it is clear that the treatment of AN is not only hampered by a limited number of RCTs, but also by the lack of studies exploring for whom treatments work best, or how one treatment versus another brings about therapeutic change. In the present study we examine predictors and moderators of outcome (i.e., eating disorder-related quality of life, mental health, depressive symptoms) for participants in the RCT of CBT-AN and SSCM briefly described above. Given the scarcity of prior work in this domain, we did not advance any specific hypotheses. Rather, we chose to investigate several variables as possible predictors and moderators, and our procedure was therefore an exploratory analysis. Findings should thus be regarded as hypothesis generating as opposed to hypothesis testing. Method Design This RCT occurred at two intervention sites (University of Sydney and St. George’s, University of London). The main outcome report, which has been published elsewhere (Touyz et al., 2013), compared CBT-AN (Pike et al., 2003) to SSCM (McIntosh et al., 2006; McIntosh, Jordan, & Bulik, 2010) among females with SE-AN. Participants (N ¼ 63) were randomly assigned to either CBT-AN

(n ¼ 30) or SSCM (n ¼ 33). This study was reviewed and approved by the Institutional Review Boards at each site. Recruitment for this RCT occurred from 2007 to 2010. After telephone screening (n ¼ 159) to determine eligibility, 73 (46%) individuals were invited for an in-person assessment. Participants were eligible if they were female (males were excluded as we estimated that the number of such cases would be negligible), aged  18 years, met DSM-IV criteria for AN excluding the amenorrhea criterion, and had an illness duration of at least seven years. Participants were also included if the met the criteria above, and were at a BMI of 18.5 or lower. Participants were excluded from the study if they presented with a current manic episode or psychosis, current alcohol or substance abuse or dependence, significant current medical or neurological illness (including seizure disorder) with the exception of nutrition-related alterations that impact on weight, were currently engaged in psychotherapy and not willing to suspend treatment for the duration of their participation in the study, had plans to move beyond commuting distance from the study site in the following 12 months, or did not live within commuting distance to the study site. Eighty-six percent (N ¼ 63) of eligible screened participants agreed to randomization. The majority of those ineligible did not meet the DSM-IV weight criterion or the illness duration criterion. Treatments Treatment was provided on an outpatient basis and involved 30 individual treatment sessions provided over the course of eight months. Sessions were conducted weekly and were 50 min in length. Participants were told that the focus of treatment was on improving quality of life rather than weight gain per se and that specific treatment goals would be decided upon collaboratively at the beginning of therapy. This does not imply that weight gain and other eating disorder symptoms were not a priority, rather, that these were somewhat deemphasized relative to quality of life improvements. The two manualized treatments and their implementation are described in detail elsewhere (McIntosh et al., 2006, 2010; Pike et al., 2003). Briefly, CBT-AN (Pike et al., 2003) focuses on the cognitive and behavioral factors that play a role in the core features of AN and on more global issues associated with AN (e.g., motivation, schema-based work). It is comprised of four phases. In Phase I, treatment is initiated, patients are oriented to CBT-AN, and motivation is addressed. In Phase II, weight gain and cognitive distortions and behavioral disturbances associated with eating and weight are addressed. In Phase III, the focus of treatment is expanded to schema-based work that addresses relevant issues that extend beyond the specific domain of eating and weight. In Phase IV, the course of treatment is reviewed, gains are consolidated and continuing the work of CBT-AN independently after therapy ends is discussed. Although the four phases are described sequentially, CBT-AN is flexible in terms of applying modules of the protocol as needed throughout the course of treatment. For the current study, weight gain and recovery from core eating disorder pathology were not assumed to be treatment priorities. Rather, treatment goals were decided upon collaboratively at the outset of treatment. Weight gain was encouraged but not identified as the primary goal or focus of therapy (although medical stability was monitored and required in order to remain in the study). In general, CBT-AN was implemented flexibly in this study. For example, the motivational enhancement section of the manual was allowed to continue as long as needed. SSCM (McIntosh et al., 2006, 2010) combines features of clinical management and supportive psychotherapy. More specifically, clinical management includes education, care, and

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

support. Attempts are made to foster the therapeutic relationship in a way that promotes treatment adherence. Supportive psychotherapy aims to help the patient through the use of praise, reassurance, and advice. As with CBT-AN, SSCM was modified for the current study so that weight gain was not prioritized. Instead, the goals of SSCM involved improving quality of life and physical well-being. The rationale for this is that improvement in domains outside of core pathology can significantly affect patient wellbeing and disease burden. In general, SSCM aims to help individuals improve their quality of life, which is believed to further motivate and enable them to work on their core eating disorder pathology. Thus, CBT-AN and SSCM were both modified to focus on quality of life and harm minimization associated with the eating disorder. Both treatments still emphasized weight gain, but as a secondary aim. CBT-AN used specific cognitive and behavioral strategies to work on these goals, while SSCM made use of more general, supportive therapeutic techniques. The outcome measures (i.e., eating disorder-related quality of life, mental health, depressive symptoms) were selected to assess the extent to which participants were able to find satisfaction in their lives as a result of treatment. Assessment and procedures Participants completed a number of self-report, as well as investigator-based measures at the baseline assessment, end of treatment (EOT), and 6- and 12-month follow-ups. Of note, independent assessors not involved in treatment delivery and blind to treatment assignment conducted all assessments. Baseline predictors and moderators of treatment outcome In an exploratory analysis of this kind, we utilized all meaningful baseline variables as possible moderators. Therefore, 13 baseline variables thought to be key clinical markers were examined as possible predictors and moderators of treatment outcome as defined in our main outcome report (Touyz et al., 2013). These were 1) age; 2) BMI (weight and height were assessed at baseline; the participant was weighed in light indoor clothing, without shoes on a balance beam scale that was recalibrated regularly); 3) relationship status (single or in a relationship); 4) duration of AN (in years); 5) AN subtype (restricting (AN-R) or binge eating/purging (AN-BP)); 6) employment status (currently studying; employed or full time home duties/caring for children; or unemployed) 7) education level (high school graduate or college graduate); 8) current medical concerns (yes or no; e.g., raised liver enzymes, mild neutropenia); 9) current psychotropic medication (yes or no); 10) eating disorder psychopathology (EDE Global; Fairburn & Cooper, 1993; higher scores indicate greater severity or frequency); 11) readiness to recover from AN (Anorexia Nervosa Stages of Change Questionnaire; ANSOCQ; Rieger, Touyz, & Beumont, 2002; higher scores indicate greater readiness to recover); 12) depressive symptoms (Beck Depression Inventory; BDI; Beck, Steer, & Brown, 1996; higher scores indicate greater depressive symptoms); and 13) social adjustment (Weissman Social Adjustment Scale; WSAS; Weissman & Bothwell, 1976; higher scores indicate greater impairment). Treatment outcome measures Eating disorder-related quality of life Eating disorder-related quality of life was assessed using the Eating Disorders Quality of Life (EDQOL; Engel et al., 2006) instrument. This measure assesses quality of life in eating disorder

93

populations across four domains: psychological, physical/cognitive, financial, and work/school. Lower scores indicate a better quality of life. Mental health Mental health was assessed using the Mental Component Summary (MCS) scale score of the Short Form-12 Health Survey (SF-12; Ware, Kosinski, & Keller, 1996). The SF-12 MCS scale score measures impairment in everyday functioning associated with mental health problems. Lower scores indicate higher levels of impairment. Depressive symptoms The Beck Depression Inventory (BDI; Beck et al., 1996) was used to assess depressive symptoms. This measure was examined as both a treatment outcome measure and a possible predictor/ mediator of treatment outcome (when this was assessed using the EDQOL and the MCS). Statistical analyses Statistical analyses were performed following the guidelines of Kraemer et al. (2002) for evaluating predictors and moderators of treatment effects in RCTs. To test whether baseline variables were predictors or moderators, we constructed a general linear regression model for each outcome measure, with models run separately for EOT score, 6-month follow-up score, and 12-month follow-up score using all available data. Independent variables included the baseline value of the outcome measure, the potential baseline predictor (grand-mean-centered), treatment group (CBT-AN or SSCM), and the predictor-by-treatment group interaction term. Follow-up tests were conducted to understand the nature of significant categorical predictor variables and the nature of significant interaction terms. The main effect for the centered baseline variable provided the test for predictors, while the centered variable-bytreatment group interaction term provided the test for moderators. Missing data for continuous outcome measures at EOT and follow-ups were imputed using multiple imputation based upon fully conditional Markov chain Monte Carlo modeling (Schafer, 1997). Both significance and effect sizes (partial h2) were used in interpreting results. To correct for the use of multiple tests, p < .01 was used as the threshold for significance given that the Bonferroni correction has been criticized for being overly conservative (e.g., Perneger, 1998). Further, it has been suggested that effect sizes rather than p-values should be the primary focus of interest in hypothesis generating studies (Kraemer et al., 2002). As such, we also considered tests with p < .05 and at least a medium effect size to be significant. Cohen (1988) provides the following guidelines for interpreting partial h2: small effect ¼ .01, medium effect ¼ .06, and large effect ¼ .14. Results Detailed description of participant baseline characteristics is available in the main report (Touyz et al., 2013). To summarize, all participants (N ¼ 63) were female and had a mean age of 33.4 years (SD ¼ 9.6; range: 20e62 years), a mean duration of illness of 16.6 years (SD ¼ 8.5; range: 7e49 years), and a mean BMI of 16.2 kg/m2 (SD ¼ 1.3; range: 11.8e18.5). Nearly three-quarters of participants (n ¼ 47; 74.6%) met criteria for AN-R. 37 (58.7%) participants had a comorbid psychiatric disorder, and 26 (41.3%) were taking psychotropic medication at baseline.

94

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

No variables were found to moderate the effect of treatment condition on eating disorder-related quality of life at EOT or at 6month or 12-month follow-up (see Table 1).

Predictors and moderators of outcome from baseline to EOT and follow-up Eating disorder-related quality of life

Mental health

Several non-specific predictors of outcome in eating disorderrelated quality of life from baseline to EOT and 6- and 12-month follow-up were identified (see Table 1). From the 13 candidate variables, four were identified as predictors of eating disorderrelated quality of life at EOT. More specifically, results indicated that lower age (B ¼ .03, t(1, 58) ¼ 2.56, p ¼ .013) and shorter duration of illness (B ¼ .03, t(1, 58) ¼ 2.31, p ¼ .024) predicted improved eating disorder-related quality of life at EOT (controlling for baseline levels) across the two treatment conditions. Being unemployed predicted worse eating disorder-related quality of life at EOT (controlling for baseline levels) across the two treatment conditions relative to those who were employed or engaged in full time home duties/child care (p ¼ .006) but not those who were currently studying (p ¼ .069). Eating disorder-related quality of life at EOT (controlling for baseline levels) did not significantly differ between those who were employed or engaged in full time home duties/child care and those who were currently studying (p ¼ .616). Taking psychotropic medication at baseline predicted worse eating disorder-related quality of life at EOT (controlling for baseline levels) relative to those who were not taking psychotropic medication at baseline (p ¼ .021) across the two treatment conditions. No variable emerged as a significant predictor of change in eating disorder-related quality of life from baseline to 6-month follow-up, but one variable, AN subtype, was identified as a predictor of eating disorder-related quality of life at 12-month follow-up controlling for baseline levels. Individuals with AN-R had improved eating disorder-related quality of life at 12-month follow-up (controlling for baseline levels) compared to individuals with AN-BP (p < .001) across the two treatment conditions.

Several non-specific predictors of mental health outcome from baseline to EOT and 6- and 12-month follow-up were identified (see Table 2). From the 13 candidate variables, four were identified as predictors of mental health at EOT. More specifically, results indicated that lower age (B ¼ .30, t(1, 58) ¼ 1.79, p ¼ .079) and shorter duration of illness (B ¼ .31, t(1, 58) ¼ 1.71, p ¼ .092) predicted improved mental health at EOT (controlling for baseline levels) across the two treatment conditions. Although the tests of the between-subjects effects for these variables were significant (see Table 2), it should be noted that the tests of these parameter estimates were approaching the level of significance. Being unemployed predicted worse mental health at EOT (controlling for baseline levels) across the two treatment conditions relative to those who were employed or engaged in full time home duties/ child care (p ¼ .001) and those who were currently studying (p ¼ .005). Mental health at EOT (controlling for baseline levels) did not significantly differ between those who were employed or engaged in full time home duties/child care and those who were currently studying (p ¼ .719). Results also indicated that better social adjustment (B ¼ .51, t(1, 58) ¼ 2.64, p ¼ .011) predicted improved mental health at EOT (controlling for baseline levels) across the two treatment conditions. Psychotropic medication and AN subtype were identified as predictors of mental health at 6month and 12-month follow-up, respectively, controlling for baseline levels of the construct. Taking psychotropic medication at baseline predicted worse mental health at 6-month follow-up (controlling for baseline levels) relative to those who were not taking psychotropic medication at baseline (p ¼ .027) across the

Table 1 Predictors and moderators of eating disorder quality of life (EDQOL) at end of treatment (EOT), 6-month follow-up (6 mF), and 12-month follow-up (12 mF). Variable

EOT variable F

EOT variable  treatment F

6 mF Variable F

6 mF Variable  treatment F

12 mF Variable F

12 mF Variable  treatment F

Status

Age

7.09* (1, 58) partial h2 ¼ .11 3.80 (1, 58) partial h2 ¼ .06 1.71 (1, 58) partial h2 ¼ .03 8.37** (1, 58) partial h2 ¼ .13 .57 (1, 58) partial h2 ¼ .01 4.10* (2, 56) partial h2 ¼ .13 .59 (1, 58) partial h2 ¼ .01 .06 (1, 58) partial h2 ¼ .00 5.59* (1, 58) partial h2 ¼ .09 .00 (1, 58) partial h2 ¼ .00 1.30 (1, 58) partial h2 ¼ .02 1.67 (1, 58) partial h2 ¼ .03 .88 (1, 58) partial h2 ¼ .02

.53 (1, 58) partial h2 ¼ 1.45 (1, 58) partial h2 ¼ .87 (1, 58) partial h2 ¼ .00 (1, 58) partial h2 ¼ 1.51 (1, 58) partial h2 ¼ 1.86 (2, 56) partial h2 ¼ .01 (1, 58) partial h2 ¼ 1.18 (1, 58) partial h2 ¼ .03 (1, 58) partial h2 ¼ .47 (1, 58) partial h2 ¼ .43 (1, 58) partial h2 ¼ 1.69 (1, 58) partial h2 ¼ .02 (1, 58) partial h2 ¼

.76 (1, 58) partial h2 ¼ 2.07 (1, 58) partial h2 ¼ 1.53 (1, 58) partial h2 ¼ .17 (1, 58) partial h2 ¼ .84 (1, 58) partial h2 ¼ .38 (2, 56) partial h2 ¼ 1.73 (1, 58) partial h2 ¼ .57 (1, 58) partial h2 ¼ 2.60 (1, 58) partial h2 ¼ .23 (1, 58) partial h2 ¼ .25 (1, 58) partial h2 ¼ .09 (1, 58) partial h2 ¼ .52 (1, 58) partial h2 ¼

.30 (1, 58) partial h2 ¼ 2.37 (1, 58) partial h2 ¼ 1.64 (1, 58) partial h2 ¼ .36 (1, 58) partial h2 ¼ 2.08 (1, 58) partial h2 ¼ 2.71 (2, 56) partial h2 ¼ .05 (1, 58) partial h2 ¼ .05 (1, 58) partial h2 ¼ .33 (1, 58) partial h2 ¼ .02 (1, 58) partial h2 ¼ .45 (1, 58) partial h2 ¼ .37 (1, 58) partial h2 ¼ .63 (1, 58) partial h2 ¼

2.16 (1, 58) partial h2 ¼ .04 2.18 (1, 58) partial h2 ¼ .04 .56 (1, 58) partial h2 ¼ .01 .71 (1, 58) partial h2 ¼ .01 17.59*** (1, 58) partial h2 ¼ .23 .02 (2, 56) partial h2 ¼ .00 .34 (1, 58) partial h2 ¼ .01 .11 (1, 58) partial h2 ¼ .00 .01 (1, 58) partial h2 ¼ .00 .63 (1, 58) partial h2 ¼ .01 1.38 (1, 58) partial h2 ¼ .02 1.82 (1, 58) partial h2 ¼ .03 .41 (1, 58) partial h2 ¼ .01

2.96 (1, 58) partial h2 ¼ .62 (1, 58) partial h2 ¼ 2.56 (1, 58) partial h2 ¼ 3.17 (1, 58) partial h2 ¼ .06 (1, 58) partial h2 ¼ .00 (2, 56) partial h2 ¼ .11 (1, 58) partial h2 ¼ .82 (1, 58) partial h2 ¼ .03 (1, 58) partial h2 ¼ 1.78 (1, 58) partial h2 ¼ .53 (1, 58) partial h2 ¼ .19 (1, 58) partial h2 ¼ .91 (1, 58) partial h2 ¼

P at EOT

BMI Relationship status AN duration AN subtype Employment status Education Medical concerns Psychotropic medication EDE Global ANSOCQ BDI WSAS

.01 .02 .02 .00 .03 .06 .00 .02 .00 .01 .01 .03 .00

.01 .03 .03 .00 .01 .01 .03 .01 .04 .00 .00 .00 .01

.01 .04 .03 .01 .04 .09 .00 .00 .01 .00 .01 .01 .01

.05 e .01 e .04 P at EOT .05 P at 12 mF .00 P at EOT .00 e .00 e .01 P at EOT .00 e .03 e .01 e .00 e .02

Note. BMI ¼ body mass index. AN ¼ anorexia nervosa. EDE ¼ Eating Disorder Examination. ANSOCQ ¼ Anorexia Nervosa Stages of Change Questionnaire. BDI ¼ Beck Depression Inventory. WSAS ¼ Weissman Social Adjustment Scale. P ¼ predictor. *p < .05. **p < .01. ***p < .001.

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

95

Table 2 Predictors and moderators of the Mental Component Summary (MCS) scale of the Short-Form-12 Health Survey (SF-12) at end of treatment (EOT), 6-month follow-up (6 mF), and 12-month follow-up (12 mF). Variable

EOT variable F

EOT variable x treatment F

6 mF Variable 6 mF Variable x F treatment F

Age

4.37* (1, 58) partial h2 ¼ .07 2.46 (1, 58) partial h2 ¼ .04 .11 (1, 58) partial h2 ¼ .00 6.42* (1, 58) partial h2 ¼ .10 1.10 (1, 58) partial h2 ¼ .02 6.32** (2, 56) partial h2 ¼ .18 .01 (1, 58) partial h2 ¼ .00 .05 (1, 58) partial h2 ¼ .00 3.45 (1, 58) partial h2 ¼ .06 .10 (1, 58) partial h2 ¼ .00 .02 (1, 58) partial h2 ¼ .00 .00 (1, 58) partial h2 ¼ .00 6.98* (1, 58) partial h2 ¼ .11

.08 (1, 58) partial h2 ¼ .00

2.22 (1, 58) partial h2 ¼ .04 1.56 (1, 58) partial h2 ¼ .03 .04 (1, 58) partial h2 ¼ .00 1.32 (1, 58) partial h2 ¼ .02 .43 (1, 58) partial h2 ¼ .01 1.09 (2, 56) partial h2 ¼ .04 2.31 (1, 58) partial h2 ¼ .04 .12 (1, 58) partial h2 ¼ .00 5.18* (1, 58) partial h2 ¼ .08 .27 (1, 58) partial h2 ¼ .01 .35 (1, 58) partial h2 ¼ .01 3.34 (1, 58) partial h2 ¼ .05 2.03 (1, 58) partial h2 ¼ .03

BMI

Relationship status

AN duration

AN subtype

Employment status

Education

Medical concerns

Psychotropic medication EDE Global

ANSOCQ

BDI

WSAS

.37 (1, 58) partial h2 ¼ .01 1.63 (1, 58) partial h2 ¼ .03 .15 (1, 58) partial h2 ¼ .00 .74 (1, 58) partial h2 ¼ .01 .92 (2, 56) partial h2 ¼ .03 .01 (1, 58) partial h2 ¼ .00 .94 (1, 58) partial h2 ¼ .02 .31 (1, 58) partial h2 ¼ .01 4.67* (1, 58) partial h2 ¼ .07 2.16 (1, 58) partial h2 ¼ .04 4.05* (1, 58) partial h2 ¼ .07 .83 (1, 58) partial h2 ¼ .01

12 mF Variable 12 mF Variable x F treatment F

Status

.68 (1, 58) partial h2 ¼ .01

1.77 (1, 58) 4.98* (1, 58) partial h2 ¼ .03 partial h2 ¼ .08

P at EOT; M at 12 mF

.02 (1, 58) partial h2 ¼ .00

3.08 (1, 58) .06 (1, 58) partial h2 ¼ .05 partial h2 ¼ .00

e

.04 (1, 58) partial h2 ¼ .00

.00 (1, 58) 3.50 (1, 58) partial h2 ¼ .00 partial h2 ¼ .06

e

.56 (1, 58) partial h2 ¼ .01

2.38 (1, 58) 2.61 (1, 58) partial h2 ¼ .04 partial h2 ¼ .04

P at EOT

3.10 (1, 58) partial h2 ¼ .05

9.53** (1, 58) .48 (1, 58) partial h2 ¼ .14 partial h2 ¼ .01

P at 12 mF

.95 (2, 56) partial h2 ¼ .03

.06 (2, 56) .02 (2, 56) partial h2 ¼ .00 partial h2 ¼ .00

P at EOT

1.77 (1, 58) partial h2 ¼ .03

.87 (1, 58) .86 (1, 58) partial h2 ¼ .02 partial h2 ¼ .02

e

.18 (1, 58) partial h2 ¼ .00

.01 (1, 58) 1.37 (1, 58) partial h2 ¼ .00 partial h2 ¼ .02

e

.96 (1, 58) partial h2 ¼ .02

1.84 (1, 58) .15 (1, 58) partial h2 ¼ .03 partial h2 ¼ .00

P at 6 mF

.86 (1, 58) partial h2 ¼ .02

.08 (1, 58) 3.08 (1, 58) partial h2 ¼ .00 partial h2 ¼ .05

M at EOT

.09 (1, 58) partial h2 ¼ .00

.73 (1, 58) .01 (1, 58) partial h2 ¼ .01 partial h2 ¼ .00

e

.01 (1, 58) partial h2 ¼ .00

.73 (1, 58) .00 (1, 58) partial h2 ¼ .01 partial h2 ¼ .00

M at EOT

.01 (1, 58) partial h2 ¼ .00

1.52 (1, 58) .01 (1, 58) partial h2 ¼ .03 partial h2 ¼ .00

P at EOT

Note. BMI ¼ body mass index. AN ¼ anorexia nervosa. EDE ¼ Eating Disorder Examination. ANSOCQ ¼ Anorexia Nervosa Stages of Change Questionnaire. BDI ¼ Beck Depression Inventory. WSAS ¼ Weissman Social Adjustment Scale. P ¼ predictor. M ¼ moderator. *p < .05. **p < .01.

two treatment conditions. Individuals with AN-R had improved mental health at 12-month follow-up (controlling for baseline levels) compared to individuals with AN-BP (p ¼ .003) across the two treatment conditions. Several moderators of mental health emerged (see Table 2). In particular, EDE Global scores and depressive symptoms emerged as moderators of treatment condition on mental health at EOT. Results revealed that those with higher EDE Global scores/depressive symptoms at baseline responded better to CBT-AN at EOT in terms of mental health. Age emerged as a moderator of this effect at 12month follow-up. Results indicated that older participants responded better to CBT-AN at 12-month follow-up in terms of mental health. Depressive symptoms Several non-specific predictors of depression outcome from baseline to EOT and 6- and 12-month follow-up were identified (see Table 3). From the 13 candidate variables, five were identified as predictors of depressive symptoms at EOT. More specifically, results indicated that lower age (B ¼ .32, t(1, 58) ¼ 1.59, p ¼ .116) and shorter duration of illness (B ¼ .32, t(1, 58) ¼ 1.51, p ¼ .136) predicted decreased depressive symptoms at EOT (controlling for baseline levels) across the two treatment

conditions. As above, although the tests of the between-subjects effects for these variables were significant (see Table 3), it should be noted that the tests of these parameter estimates were approaching the level of significance. Regarding employment status, being unemployed predicted higher depressive symptoms at EOT (controlling for baseline levels) across the two treatment conditions relative to those who were employed or engaged in full time home duties/child care (p ¼ .004) and those who were currently studying (p ¼ .036). Depressive symptoms at EOT (controlling for baseline levels) did not significantly differ between those who were employed or engaged in full time home duties/child care and those who were currently studying (p ¼ .772). Taking psychotropic medication at baseline predicted increased depressive symptoms at EOT (controlling for baseline levels) relative to those who were not taking psychotropic medication at baseline (p ¼ .006) across the two treatment conditions. Finally, results indicated that better social adjustment (B ¼ .67, t(1, 58) ¼ 2.64, p ¼ .011) predicted decreased depressive symptoms at EOT (controlling for baseline levels) across the two treatment conditions. AN subtype was identified as a predictor of depressive symptoms at both 6-month and 12-month follow-up controlling for baseline levels of the construct. Individuals with AN-R had decreased depressive symptoms at both 6- and 12month follow-ups (controlling for baseline levels) compared to

96

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

Table 3 Predictors and moderators of the Beck Depression Inventory (BDI) at end of treatment (EOT), 6-month follow-up (6 mF), and 12-month follow-up (12 mF). Variable

EOT variable F

EOT variable x treatment F

6 mF Variable F

6 mF Variable x treatment F

12 mF Variable F

12 mF Variable x treatment F

Status

Age

3.97 (1, 58), p ¼ .05 partial h2 ¼ .06 2.89 (1, 58) partial h2 ¼ .05

.01 (1, 58) partial h2 ¼ .00

1.99 (1, 58) partial h2 ¼ .03 .97 (1, 58) partial h2 ¼ .02 2.86 (1, 58) partial h2 ¼ .05 .79 (1, 58) partial h2 ¼ .01 4.60* (1, 58) partial h2 ¼ .07

.31 (1, 58) partial h2 ¼ .01

1.24 (1, 58) partial h2 ¼ .02

P at EOT

.66 (1, 58) partial h2 ¼ .01

e

2.23 (1, 58) partial h2 ¼ .04

e

.19 (1, 58) partial h2 ¼ .00

P at EOT

1.17 (1, 58) partial h2 ¼ .02

P at 6 mF and 12 mF; M at 6 mF

1.89 (2, 56) partial h2 ¼ .06 3.00 (1, 58) partial h2 ¼ .05 .09 (1, 58) partial h2 ¼ .00 2.77 (1, 58) partial h2 ¼ .05 .60 (1, 58) partial h2 ¼ .01 .63 (1, 58) partial h2 ¼ .01 1.35 (1, 58) partial h2 ¼ .02

1.12 (2, 56) partial h2 ¼ .04

.12 (1, 58) partial h2 ¼ .00 1.99 (1, 58) partial h2 ¼ .03 .86 (1, 58) partial h2 ¼ .02 .28 (1, 58) partial h2 ¼ .01 36.02*** (1, 58) partial h2 ¼ .38 .36 (2, 56) partial h2 ¼ .01 .45 (1, 58) partial h2 ¼ .01 .78 (1, 58) partial h2 ¼ .01 1.08 (1, 58) partial h2 ¼ .02 .28 (1, 58) partial h2 ¼ .01 .45 (1, 58) partial h2 ¼ .01 1.95 (1, 58) partial h2 ¼ .03

.01 (2, 56) partial h2 ¼ .00

P at EOT

.57 (1, 58) partial h2 ¼ .01

e

1.16 (1, 58) partial h2 ¼ .02

e

.10 (1, 58) partial h2 ¼ .00

P at EOT

4.24* (1, 58) partial h2 ¼ .07

M at 12 mF

.69 (1, 58) partial h2 ¼ .01

e

.34 (1, 58) partial h2 ¼ .01

P at EOT

BMI

.35 (1, 58) partial h2 ¼ .01

Relationship status

.07 (1, 58) .01 (1, 58) partial h2 ¼ .00 partial h2 ¼ .00

AN duration

1.27 (1, 58) 9.04** (1, 58) partial h2 ¼ .14 partial h2 ¼ .02

AN subtype

3.71 (1, 58) 3.09 (1, 58) partial h2 ¼ .06 partial h2 ¼ .05

Employment status 4.64* (2, 56) .59 (2, 56) partial h2 ¼ .14 partial h2 ¼ .02 Education

.14 (1, 58) .07 (1, 58) partial h2 ¼ .00 partial h2 ¼ .00

Medical concerns

2.02 (1, 58) .29 (1, 58) partial h2 ¼ .01 partial h2 ¼ .03

Psychotropic medication

8.29** (1, 58) .32 (1, 58) partial h2 ¼ .13 partial h2 ¼ .01

EDE Global

1.59 (1, 58) 1.04 (1, 58) partial h2 ¼ .02 partial h2 ¼ .03

ANSOCQ

.42 (1, 58) .82 (1, 58) partial h2 ¼ .01 partial h2 ¼ .01

WSAS

.38 (1, 58) 9.75** (1, 58) partial h2 ¼ .14 partial h2 ¼ .01

.18 (1, 58) partial h2 ¼ .00 .33 (1, 58) partial h2 ¼ .01 .52 (1, 58) partial h2 ¼ .01 6.27* (1, 58) partial h2 ¼ .10

.04 (1, 58) partial h2 ¼ .00 .36 (1, 58) partial h2 ¼ .01 2.04 (1, 58) partial h2 ¼ .03 .17 (1, 58) partial h2 ¼ .00 .07 (1, 58) partial h2 ¼ .00 .03 (1, 58) partial h2 ¼ .00

Note. BMI ¼ body mass index. AN ¼ anorexia nervosa. EDE ¼ Eating Disorder Examination. ANSOCQ ¼ Anorexia Nervosa Stages of Change Questionnaire. WSAS ¼ Weissman Social Adjustment Scale. P ¼ predictor. M ¼ moderator. *p < .05. **p < .01. ***p < .001.

individuals with AN-BP (p ¼ .036 and p < .001, respectively) across the two treatment conditions. Two moderators of depressive symptoms emerged (see Table 3). In particular, AN subtype and EDE Global scores emerged as moderators of treatment condition on depressive symptoms at 6-month follow-up and 12-month follow-up, respectively. Results revealed that individuals with AN-BP responded better to CBT-AN than SSCM at 6-month follow-up in terms of depressive symptoms. Additionally, those with higher EDE Global scores at baseline responded better to CBT-AN at 12-month follow-up in terms of depressive symptoms. Discussion We examined moderators and mediators of outcome in a sample of 63 adult women with SE-AN who participated in an RCT comparing CBT-AN and SSCM. Thirteen pre-randomization variables were examined in this exploratory analysis, with four moderators of outcome emerging. EDE Global and BDI scores moderated outcome on mental health (MCS) at EOT, age moderated outcome on mental health at 12-month follow-up, and AN subtype and EDE Global score moderated outcome on depression at 6-month and 12-month follow-up, respectively. More specifically, participants with higher EDE Global and BDI scores did better in CBT-AN at EOT on mental health outcomes. Older participants did better in CBT-AN at 12-month follow-up on mental health outcomes, as well. Participants with AN-BP did better in

CBT-AN at 6-month follow-up, and participants with higher EDE Global scores did better in CBT-AN at 12-month follow-up on depression outcomes. Whereas results from the main outcome study indicated statistically significant improvements in most clinical domains across both treatments, there were no differences between CBT-AN and SSCM (Touyz et al., 2013). For this exploratory analysis of moderators on treatment outcome, CBT-AN was more beneficial than SSCM when participants showed greater eating disorder symptomatology (as measured on the EDE), or more depression (as measured on the BDI), or when they presented with AN-BP or older age. Given that CBT-AN, compared to SSCM, is more focused on targeting and changing eating disorder behaviors (e.g., starvation, binge eating and purging), and cognitions (e.g., thinking one is fat, feeling depressed), it is not altogether surprising that CBT-AN seems to outperform SSCM in these domains. SSCM, by contrast, has a broader therapeutic focus and it is therefore not an entirely unexpected finding for CBT-AN to be superior to SSCM in this regard. It is, however, disappointing not to have identified a subset of patients for whom SSCM would be the preferred treatment. Such a finding would allow clinicians to provide clinical care in a more discerning manner in terms of matching patients with treatment. However, as this was an exploratory study, such a finding at this time is still premature. Several baseline variables were non-specific predictors of eating disorder-related quality of life, mental health, and depression outcomes at both EOT and for the maintenance period. These were age,

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

duration of illness, employment status, the use of psychotropic medications, AN subtype, and social adjustment. Younger age, briefer duration of illness, and better social adjustment all predicted improved outcomes at EOT. Lack of employment and use of psychotropic medications predicted poorer outcomes at EOT. Having AN-BP subtype also predicted poorer outcome for the maintenance period, as did the use of psychotropic medication. Several studies have provided guidance regarding baseline patient characteristics that may be predictive of eventual treatment outcome (e.g., Lock, Agras, Le Grange, et al., 2013; Lockwood et al., 2012; McIntosh et al., 2005; Pike et al., 2003). Findings from the current study were not surprising in that, among a population with SE-AN, a relative younger age, briefer duration of illness, better social adjustment, being employed, not being on medication, and having AN-R subtype all predicted improved outcomes. To some extent, this finding reinforces the primary finding from our main report (Touyz et al., 2013) in that this patient population can indeed make meaningful clinical gains, regardless of treatment modality, provided some or all of these parameters are met. However, what is interesting is that many of these variables were related to outcome only in the short-term (i.e., at EOT). Thus, even in this group, the influence of some factors, such as illness chronicity, on outcome was limited temporally. Strengths and limitations As is the case for most adult AN RCTs, the current study suffers from the limitation of moderate sample size. This only allowed for an exploratory moderator analysis, but was insufficient for a mediator analysis. In addition, the follow-up period of 12-months could be considered short for a participant group with SE-AN. We were also not able to conduct a closed follow-up, and for clinical reasons, participants could seek follow-up care as necessary. Most of the outcome measures were self-report, and given the possibility for potential biases in such self-evaluative assessments, more objective observations from relatives to corroborate the patients’ account would have been helpful. Finally, the limited modifications to both treatments have not yet been shown to be efficacious in treating the key outcome variables related to AN, and therefore probably restrict the generalizability of our findings. Despite these limitations, this remains the only RCT to date to exclusively recruit AN patients with an illness duration of seven years or more, use manualized treatments with supervision conducted on a weekly basis throughout the trial, record sessions for quality control, and retain 85% of the study sample in treatment and follow-up. Conclusions This study identified four moderators of treatment outcome: EDE Global scores, BDI scores, age, and AN subtype. For each moderator, whether it is higher eating disorder psychopathology, poorer mood, older age, or a binge/purge profile, patients did better if they received CBT-AN rather than SSCM. While it is helpful to have identified a subset of patients for whom CBT-AN would be advantageous, it was disappointing that no such patient group was identified for SSCM. Larger samples and interviewer-based outcome measures may strengthen future studies and will increase power and provide a better opportunity to detect such moderation. Findings from this study are exploratory in nature but nevertheless provide an important rationale for testing specific moderation and mediation effect hypotheses in future studies. If such studies are feasible, heterogeneous treatment effects on outcome for patients in different treatments and with different levels of psychopathology can be examined.

97

Acknowledgments Funding Support: Australian National Health and Research Council PG 457419 (Drs. Touyz, Le Grange, Lacey and Hay), Southwest London and St. George’s NHS Trust (Dr. Lacey), the Butterfly Foundation (Dr. Touyz), and University of Western Sydney (Dr. Hay). Dr. Le Grange receives royalties from Guilford Press and Routledge, as well as honoraria from the Training Institute for Child and Adolescent Eating Disorders, LLC.

References Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck Depression Inventory-II. San Antonio: Psychological Corporation. Castellini, G., Mannucci, E., Lo Sauro, C., Benni, L., Lazzaretti, L., Ravaldi, C., et al. (2011). Different moderators of cognitive-behavioral therapy on subjective and objective binge eating in bulimia nervosa and binge eating disorder: a threeyear follow-up study. Psychotherapy and Psychosomatics, 81, 11e20. http:// dx.doi.org/10.1159/000329358. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Dare, C., Eisler, I., Russell, G., Treasure, J., & Dodge, L. (2001). Psychological therapies for adults with anorexia nervosa: randomized controlled trial of out-patient treatments. British Journal of Psychiatry, 178, 216e221. http://dx.doi.org/ 10.1192/bjp.178.3.216. Eisler, I., Dare, C., Hodes, M., Russell, G., Dodge, E., & Le Grange, D. (2000). Family therapy for adolescent anorexia nervosa: the results of a controlled comparison of two family interventions. Journal of Child Psychology and Psychiatry and Allied Disciplines, 41, 727e736. http://dx.doi.org/10.1111/1469-7610.00660. Engel, S. G., Wittrock, D. A., Crosby, R. D., Wonderlich, S. A., Mitchell, J. E., & Kolotkin, R. L. (2006). Development and psychometric validation of an eating disorder-specific health-related quality of life instrument. International Journal of Eating Disorders, 39, 62e71. http://dx.doi.org/10.1002/eat.20200. Fairburn, C. G., & Cooper, Z. (1993). The eating disorder examination. In C. G. Fairburn, & T. Wilson (Eds.), Binge eating: Nature, assessment, and treatment (12th ed.) (pp. 317e360). New York: Guilford. Kraemer, H. C., Wilson, G. T., Fairburn, C. G., & Agras, W. S. (2002). Mediators and moderators of treatment effects in randomized clinical trials. Archives of General Psychiatry, 59, 877e883. http://dx.doi.org/10.1001/archpsyc.59.10.877. Le Grange, D., Eisler, I., Dare, C., & Hodes, M. (1992). Family criticism and selfstarvation: a study of expressed emotion. Journal of Family Therapy, 14, 177e 192. http://dx.doi.org/10.1046/j..1992.00451.x. Le Grange, D., Lock, J., Agras, W. S., Moye, A., Bryson, S. W., Jo, B., et al. (2012). Moderators and mediators of remission in family-based treatment and adolescent focused therapy for anorexia nervosa. Behaviour Research and Therapy, 50, 85e92. http://dx.doi.org/10.1016/j.brat.2011.11.003. Lock, J., Agras, W. S., Bryson, S., & Kraemer, H. C. (2005). A comparison of short- and long-term family therapy for adolescent anorexia nervosa. Journal of the American Academy of Child & Adolescent Psychiatry, 44, 632e639. http:// dx.doi.org/10.1097/01.chi.0000161647.82775.0a. Lock, J., Agras, W. S., Fitzpatrick, K. K., Bryson, S. W., Jo, B., & Tchanturia, K. (2013). Is outpatient cognitive remediation therapy feasible to use in randomized clinical trials for anorexia nervosa? International Journal of Eating Disorders, 46, 567e 575. http://dx.doi.org/10.1002/eat.22134. Lock, J., Agras, W. S., Le Grange, D., Couturier, J., Safer, D., & Bryson, S. W. (2013). Do end of treatment assessments predict outcome at follow-up in eating disorders? International Journal of Eating Disorders. http://dx.doi.org/10.1002/ eat.22175. Lock, J., Couturier, J., Bryson, S., & Agras, S. (2006). Predictors of dropout and remission in family therapy for adolescent anorexia nervosa in a randomized clinical trial. International Journal of Eating Disorders, 39, 639e647. http:// dx.doi.org/10.1002/eat.20328. Lockwood, R., Serpell, L., & Waller, G. (2012). Moderators of weight gain in the early stages of outpatient cognitive behavioral therapy for adults with anorexia nervosa. International Journal of Eating Disorders, 45, 51e56. http://dx.doi.org/ 10.1002/eat.20885. McIntosh, V. V. W., Jordan, J., & Bulik, C. M. (2010). Specialist supportive clinical management for anorexia nervosa. In C. M. Grilo, & J. E. Mitchell (Eds.), The treatment of eating disorders: A clinical handbook (pp. 108e129). New York: Guilford. McIntosh, V. V. W., Jordan, J., Carter, F. A., Luty, S. E., McKenzie, J. M., Bulik, C. M., et al. (2005). Three psychotherapies for anorexia nervosa: a randomized, controlled trial. American Journal of Psychiatry, 162, 741e747. http://dx.doi.org/ 10.1176/appi.ajp.162.4.741. McIntosh, V. V. W., Jordan, J., Luty, S. E., Carter, F. A., McKenzie, J. M., Bulik, C. M., et al. (2006). Specialist supportive clinical management for anorexia nervosa. International Journal of Eating Disorders, 39, 625e632. http://dx.doi.org/10.1002/ eat.20297. Perneger, T. V. (1998). What’s wrong with Bonferroni adjustments. British Medical Journal, 316, 1236e1238. http://dx.doi.org/10.1136/bmj.316.7139.1236.

98

D. Le Grange et al. / Behaviour Research and Therapy 56 (2014) 91e98

Pike, K. M., Walsh, B. T., Vitousek, K., Wilson, G. T., & Bauer, J. (2003). Cognitive behavior therapy in the posthospitalization treatment of anorexia nervosa. American Journal of Psychiatry, 160, 2046e2049. http://dx.doi.org/10.1176/ appi.ajp.160.11.2046. Rieger, E., Touyz, S. W., & Beumont, P. J. (2002). The Anorexia Nervosa Stages of Change Questionnaire (ANSOCQ): Information regarding its psychometric properties. International Journal of Eating Disorders, 32, 24e38. http://dx.doi.org/ 10.1002/eat.10056. Russell, G. F., Szmukler, G. I., Dare, C., & Eisler, I. (1987). An evaluation of family therapy in anorexia nervosa and bulimia nervosa. Archives of General Psychiatry, 44, 1047e1056. http://dx.doi.org/10.1001/archpsyc.1987.01800240021004.

Schafer, J. L. (1997). Analysis of incomplete multivariate data. London, UK: Chapman & Hall. Touyz, S., Le Grange, D., Lacey, H., Hay, P., Smith, R., Maguire, S., et al. (2013). Treating severe and enduring anorexia nervosa: a randomized controlled trial. Psychological Medicine. http://dx.doi.org/10.1017/S0033291713000949. Ware, J. E., Kosinski, M., & Keller, S. D. (1996). A 12-item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Medical Care, 34, 220e233. http://dx.doi.org/10.1097/00005650-199603000-00003. Weissman, M. M., & Bothwell, S. (1976). Assessment of social adjustment by patient self-report. Archives of General Psychiatry, 33, 1111e1115. http://dx.doi.org/ 10.1001/archpsyc.1976.01770090101010.