a r t i c l e i n f o

Article history:Received 11 December 2012Received in revised 7 November 2013

2013 Elsevier Ltd. All rights reserved.

. . . . . . . . . . 56

. . . . . . . . . . 56

Clinical Psychology Review 34 (2014) 5472

Contents lists available at ScienceDirect

Clinical Psychology Review3.2. Randomised Controlled Trials (RCTs) (Studies 15) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563.2.1. Prominent ndings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563.2.2. Methodological issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551.1. Rationale & aims of the present review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

2. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562.1. Inclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562.2. Search strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562.3. Study selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562.4. Data synthesis & appraisal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Corresponding author.E-mail address: sa172@leicester.ac.uk (S. Allan).

0272-7358/$ see front matter 2013 Elsevier Ltd. All rihttp://dx.doi.org/10.1016/j.cpr.2013.11.001evidence.Further research and ongoing review is needed to evaluate the settings, patient groups and formats inwhich CBTmay be effective as a treatment for AN.Accepted 18 November 2013Available online 27 November 2013

Keywords:Cognitive Behavioural TherapyAnorexiaEffectivenessa b s t r a c t

Evidence for the effectiveness of psychological therapies for anorexia nervosa (AN) is inconsistent. There havebeen no systematic reviews solely on the effectiveness for Cognitive Behavioural Therapy (CBT) for AN. This re-view aimed to synthesise and appraise the recent evidence for CBT as a treatment for AN. Using specic searchcriteria, 16 relevant articles were identied which evaluated CBT alone or as part of a broader randomised/non-randomised trial. Various formats of CBT were utilised in the reviewed papers. Studies were evaluatedusing established quality criteria.The evidence reviewed suggested that CBT demonstrated effectiveness as a means of improving treatment ad-herence andminimising dropout amongst patients with AN.While CBT appeared to demonstrate some improve-ments in key outcomes (body mass index, eating-disorder symptoms, broader psychopathology), it was notconsistently superior to other treatments (including dietary counselling, non-specic supportive management,interpersonal therapy, behavioural family therapy). Numerous methodological limitations apply to the available CBT showed promise in reducing dropout compared to other treatments. Patients treated with CBT showed improvements in physical and psychologic CBT did not appear to be superior to other types of treatment on these outcoCognitive Behavioural Therapy for anorexia nervosa: A systematic review

Lisa Galsworthy-Francis , Steven AllanDepartment of Clinical Psychology, University of Leicester, 104 Regent Road, Leicester LE1 7LT, UK

H I G H L I G H T S

al outcomes.mes.ghts reserved.

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

. .

psychotherapy (IPT), focal psychodynamic therapy and family interven- for AN but did not identify any additional studies to those in Kaplan

55L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472tions (NICE, 2009).A number of reviews have been conducted on the effectiveness

of psychological therapies for eating disorders (Bulik, Berkman,Brownley, Sedway, & Lohr, 2007; Hay, Touyz, & Sud, 2012;Kaplan, 2002; Lock & Fitzpatrick, 2007; Peterson & Mitchell,1999; Rosenblum & Forman, 2002; Rutherford & Couturier, 2007;Watson & Bulik, 2012; Wilson, 2005; Wilson, Grilo, & Vitousek,2007). The consensus of these reviews was of a paucity of evidence(specically RCTs) to support any particular treatment for adultswith AN. This is in contrast to bulimia nervosa, where CBT is con-sidered the treatment of choice (National Institute for Health andClinical Excellence [NICE], 2004).

There are methodological difculties in conducting RCTs with peo-ple with AN, particularly with respect to recruitment and compliance(Treasure & Kordy, 1998), so RCTs are relatively rare: making the at-tempt to reach for a gold standard of treatment for AN difcult to

(2002) and unsurprisingly, this review drew similar conclusions. Alater review by Bulik et al. (2007) identied one additional RCT whichsuggested that outcomes in the CBT condition were superior to one ofthe comparison treatments but inferior to a second. After summarisingthe methodological limitations of the reviewed papers, Bulik et al.(2007) concluded there was tentative evidence that CBT reduces re-lapse risk for adults, after weight restoration has been accomplished(p.317).

These previous reviews are themselves open to a number of meth-odological limitations. The Kaplan (2002) paper is a descriptive ratherthan a systematic review and it is unclear how papers were selected,assessed for quality and appraised. The Bulik et al. (2007) review wasmore systematic but employed a subjective and unvalidated ratingscale to evaluate strength and quality of evidence. While previous re-views represent considerable breadth of treatments, there are no specif-ic reviews of the effectiveness of CBT for AN. Given the rarity of (and3.3. Non-randomised clinical trials (Studies 67) . . . . . . . . .3.3.1. Prominent ndings . . . . . . . . . . . . . . . .3.3.2. Methodological issues . . . . . . . . . . . . . . .

3.4. Non-comparative clinical trials (Studies 816) . . . . . . . .3.4.1. Prominent ndings . . . . . . . . . . . . . . . .3.4.2. Methodological issues . . . . . . . . . . . . . . .

4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . .4.1. CBT and treatment acceptance/adherence . . . . . . . . . .4.2. CBT and eating disorder symptomatology . . . . . . . . . .4.3. CBT and broader symptomatology . . . . . . . . . . . . .4.4. Overall quality of reviewed studies . . . . . . . . . . . . .4.5. Clinical implications & suggestions for further research . . . .4.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . .

Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction

Anorexia nervosa (AN) is characterised by deliberate weight lossthrough food restriction and/or compensatory strategies including ex-cessive exercise, bingeing and purging. It is accompanied by a distortionof body image and an intense fear of gainingweight despite emaciation.Outcomes for individuals with AN have improved little in the secondhalf of the past century (Crow & Peterson, 2003), and AN continues tobe associated with poor prognosis and signicant physical and psycho-logical complications. This review intended to evaluate the evidence forone particular approach to the treatment of AN: Cognitive BehaviouralTherapy (CBT).

Longitudinal research has suggested fewer than 50% of individualsdiagnosed with AN recover fully; 2030% continue to experience resid-ual symptoms, 1020% remain signicantly ill and 510% die from theirillness (Steinhausen, 2002). Mortality rates in AN are ten times that ofthe general population (Morris, 2008), and are the highest of all psychi-atric disorders (Harris & Barraclough, 1998). Such statistics highlight theimportance of research into developing effective prevention and treat-ment strategies for AN.

Evidence for drug therapy alone in eating disorders is weakmoderate (Hay & Claudino, 2012); although low-dose antipsychoticmedication has been found to be benecial in some trials, it has longbeen recognised that treatments for AN need to target both physical(i.e. promotion of weight gain, reducing risk of physical complications)and psychological aspects of the disorder (e.g. working with disorderedcognitions, harmful behaviours, body image issues and associated emo-tional disturbances). Current guidance suggests a range of psychologicaltherapies to consider for the treatment of AN, including cognitive ana-lytic therapy (CAT), cognitive behaviour therapy (CBT), interpersonal

Referencesachieve (Goldstein et al., 2011, p.29). NICE (2004) made over 100. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

recommendations for eating disorders. CBT for bulimia and binge-eating disorder received strong empirical support, however no specicrecommendations were made for AN.

CBT seeks to help patients overcome difculties by identifying andaltering dysfunctional thinking, behaviour, and emotional responses/behaviours. CBT has been shown to be effective in treating many ofthe problems which are often a feature of AN (depression, anxiety,low self-esteem, obsessions/compulsions). Cognitive and attentionalbiases towards food/eating/shape-related stimuli are a signicant fea-ture in eating disorder presentations (e.g. Brooks, Prince, Stahl,Campbell, & Treasure, 2011), therefore CBT would appear to be a logicalchoice for treatment. Furthermore, the stylistic features of CBT(structured, time-limited, directive, focused on the present) appearsuited to the typical individual with AN who is described as com-fortable with order and control, and not prepared to delve into thepast (Freeman, 2002). CBT appears to have been accepted by profes-sionals as a useful intervention for AN. Herzog et al. (1992) reportedthat 8892% of clinicians at eating disorder conferences consideredCBT (alone or combined with a psychodynamic approach) to be indi-cated in AN. However, despite the apparent theoretical suitabilityand acceptability of CBT for AN, evidence for its effectiveness islimited.

Previous reviews have evaluated the evidence for a range of treat-ments for ANwithCBT as one type of treatment. Kaplan (2002) reportedthree RCTs which included CBT. Two of these RCTs suggested a positiveeffect on outcome for CBT compared to other treatments,while the thirdstudy showed no difference in outcome between treatments. However,Kaplan (2002) noted the methodological limitations of these studies(e.g. small samples, power issues, the impact of dropout on results). ACochrane review (Hay et al., 2003) evaluated multiple psychotherapiesmethodological difculties associated with) RCTs in AN, previous

56 L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472reviews may also be overly-limiting in their exclusion of studies whichdeviate from strict RCT procedures.

1.1. Rationale & aims of the present review

In an attempt to overcome some of the limitations of previous re-views and to provide an up-to-date synthesis of the evidence withoutexcessive duplication, the current paper sought to review the recent lit-erature in order to appraise the evidence for CBT for AN. This shouldcontribute to a more informed understanding of the effectiveness ofCBT in the treatment of this important disorder. Unlike previous re-views, the present review focused solely on CBT as the treatment of in-terest. Also, the current paper included studies which utilised designsother than the RCT given that such highly controlled procedures areoften not pragmatically or ethically possible in clinical practice, andbecause alternative methodologies may be considered a helpful contri-bution to the evidence base in a limited area.

2. Method

2.1. Inclusion criteria

Anumber of a priori limitswere set. Firstly, papersmust be in English(for pragmatic reasons). A date cut-off was set at 1995present to ex-pand upon previous reviews by including more recent studies, and toavoid duplication of discussions in earlier reviews. Papers must bepeer-reviewed journal articles (with the expectation that minimumquality standards have been met through the peer review process).Papers must describe methodologies at a higher level than the singleclinical case (for increased generalisability), andmust employ quantita-tive designs (to allow for objectivemeasurement of treatment effective-ness). Finally, papers needed to include at aminimummeasurements attwo time points (pre and post-intervention).

2.2. Search strategy

A scoping exercise was conducted prior to the main search, inorder to gauge the amount, type and breadth of the available litera-ture and to identify previous reviews. This exercise informed thesearch terms and shaped the focus of the review. The scoping exer-cise also identied that much of the research in this area combinedadolescent and adult samples, making it difcult to focus upon aspecic age group.

Keywords selected for searches included the terms anorexia andCBT or cognitive behav* therapy (truncation applied to include var-iations on cognitive behaviour therapy such as behavioural, andalso to include non-American spelling of behaviour/behavioural).At the initial search stage, specic study outcomes (e.g. effect ofCBT on weight, dysfunctional thoughts etc.) were not includedwithin the search terms given their heterogeneity (as revealed bythe scoping exercise).

The search terms and a priori limits were applied withinthe databases Scopus, PsycInfo (incorporating PsycArticles andPsycExtra), Science Direct and Ovid SP (incorporating Ovid Medlineand Embase). The Cochrane Library database was searched forexisting reviews.

2.3. Study selection

Fig. 1 shows the shortlisting process. Only articles dealing with ANwere considered for focus; however articles which discussed AN treat-ment separately (within a paper on general eating disorders) were in-cluded. If evaluation of a CBT programme was presented as part of awider paper, this was included.

Papers were included if they explicitly referred to using cognitive

behavioural interventions/principles, and gave details of the corecomponents of the treatment. The particular format of CBT programmes(i.e. length, individual or group-based, inpatient or outpatient) did notpreclude selection.

Further papers were excluded if found to group eating disorderdiagnoses together (i.e. preventing analysis for AN separately) or ifthey described multidimensional treatment programmes which didnot allow for CBT to be examined in isolation.

2.4. Data synthesis & appraisal

Key features of each paper (aims, design, format, sample, key results)were extracted and summarised in tabular format (see Tables 13).Morethorough appraisal was guided by a quality assessment tool, whichallowed closer examination of the methodological soundness of eachstudy.

Due to variations in therapeutic content and length, age of samplesand other heterogeneous features, meta-analysis was not possible. Anarrative discussion of papers follows.

3. Results

3.1. Overview

The nal 16 papers consisted of 5 randomised controlled trials, 2non-randomised controlled trials and 9 individual clinical trials (caseseries trials, no comparison group). For the purpose of the presentreview, papers have been grouped by design type (RCTs, non-randomised controlled trials, individual clinical trials) and for eachdesign type, all studies are summarised in tabular form, followed by adiscussion of main study ndings (e.g. treatment adherence/attrition,effect of the treatment on physical, ED-specic and broader outcomes)and nally a critique of methodological issues. More details of theCBT programmes delivered in each study are presented in Table 4(appendix).

3.2. Randomised Controlled Trials (RCTs) (Studies 15)

Characteristics of samples are presented in Table 1. Sample sizeranged from 25 to 167, with a total of 316. A range of ages were repre-sented. Of the studies reporting gender (14), 94% were female. Allstudies included both subtypes of AN (restricting or bingepurge). De-tails of treatment settings and outcome measures are also presented.For the RCTs, all CBT programmes were conducted on an outpatientbasis, utilising an individual (11), manualised approach. Ethnicity ofparticipants was not reported in any of the RCTs.

3.2.1. Prominent ndings3.2.1.1. Adherence/Attrition. Dropout was a factor in all RCTs; reasons in-cluded relocation, hospitalisation, treatment refusal and dropout due toperceived improvement. Study 1 reported equal dropouts in eachgroup, with no signicant pre-treatment differences between com-pleters and non-completers and no signicant association betweentype and likelihood of completion. Study 3 found a signicant differencein mean weight at baseline for completers and non-completers; therewere no group differences on the likelihood of completing therapy. Incontrast, study 5 reported signicantly higher voluntary dropout (be-fore session 10) from dietary counselling (3 out of 15; 20%) comparedto CBT (0). Study 4 found higher adherence (74.5%) with specialist out-patient treatment involving CBT, compared to 49.1% for inpatient treat-ment and 69.1% for non-specialised outpatient treatment. There wereonly two dropouts in the cognitive therapy group in study 2, with allof the dietary counselling group disengaging.

3.2.1.2. Effect of CBT on physical outcomes. Of the RCTs reporting BMI as

an outcome, all demonstrated increases following CBT. Improvements

h h

ousrap

57L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472Initial searc

Titles scanned for relevance (exclusion of obvion BN/BED/EDNOS, articles discussing theover time were statistically signicant in study 1. However there wereno statistically signicant differences across groups and similar im-provements over time were observed in the comparison treatment.Study 2 also reported a statistically signicant increase in BMI followingcognitive therapy. However due to attrition, post-treatment BMI forthe comparison group was unobtainable and so it was not possibleto compare which treatment showed most gain. Study 3 did notnd increases in BMI over time to be statistically signicant, norwere there differences across groups. Actual weight followed asimilar pattern. Study 4 did not analyse the statistical signicance

Export to Refwo

Collation and removal o

Abstract retrieN = 65

Abstract scanning (exclusion of rdescription, discussion papers, stu

closer investig

Full Text retrieN = 40

Quality assessment too

Final review: N

Exclusion of papers where CBT was not an indivwhere AN grouped with other diagnoses/not

treatment comparison u

Fig. 1. Shortlistiits

ly non-relevant articles, articles focused solely ies other than CBT, articles not in English)of improvement in BMI over time, however found no difference be-tween treatment conditions. Interestingly none of these results dem-onstrated statistically signicant differences between CBT andcomparison treatment(s).

3.2.1.3. Effect of CBT on ED symptomatology. Study 1 found signicantmain effects for time on measures/subscales of eating-disordered pa-thology; there were no differences between treatment conditions and,despite improvements over time, scores remained in the clinical rangeat follow-up. Study 2 also found those in the CT group showed

rks

f duplicates

val

eviews, theory/treatment dies where not relevant on ation)

val

l applied

= 16

Reference list

scanning

idual treatment; prospective studies; examined individually; pre/post navailable

ng process.

Table 1Randomised controlled trials.

StudyID

Author(s)& date

Aims of study Design Format oftreatment

Sample/Participants Key results of study

1 Ball andMitchell(2004)

To investigate theeffectiveness of CBTfor anorexia inoutpatientadolescents/young adults

Between groups (treatment type)and within groups (time)comparisons 2 groups: CBT vs. Behavioral FamilyTherapy (BFT) Randomisation to treatment group(process not described)

Outpatient Individual

Manualised

Total N = 25 (13 CBT, 12 BFT) Completers = 18 (9 CBT, 9 BFT) All female outpatients Ages 1323, all lived with family CBT group: 7 AN-R, 6 AN-BP Mean age of CBT group = 18.45 years 64% had received some sort of treatment prior to studyentry No signicant pre-treatment differences between treat-mentgroups on demographics/BMI DSM diagnosis met; also included subthreshold cases(individuals weighing between 85 and 90% of normalweight for age and height) Included both subtypes of ANSpecic exclusion criteria: BMI b 13.5 Currently receiving other treatments Comorbid physical/psychiatric disorder (not includingdepression/anxiety secondary to AN) Current drug/alcohol abuse Self-harm in last 12 months Indications for hospitalisation (suicidal ideation, severephysical complications) Untreated trauma/abuse

No signicant association between type of therapy and likelihoodof completionPhysical measures: Primary outcome: no difference across treatments at post-treatmentor follow-up BMI: signicant effect for time (improvement from pre-posttreatment) however no effect of treatment typeED-specic: Signicant effect for time (improvement from pre-post treatment)on all measures in both treatment groups, however scores remainedin clinical range at follow-upBroader psychopathology: Signicant improvement over time in both groups for anxiety anddepression, with changes maintained at follow-upOther: Signicant differences in self-esteem over time (maintained atfollow-up) for both treatment groups, although remaining below average Trends suggestive of greater improvements in family functioningin CBT group, but BFT group showed trend towards less negativecommunication

2 Serfaty,Turkington,Heap,Ledsham,and Jolley(1999)

To evaluate theeffectiveness ofcognitive therapyversus dietarycounselling inoutpatients

Predominantly within-groupscomparisons (time) 2 groups: Cognitive Therapy (CT) vs.dietary counselling (DC) Randomisation to groups (processadequately described)

Outpatient Individual

Manualised

Total N = 35 (25 CT, 10 DC) 33 females, 2 males (both in CT group) Completers = 23 (23 CT, 0 DC) Only new GP referrals (excluding CMHT andinterprofessional referrals) 28 restrictive type (all DC group, 18/25 CT group); 7 inCT group bulimic subtype No signicant pre-treatment differences betweentreatment groups on key variables, except duration ofillness (CT group signicantly longer duration)Specic inclusion criteria: 16+ years of age DSM diagnosis met and conrmed

(CT completers only)Physical measures: Signicant improvement in BMI from initial assessment to follow-upED-specic measures: Signicant improvement from initial assessment to follow-upBroader psychopathology: Depression: signicant improvement from initial assessment tofollow-upOther: Dysfunctional attitudes: no signicant difference from initialassessment to follow-up Locus of control of behaviour: signicant improvement from initialassessment to follow-up Overall, of CT completers 70% no longer met diagnostic criteria; 87%had increased BMI

58L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

3 McIntoshet al.(2005)

To compare theeffectiveness ofCBT, interpersonalpsycho-therapyvs. controltreatment

Comparison across 3 groups: CBTvs. Interpersonal psychotherapy (IP)vs. non-specic supportive clinicalmanagement (NSCM; control group) Randomisation to treatment group(procedure not described)

Outpatient Individual

Manualised

Total N = 56 (19 CBT, 21 IP, 16 NSCM) Completers = 35 (12 CBT, 12 IP; 11 NSCM) Broad referral base including self-referralSpecic inclusion criteria: Female 1740 years old Primary diagnosis of AN (also included BMI 17.519)Specic exclusion criteria: BMI b 14.5 Current severe major depression Psychoactive substance dependence Major medical/neurological illness Developmental learning disorder/cognitive impairment Bipolar disorder Schizophrenia Chronic refractory course of AN

No group differences on likelihood of completionALL PPs Primary outcome: NSCM superior to IP; no difference IP vs. CBT; nodifference CBT vs. NSCM Secondary outcomes: no signicant difference in physical measuresor eating disorder specic subscales; Restraint subscale only signicantlydifferent to baseline, with both CBT and NSCM superior to IP, butNSCM superior to CBT; signicant differences on global functioning,with NSCM superior to both IP and CBT, and CBT superior to IPCOMPLETERS ONLY Primary outcome: NSCM superior to both IP and CBT; no differenceIP vs. CBT Secondary outcomes: no differences in physical measures; signicantdifferences on all subscales of one eating specic measurerestraintand shape concerns NSCM superior to IP, eating and weight concernsNSCM and CBT superior to IP; drive for thinness NSCM superior to IP;on measures of global functioning, NSCM superior to CBT and IP,no difference between CBT and IP

4 Gowerset al.(2007)

To evaluate theeffectiveness of 3treatments foradolescents withAN

3 groups: Inpatient, specialistoutpatient and general CAMHSoutpatient care

Outpatient Individual

Manualised

Multicentre Referral and identication via audit Total N = 167 (57 inpatient, 55 specialistoutpatient, 55 treatment as usual) Ages 1218 153 (92%) females Modied DSM diagnosis met Mixed duration and subtype

Treatment adherence varied between groups: inpatient 49.1%adherence, specialist outpatient 74.5%, treatment as usual 69.1%Outcomes at 1 year All groups substantial improvement on weight, global measures,self-reported psychopathology No statistically signicant differences between groups on any measures Relatively poor outcome for inpatient group Fewer than 1 in 5 fully recoveredOutcomes at 2 years Further improvement in all groups No statistically signicant differences in groups on any measure One third recovered

5 Pike,Walsh,Vitousek,Wilson,and Bauer(2003)

To evaluate theeffectiveness ofCBT as a posthospitalisationtreatment for AN

2 groups: CBT and nutritional/dietary counselling (DC) Randomisation to treatmentgroup (procedure adequatelydescribed)

Outpatient Individual

Manualised

Total N = 33 (18 CBT, 15 DC) Adults only, ages 1845 All pps had completed inpatient weightrestoration treatment prior to study Rates of restricting vs. binge/purge subtype werenot signicantly different between groups; in CBTgroup 56% AN-R, 44% AN-BP No signicant differences between groups onbaseline characteristics

Signicantly less relapse in CBT group (p b .004) and remained in treatmentlonger 53% of DC met criteria for relapse in 1 year follow-up, vs. 22% CBT Higher voluntary dropouts for DC vs. CBT (p b .05) Higher total dropouts for DC vs. CBT (p b .003) Higher percentage of pps in CBT group met good outcome criteria

Abbreviations/Acronyms used: AN-BP= Anorexia Nervosa binge-purge subtype; AN-R= Anorexia Nervosa restrictive subtype; BMI= BodyMass Index; CAMHS= Child and Adolescent Mental Health Service; CMHT= Community Mental HealthTeam.

59L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

60 L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472signicant changes in scores on EDmeasures over time. Study 4 demon-strated improvements from baseline to one and two-year follow ups onED measures, although did not report statistical analysis of scores overtime. Therewere no signicant differences between the three treatmentconditions on these variables.

In analysing the total sample, Study 3 showed improvement frombaseline to end of treatment with CBT on all subscales of oneED-specic measure, however only scores on one subscale (Restraint)reached statistical signicance. Post-hoc tests showed that both CBT andNSCM were each superior to IP. There were also improvements(although not statistically signicant) from baseline on all but onesubscales of another ED measure following CBT. For completersonly, there were signicant differences on all four subscales of onemeasure, and on the Drive for Thinness subscale of the other EDmea-sure used; post-hoc tests indicated that NSCM was superior to IP forsome subscales, while for others both NSCM and CBT were superiorto IP.

Overall, these studies suggested some improvement following CBTon measures of eating disordered symptoms, but these differenceswere not superior to other treatments.

3.2.1.4. Effect of CBT on general psychopathology and functioning. Study1 reported a signicant decrease on both depression and anxietyfrom pre- to post-treatment with CBT (and BFT), with changes main-tained at follow-up. Study 2 corroborated results for depression, andreported a signicant difference for the CT group. Using an alterna-tive measure, study 3 also found improvements in depressionfollowing CBT (and both other treatments), and study 4 found a re-duction over time in a measure of mood symptoms in all treatmentgroups.

CBT (and both other treatments) improved scores on global func-tioning in study 3, with a signicant difference over time; there wasalso a signicant overall difference among groups, with post-hocanalysis indicating NCSM to be superior to both IP and CBT. Study 1also found statistically signicant improvements over time for self-esteem for both CBT and BFT groups, although these remained inthe below average range. Study 2 reported improvement (althoughnot signicant) in dysfunctional attitudes, and statistically signi-cant improvement on participants' locus of control of behaviour.Study 4 found improvements in both parent and child-rated scoresfor general functioning over time, but no signicant differencesbetween groups were found.

In terms of overall functioning, study 1 reported 77.8% of both CBTand BFT groupsmet criteria for good/intermediate outcome followingtherapy, maintained at follow-up. There were also signicant main ef-fects (for time only) on a specic measure of outcome. Using similarcriteria, this pattern was corroborated in study 4 with differences inscores over time, although no signicant differences were reportedbetween groups. In contrast, study 5 reported a signicant differencebetween groups in meeting criteria for good outcome (44% in CBTgroup vs. 7% in DC); however authors' modied criteria for full recoverywas not signicantly different between groups (met by 17% in CBTgroup, none in DC).

To summarise, all studies reported some improvements inmeasuresof mood following CBT. However, as with other ndings, evidence forCBT as superior to other therapies (BFT, Dietary/Nutritional Counselling,IP, Non-Supportive Clinical Management) was weak.

3.2.1.5. Effect of CBT on family functioning. Of the two studies which in-cluded family functioning as an outcome, results were unclear.While study 1 found trends towards improved communication inthe CBT group from pre- to post-treatment, the BFT group showedgreater trends from post-treatment to follow-up, suggesting slowerbut steadier progress. Changes on measures of family functioningacross time and treatment type in study 4 were small and no clear

patterns emerged.3.2.2. Methodological issuesWhile the RCT is widely regarded as the gold standard in research,

certain recommendations for quality were not met by these studies.Blinding was not possible: therapists were aware of which therapythey were delivering. Only studies 2 and 5 described the randomisationprocedure adequately. Presentation of baseline data and outcomedata was inconsistent and often incomplete, which prevented key com-parisons from being made. Participant numbers were low and the ab-sence of a no-treatment or waiting list control group (presumably forethical reasons) questions the status of these studies as true RCTs andmade it difcult to be certain that any differences were due to therapyand not other factors.

Characteristics of the samples may have introduced additionalbiases threatening validity. For example mixed severity and dura-tion of illness was present in all studies; these variables have beenshown to impact on treatability and prognosis (Steinhausen, 2002;Treat et al., 2005). More specically, the sample in study 4 includedinpatients who had already failed outpatient treatment, likely tohave had poorer prognosis from the outset; study 1 reported 64%of patients had received some sort of treatment before, raising ques-tions regarding prior exposure confounding results. Study 1 includ-ed sub-threshold cases, making comparison with other samples(where all diagnostic criteria were met) difcult. The inclusion ofadolescents and adults within studies made it difcult to disentan-gle the comparative effectiveness of treatments for the younger ver-sus the older patient with AN.

The different outcome measures employed in the studies made itdifcult to nd consistent patterns in results. Alterations made to stan-dard outcome measures also shed doubt on validity of such measures,for example modication of the MROS for adolescents in study 4. Alsothe large number of dependent variables (measures/subscales ofmeasures) studied may have restricted the power to demonstratedifferences between treatments.

Attrition may have introduced bias as dropouts have been shown todilute the effect of treatment (Ellenberg, 1994), and completers fromdifferent groups may not be comparable. Analysing only those whocompleted therapy may undo the benets of randomisation (thus af-fecting internal validity). Intention to treat analyses (as used in threeRCTs) may also be misleading by virtue of the methods of imputationemployed.

Overall, the results of these RCTs suggested that CBT may bemore effective than other treatments in reducing treatment drop-out. In terms of physical, eating-disordered and broader psycho-pathological outcomes, CBT did lead to some positive changes butit was unclear whether CBT was any more effective than othertreatments.

More recently, Carter et al. (2011) followed up 77% of the originalparticipants in study 3 (mean 6.7 years post-treatment; minimum5 years) and reported no signicant differences on any pre-selectedprimary (clinician ratings of outcome), secondary (physical andED-specic) or tertiary (broader psychological) measures across the 3different treatments. Weight and BMI had increased across all groupsfrom post-treatment levels, however only 49% of the total samplewere rated as having a good outcome at follow-up. Differences werefound in terms of patterns of change over time; whereas 75% of partic-ipants who received non-specic supportive clinical managementwere rated as having a good global outcome by clinicians post-treatment, this dropped to 42% at follow-up; these gures were33% (post-treatment) and 41% (follow-up) for CBT, and 15% (post-treatment) and 64% (follow-up) for interpersonal therapy. While it isnot known when these changes occurred, these results suggestthat CBT may have a more stable course than other treatments.Post-treatment, patients who had received CBT had an intermediateglobal outcome rating (not signicantly different from comparisontreatments), and were likely to have improved by long-term follow-

up (though not as commonly as those who received IPT). A similar

Table2

Non

-rando

mised

clinicaltrials.

Stud

yID

Autho

r(s)

&date

Aim

sof

stud

yDesign

Form

atof

treatm

ent

Sample/Participants

Key

resultsof

stud

y

6Fernndez,Turn,

Siegfried,

Meerm

ann,

andVallejo(199

5)

Todeterm

inethe

effectiveness

ofmulti-m

odalCB

Twithaddition

albody

therapy

CB

Twithbody

therapyvs.BFT

withno

body

therapy

Bo

thwithin-

subjectsand

betw

een-subjectscomparisons

Inpatient

Co

mbination

ofindividu

alandgrou

pform

ats

ANon

lyTotalof3

8inpatientswithAN(D

SM-IIIcriteria)

matched

forageacrossgrou

psN=

19in

CBTgrou

pAllfemale

Signicant

difference

inheight/w

eightacross

grou

psbu

tno

tBM

INoothersignicant

differencesbetw

een

grou

pson

demograph

ic/clin

ical/psychom

etric

characteristics

CB

Tgrou

ptook

sign

icantlylonger

toreachtarget

weigh

tthan

BFT

grou

p(14.4mon

thsvs.9.7mon

ths)

Atp

ost-treatm

ent:

Signicant

increase

inBM

Ifor

both

grou

psSignicant

redu

ctionin

globalED

measure

scores,and

5of

8subscales

ofanotherED

measure,for

both

grou

psSignicant

redu

ctionin

depression

scores

inboth

grou

psGains

generally

maintainedforboth

grou

psGoodou

tcom

epo

st-treatment(+follow-up):32%

(30%

)CB

T,10%(20%

)BFT

Interm

ediate:42%(30%

)CB

T,38

%(10%

)BFT

Poor:

26%(40%

)CB

T,52

%(70%

)BFT

20

drop

outs(8

CBT,12

MTA

U)

nosign

icant

diffe

rencein

drop

outs

7Carter

etal.(20

09)

Tocompare

therate

andtimingofrelapse

forp

atientsreceiving

oneof

two

maintenance

treatm

entsforAN

Tw

ogrou

ps:C

BTand

maintenance

treatm

ent

asusual(MTA

U)

Betw

een-grou

pscomparison

Outpatient

Individu

alManualised

ANon

lyTotalN

=88

(46CB

T,42

MTA

U)

Allpp

shadachieved

weigh

trestoration

(BMI

of19.5)followingspecialised

hospitalprogramme

Allfemales

Meanage=

24.1years(SD=

5.1)

37

pps(42%

)binge/pu

rgesubtype;51

(58%

)restrictingsubtype

Ethn

icity,maritalstatus,age

ofon

set,du

ration

ofillness

allreported

Relapsedenedas

BMI17.5for3mon

ths:

Timeto

relapsesignicantlylong

erin

CBTgrou

pthan

MTA

U(p

=.05)

At1

year

24.4%of

CBTrelapsed

vs.50%

ofMTA

URelapsedenedas

aboveor

resumptionof

bing

eing/purging:

Timeto

relapsesignicantlylong

erin

CBTgrou

pthan

MTA

U(p

=.007)

At1

year

32.5%of

CBTrelapsed

vs.65.6%

ofMTA

U65%of

CBTvs.34%

ofMTA

Uremainedremittedat1year

61L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472pattern was also found for secondary and tertiary outcome measures.This study proves how longer-term follow-up of RCTs in AN is essentialto monitor treatment effectiveness.

3.3. Non-randomised clinical trials (Studies 67)

Details of samples, settings and outcomemeasures used are present-ed in Table 2. In total, the non-randomised trials sampled 126 femaleswith AN, 65 of whom received CBT. One of the non-randomised clinicaltrials was conducted on an inpatient basis with both group-based andindividual elements; the second was on an outpatient basis and wasindividual treatment only. Ethnicity was reported in one study(see Table 2 for further information).

3.3.1. Prominent ndings3.3.1.1. Adherence/Attrition. Study 6 did not report attrition, implying all38 participants (19 in each treatment condition) remained in treatment.However at follow-up data was available for only 10 participantsfrom each treatment group. In study 7, a total of 20 participants(22.72%) dropped out of the study, 8 from CBT treatment and 12from MTAU. There was no signicant difference in dropout fromthe two groups. Follow-up data were still available for the major-ity of CBT dropouts.

3.3.1.2. Effect of CBT on physical outcomes. There was a signicant in-crease in BMI over time for both treatment groups in study 6, althoughparticipants in the CBT group took signicantly longer to reach theirtarget weight.

3.3.1.3. Effect of CBT on ED symptomatology. In study 6, both treatmentgroups showed reduced ED symptoms over time, with signicant im-provement on a global ED measure, and on most subscales of anothermeasure. When measured categorically, the CBT group showed betteroverall outcomes than the BFT group.

3.3.1.4. Effect of CBT on general psychopathology and functioning. Bothtreatment groups in study 6 showed a signicant reduction in depres-sive symptoms over time.

3.3.1.5. Relapse rates. In study 7, 65% of participants who received CBTmaintenance treatment remained remitted at one year compared to34% who received MTAU. Whether dened in terms of solely BMI orBMI plus eating behaviours, CBT proved signicantly more effectiveat preventing relapse. In study 6, both CBT and BFT groups partiallymaintained gains at follow-up.

3.3.2. Methodological issuesNon-randomised trials are an important alternative to RCTs when

the latter are not practically possible or ethically appropriate (Black,1996). However, by not randomly allocating participants to treatments,the possibility that selection bias and pre-treatment differences mayaffect results increases. For example in study 7, it is possible that thosewho consented to further active treatment (CBT group) were alsothose with greater motivation to change and so may have had a betterprognosis than non-consenters. Additionally, this study did not distin-guish between participants whose pre-maintenance (weight restora-tion) treatment was inpatient or outpatient based and thereforeinherent differences in setting may have affected participants.Non-randomisation in study 6 meant that the two interventions tookplace in different countries. Differences in services, training of therapistsand cultural differences across these two countries may therefore haveinuenced results.

The authors of study 7 recognised that MTAU was not controlled,and does not act as a no aftercare treatment alternative. In fact 97.1%

of the MTAU group sought at least one form of follow-up treatment

Table 3Non-comparative clinical trials.

StudyID

Author(s)& date

Aims of study Design Format oftreatment

Sample/Participants Key results of study

8 Leung, Waller,and Thomas(1999)

To investigate the effectivenessof group CBT for AN, and whethercore beliefs predict outcome

Pre-post within-groupscomparison

Outpatient Group Manualised

AN only Total N at start of study = 30 (completers = 20) All female Age range 1750 Mean age 26 (SD = 7.66) 6 (30%) binge/purge subtype; 14 (70%) restricting subtype

Eating behaviours/cognitions: No signicant differences from pre-post treatment onany subscale of ED measure Non-signicant reduction in anorexic cognitionspost-treatment

9 Bowers andAnsher(2000)

To examine cognitions (automaticthoughts and schemas) in AN,before and after treatment

Within-groups comparison Inpatient Mixedindividual andgroup/familywork

AN only 32 Caucasian adults 29/32 (91%) female 19/32 AN-R, 13/32 AN-BP Mean age 27.8 years Mean education, days in hospital, age of onset, duration ofillness, previous hospital admissions provided AN diagnosed by DSM-IV criteria

BMI increased from (sample mean) 16.7 at admissionto 20.5 at discharge (statistics not calculated) Signicant reduction on measure of automatic thoughtsfollowing treatment Signicant reduction in anorexic cognitions Signicant reduction in many subscale scores onmeasure of maladaptive schema

10 Dalle Graveet al. (2007)

To investigate the effect ofinpatient CBT on temperamentand character of ED patients

Pre-post comparison 3 diagnostic groups(NB: only AN group reportedfor purpose of present paper)

Inpatient plusresidential dayhospital Mixedindividualand group/family work Manualisedprotocoladapted forinpatientsetting

AN, BN, EDNOS sub groups Total N = 149 136 females, 13 males Adults and adolescents Active substance abuse or psychosis excludedAN group: N = 60 (40.3% of total sample) Mean age 24.6 years, age of onset 16.9 years, 2.5 previousinpatient admissions 12 did not complete full 20 weeksSpecic exclusion criteria: Active substance abuse Schizophrenia/psychosis

Signicant treatment effect on measure of temperamentand character (including harm avoidance, persistence,self-directedness) Mean BMI increased signicantly from 14.5 on admissionto 19.6 at discharge (p b .001) Signicant reduction in reporting of objective andsubjective bingeing episodes, frequency of self-inducedvomiting, laxative misuse, excessive exercise Signicant reduction on scores for restraint, eating concern,weight concern and shape concern Signicant reduction in scores for depression

11 Bowers andAnsher(2008)

To assess changes in ED andgeneral psycho-pathologyfollowing inpatient treatment,and at one-year follow-up

Within-groups comparisonacross three timepoints

Inpatient Individual andgroup formats plusfamily work Manualisedprotocol modiedfor inpatients

AN only Total N at start of study = 32 100% Caucasian 29 (91%) females 3 (9%) males Mean age 27.8 years Average 14.1 years' education Mean age of onset of AN: 18.2 years Mean duration of AN: 9.6 years Average hospital stay at recruitment: 63.7 days 13 (40%) binge/purge subtype; 19 (60%) restricting subtype

Following treatment: Signicant reduction on global ED measure Signicant differences on 8 of 11 subscales of a secondED-specic measure Signicant differences on 5 of 13 scales measuringpersonality traits Signicant differences on 2 measures of depression Mean BMI 16.7 at admission, increased to 20.5 at dischargeAt follow-up: Signicant change sustained on many EDmeasures/subscales and depression

12 Brambillaet al.(2010)

To see whether CBT modies thesecretion of central DA, NE and5HT and if physical/psychologicaleffects of CBT correlate withthese changes

Mixed within-groups(specic diagnoses) andbetween-groups comparison Random recruitment

As in DalleGrave et al.(2007)

AN-R, AN-BP and BN subgroups Total sample N = 50; AN = 28 (14 AN-BP, 14 AN-R)remainderBN (not reported here) All aged 18+ All female AN-R mean age 27 years; AN-BP 22 AN-R mean age of onset 16 years; AN-BP 18 years AN-R mean duration of AN 125.8 months; AN-BP 119.5 months All previously received outpatient treatment with no benetSpecic exclusion criteria: Medical conditions not linked to AN (endocrine or metabolicdisorders, epilepsy, head injury) Substance abuse Comorbid psychiatric disorders

Physical: AN-R group pre-treatment BMI 13.7 rising to19.4 post-treatment AN-BP group pre-treatment BMI 16.5 rising to20.0 post-treatment Both BMI increases reached signicance No signicant changes in DA, 5HT or NE after CBTin either AN subtypePsychological: Signicant improvements for both AN subtypes onED symptoms, depression, anxiety, impulsiveness,self-esteem at the end of CBT

13 Ricca et al.(2010)

To evaluate the effectiveness ofindividual CBT for threshold andsub-threshold AN, and identifypotential predictors for outcome

Both within groups(measurement at differenttimepoints) and betweengroups (threshold and

Outpatient Individual Manualised

Subthreshold and clinical AN subgroups Total N = 103 53 diagnosed AN (diagnostic interview for DSM-IV criteria);mean age 27.48

10/53 AN patients (18%) withdrew/did not completetreatment; 2 were not available for follow-upAt end of treatment: 19 (37%) recovered (did not meet DSM criteria for ED),

62L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

subthreshold AN)Note: subthreshold AN(N = 50; met allDSM criteria exceptamenorrhea orBMI N 17.5)not includedin review

Specic inclusion criteria: Female Aged 1645Specic exclusion criteria: BMI b 14 Severe physical conditions Comorbid psychiatric disorders Illiteracy, cognitive impairment Prior psychological treatment for ED Illness duration b 1 year

12 (22%) were sub-threshold, 22 (41%) remained AN Signicant increase in BMI from baseline to end oftreatment (maintained at follow-up) Signicant reduction in scores on depression measure Signicant reduction in ED-specic scores (maintainedat follow-up) Signicant reduction in scores on body uneasiness measureAt 3 year follow-up: 1 pp recovered, 3 changed to subthreshold AN Signicant increase in body uneasiness

14 Byrne,Fursland,Allen, andWatson(2011)

To examine the effectiveness ofCBT-E for ED's in the community(including low weight patients,excluded from previousCBT-E trial)

Naturalistic open effectivenesstrial Mixed AN, BN, EDNOS

Outpatient Individual Manualised

AN, BN, EDNOS subgroups Total N = 125 AN N = 34 (BN N = 40, EDNOS N = 51) Groups similar on baseline characteristics except AN lowermin/max adult weights and more likely prior inpatient admissionAN sample: Mean age 26.82 years 32 (94.1%) females 88.2% White Mean duration of ED 10.13 yearsSpecic inclusion criteria: 16 years+ Meet DSM-IV diagnostic criteriaSpecic exclusion criteria: Acutely suicidal/psychotic; current substance misuse BMI b 14

All groups indicated treatment was credible andexpected treatment to be useful Dropout for AN 50% (vs. 35% BN, 37.3% EDNOS)Intent to treat sample: AN sample achieving full remission = 6/34 (17.6%) Full or partial remission = 6/34 (17.6%) EDE-Q global score criteria: 13/34 (38.2%) EDE-Q plus BMI criteria: 3/34 (8.8%)Completers only: AN sample achieving full remission = 6/12 (50%) Full or partial remission = 6/12 (50%) EDE-Q global score criteria: 8/12 (66.7%) EDE-Q plus BMI criteria: 3/12 (25%) Both completers and total sample: signicantimprovements (medium-large effect sizes) over time onmost ED and general measures, however results bydiagnosis not presented

15 Fairburnet al.(2013)

1. What proportion of pps areable to complete treatment?2. What are the outcomes forcompleters?3. Are changes sustained?4. Which baseline variablespredict treatment completion?Plus consistency in responseto CBT-E across 2 sites.

Both within-groupscomparison (across threetimepoints) and betweengroup comparisons(UK & Italian sites)

Outpatient Individual Manualised

Total N = 99 (50 UK, 49 Italy) Mean age 24 years 97% female Duration of illness: 3 years (median) No signicant differences on key demographic variablesacross UK & Italian samples However signicantly lower mean weight in Italian study(no lower BMI limit)Specic inclusion criteria: Aged 1865 BMI 17.5 or belowSpecic exclusion criteria: BMI b 15 (UK only) Patient unsafe to manage on outpatient basis Comorbid psychiatric disorder precluding ED treatment(e.g. psychosis, drug dependence) Psychotherapeutic treatment for ED in previous year

Intent to Treat Analysis: Marked increase in weight from pre-post treatment,maintained at follow-up Marked decrease in ED & general psychopathologyCompleters: 63.6% total sample completed treatment (dropoutsor withdrawn on clinical grounds)no signicantdifference across sites Signicant weight gain amongst completers(62% reached BMI18.5) Signicant reduction in ED-specic scores and broaderpsychopathology scores Generally maintained at follow-up however % of ppswith BMI 18.5 dropped to 55% 12 pps required additional treatment/booster sessions Higher levels of psychopathology & severity of EDassociated with decreased likelihood of completion

16 Dalle Grave,Calugi, Doll,and Fairburn(2013)

Overall aim: is CBT-E a viablealternative to FBT in adolescentswith AN?Specic aims:1. What proportion of pps areable to complete treatment?2. What are the outcomes forcompleters?3. How well are changesmaintained?

Within-subjects (pre-,post- and follow-up)

Outpatient Individual(plus somefamilysessions) Manualised

Total N = 46 Mean age 15.5 years (range 1317 years) All single white females Duration of illness: 0.5 years (median); range 05 yearsSpecic inclusion criteria: Aged 1317 Parental consentSpecic exclusion criteria: Unsafe to manage on outpatient basis Comorbid psychiatric disorder precluding ED treatment Psychotherapeutic treatment for ED in previous year

Intent to Treat Analysis: Marked increase in weight from pre-post treatment,maintained at follow-up Marked decrease in ED & general psychopathologyCompleters: 63% total sample completed treatment withoutadditional input (19.6% non-responders withdrawn,17.4% ceased to attend) Signicant weight gain amongst completers (32.1%gained sufcient weight to reach 95% of expected weightfor age/sex)weight gains maintained at follow-up Signicant reduction in scores on ED measure frompre-post treatment (96.6% minimal residual EDpsychopathology post-treatment); partially maintainedat follow-up (89.7% minimal residual ED psychopathology) Signicant reduction in broader psychopathology frompre-post treatment, maintained at follow-up 7 pps required additional treatment/booster sessions

63L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

follow-up data.

64 L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472Attrition amongst outpatient/mixed inpatient and outpatient treat-ments varied (where reported), with dropout reported at 20% (study10), 18% (study 13), 50% (study 14). Approximately two-thirds ofparticipants in studies 8, 15 and 16 completed treatment (with a highproportion available for follow-up in study 15, and 100% in study 16).Analysis of differences between completers and non-completersshowed no signicant differences on demographic variables in eitherstudy, however in study 15 (in contrast to study 8) increased EDseverity and psychopathology at pre-treatment was associated with adecreased likelihood of completion. Study 16 did not analyse com-pleters vs. non-completers.

3.4.1.2. Effect of CBT on physical outcomes. Of those studies which report-ed BMI, all reported increases from pre- to post-treatment. Signicantincreases were reported in studies 10, 12, 13, 15 and 16; no statisticaltests were performed in studies 9 and 11. Study 14 reported anon-signicant increase in BMI for the overall group, but did notreport on AN individually. Study 12 further subdivided into ANsubtype, nding signicant increases for both AN-R and AN-BP(with no difference between subtypes). Signicant weight gainin study 15 was found for both UK and Italian sites, despite the lat-mixed inpatient and outpatient basis. Treatment in 5 of the studiesconsisted of elements of both individual and group/family work, whilein 3 studies treatment was on an individual only basis, and 1 studywas group-based treatment only. Ethnicity was reported in 4 studies(see Table 3 for further information).

3.4.1. Prominent ndings3.4.1.1. Adherence/Attrition. In studies 9 and 11 (same sample; inpatienttreatment) all 32 participants nished treatment and completed preand post measures, however at follow-up (study 11 only) return ratewas only 50%. No signicant differences were present on demo-graphic variables between participants who did/did not provide(individual therapy, physician/dietician advice, support groups) andthiswas an obvious confound. In addition the CBT group simultaneouslyreceived uoxetine (for relapse prevention) or placebo, which mayhave had a direct (in the case of the active drug condition) or indirect(via expectancy effects) impact on treatment effectiveness.

Study 6 used a categorical classication of outcome, based on theresults of a single measure. This constituted a rather narrow measureof recovery in AN and omitted aspects of broader functioning thatmay be important to consider when deciding a treatment's effective-ness. Study 7 only used measures at baseline which did not allowcomparisons on key variables to be made.

Study 6 sought to investigate whether the additional element ofbody therapy was effective. However, the comparison between CBTwith body therapy with BFT with no body therapy did not make it pos-sible to disentangle the impact of body therapy on outcome. Duration oftreatment for each group varied dramatically.

To summarise, the non-randomised clinical trials suggested that CBTled to greater improvements compared to MTAU in an outpatient set-ting once weight was restored. However for inpatients the evidencesuggested that CBT was effective but not signicantly more effectivethan alternative treatments.

3.4. Non-comparative clinical trials (Studies 816)

Study characteristics are summarised in Table 3. This review reportsonly on participants with AN within wider samples. Sample size (ANonly) ranged from 28 to 99. There was wide variability in the CBTprogrammes across these 9 studies. Treatment in 2 studies was entirelyon an inpatient basis, 5 entirely on an outpatient basis, and 2 others ater having a signicantly lower weight than the UK sample at pre-treatment. Where measured, weight gain was generally main-tained at follow-up.

3.4.1.3. Effect of CBT on ED symptomatology. Amongst the studies con-ducted on an outpatient basis results were mixed. Study 8 reportedsmall improvements on ED measures but these did not reach statisticalsignicance. Studies 13, 15 and 16 did nd signicant improvementspost-treatment, with some gains maintained at follow-up. Study 14also found signicant positive change in ED symptoms, however thiswas reported for the whole group (not specically AN). However anadditional measure (body uneasiness) used in study 13 which showedimprovement from pre-post treatment appeared to deteriorate atfollow-up.

Inpatients in study 9 demonstrated signicant improvements interms of anorexic cognitions at post-treatment. Study 11 reportedsignicant improvements on eating attitudes, plus on 8 of 11 subscalesof another ED measure at post-treatment, with some of these gainsmaintained at follow-up.

Studies 10 and 12 (combination of inpatient and day hospital treat-ment) also reported signicant improvement in ED symptomatology(episodes of ED behaviour and scale scores). Study 12 furthersubdivided into AN subtype, with both subtypes showing signicantimprovements.

3.4.1.4. Effect of CBT on general psychopathology & functioning. All 4 stud-ies measuring depression found signicant reductions in severity fol-lowing treatment. Study 11 suggested that such gains weremaintained at follow-up, however study 13 did not support this. Studies13, 15 and 16 reported signicant reductions in a broadmeasure of psy-chopathology/distress following treatment, maintained at follow-up.However on a specic anxiety measure, study 13 actually sug-gested increased anxiety levels from pre-post treatment andfollow-up. This is in contrast to study 14 which reported signi-cant reductions in anxiety, stress and interpersonal problems,and signicant improvements in self-esteem and quality of life, al-though these results were for the whole sample (AN results werenot reported separately).

Study 9 reported signicant reductions in negative automaticthoughts and maladaptive schema. Study 10 found pre-post treatmentdifferences on temperament variables of harm avoidance, persistence,self-directedness, self-transcendence independent of ED diagnosis.Study 11 also found signicant pre- to post-treatment differenceson the personality traits of Hypochondriasis, Depression, Hysteria,Psychasthenia and Social Isolation.

3.4.2. Methodological issuesGeneral limitations relating to the size and characteristics of the

sample made to other studies in this review apply and may affectgeneralisability. These individual trials all investigated the effectivenessof CBT alone, with no comparison treatment. While adding to the evi-dence base, such studies are less robust in singling out the effectivenessof specic treatments.

The short duration of therapy in study 8 (just 10 sessions, excludingfollow-up) was unlikely to have been long enough for signicant im-provement to take place. AN is a complex difculty and likely to requireextended intervention. Treatment length in study 10was variable, mak-ing it difcult to compare with treatments in other studies. Studies 11,13 and 15 allowed pragmatic follow-up periods, although with theexception of study 15 these were not controlled in terms of access tofurther/additional treatment.

Treatment programmes in studies 912 were a combination of ele-ments (individual and group work, plus some family work), making itimpossible to examine the impact of these individual components onoutcomes. Consistency in programme delivery/content within studiesmay also be an issue, even where the same manualised protocol was

used; for example, while study 16 followed a manualised approach

65L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472utilised in other studies, it also included additional family work. Study12 included a series of specic, optional modules in the treatment pro-grammewhich meant not all participants received the same treatment.Few studies explicitly reported supervision arrangements making itdifcult to ascertain therapist adherence to treatment protocols.Study 13 acknowledged the limited experience and training of staffand high staff turnover which may have inuenced treatmentquality.

Again, the choice of outcome measures varied making comparisonsacross studies difcult. Study 9 acknowledged that some measureshad not been validated in ED populations. Despite several treatmentprogrammes incorporating family work, no measure(s) of family func-tioning were included.

In summary, the non-comparative trials suggested that participantsreceiving CBT (both as inpatients and outpatients) made progress interms of weight gain/BMI (even those with very low initial weight).Results with respect to ED-specic symptomatology and broaderpsychopathology were mixed but improvements were noted in manystudies.

4. Discussion

After synthesizing and appraising the evidence, a number of keyndings emerge which are discussed in detail below. Overall, CBTappeared to be an accepted and effective treatment for AN based on avariety of outcomes, however CBT did not demonstrate such favourableresults as to appear to be the treatment of choice for AN. Generalmethodological limitations of the reviewed papers are discussed, andnally the implications and recommendations following the presentreview.

4.1. CBT and treatment acceptance/adherence

The ndings of the reviewed papers suggested that CBT may bemore acceptable than some therapies and no worse than others withrespect to acceptance and adherence. Of the seven studies where CBTwas compared to other treatments, four suggested that dropout in CBTwas lower than in comparison treatment(s). This included dietarycounselling (studies 2 and 5), inpatient psychiatric treatment andnon-specialist outpatient treatment (study 4), and maintenance treat-ment as usual (study 7). Where no improved completion for CBT wassuggested, comparison treatments were Interpersonal Therapy andnon-supportive clinical management (study 3) and Behavioural FamilyTherapy (studies 1 and 6).

4.2. CBT and eating disorder symptomatology

CBT led to improvements over time in eating disordered symptoms;whether assessed using standardised measures or weight/body massindex (BMI). This was the case in all but one of the studies and forboth inpatient and outpatient, and individual and group formats. Theone study (study 8) which failed to show improvement on eating disor-der symptomatology followingCBTwas the only studywhere treatmentwas delivered on an outpatient group basis (all the other studies dem-onstrated positive results for outpatient one to one and inpatientgroup treatment). However, as discussed, the intervention in Study 8was of short duration and unlikely to lead to signicant change insymptoms.

Most of the studies which compared CBT to other treatment(s) didnot demonstrate superiority over comparison treatment(s). In factimmediate post-treatment results from study 3 suggested a role fornon-specic supportive clinical management over CBT, and proposedthe most important factors (besides psychoeducation and normalisa-tion of eating) to be empathy, regard, the therapeutic alliance andincreasing autonomy. These factors have long been proposed to be the

core conditions for change in psychotherapy, and may explain thepositive results seen in both the CBT and alternative treatments. How-ever, results from longer-term follow-up (Carter et al., 2011) suggestthat it is the more specic therapeutic approaches (CBT, IPT) whichare associated with longer-term positive outcomes. Authors suggest astepped approach to treatment should be evaluated, whichmay involvenon-specic support followed by a change in approach after early im-provements have been made. Improvements maintained at follow-up(where measured) and specic studies focusing on weight-restoredpatients also suggested that CBT may be useful for preventing relapseinto AN.

4.3. CBT and broader symptomatology

In general, there was evidence for positive effects of CBT on depres-sive symptoms, self-esteem and negative thinking. Interpersonaldifculties and global ratings of mood also improved. However, the ev-idence for anxiety was less clear. Studies 1 and 12 found improvementfrom pre-post treatment but in study 13 state and trait anxiety werethe only variables which did not show improvement. Results using analternative anxiety measure (study 14) demonstrated signicant posi-tive effects.

These results were perhaps not surprising given the known ef-fectiveness of CBT for these conditions alone (NICE, 2009, 2011).However in the reviewed studies CBT was not superior to othertreatments (where compared) suggesting that, while potentiallyuseful, CBT was not unique in its capacity to improve broaderpsychopathology.

4.4. Overall quality of reviewed studies

It can be difcult to recruit enough participants with anorexianervosa (AN) to enable meaningful comparison and calculation ofpower or effect size. Furthermore, with the risk of medical complica-tions, it is often difcult to maintain sample size and, as studiesdiscussed in this review demonstrated, dropout in treatment remainshigh.

There were a disproportionate number of female participants inthe reviewed studies. Prevalence rates suggest that approximately1 in 250 females, and 1 in 2000 males will experience AN (NationalInstitute for Health and Clinical Excellence [NICE], 2004). Lifetimeprevalence of AN in men has been reported at 0.3% (versus 0.9% infemales) (Hudson, Hiripi, Pope & Kessler, 2007). While it may beargued that samples in the reviewed studies broadly reectthe prevalence of AN across genders, 2 studies explicitly excludedmales. The failure to include male participants presents a potentialselection bias, and prevents exploration of potential gender differ-ences in AN. While some have suggested similar prognoses andresponse to treatment for men and women with eating disorders(Woodside, 2002), others have suggested that treatments mayneed to take gender-specic factors into account (Andersen &Holman, 1997). Further research is needed to evaluate outcomesfor men with AN.

Some studies included adolescents and adults in the samesample however current UK recommendations are for familytherapies to be the treatment of choice for children with AN(NICE, 2004). Services are often structured to reect differencesin treatment approaches for adults and children. The one studyreviewed here which was a purely adolescent sample did ndthat manualised CBT-E was associated with similar positiveoutcomes to adult samples, however this programme also incor-porated parent sessions. Further research may seek to investigatewhether there are differences in response to CBT between adultsand children/adolescents.

Only 5 of the reviewed studies report on the ethnicity of their sam-ples, and of these, samples were predominantly white/Caucasian.

Whilst some research suggests that eating disturbances may be more

As discussed in the previous section, further investigation ofdemographic characteristics including age, ethnicity and gender may

66 L. Galsworthy-Francis, S. Allan / Clinical Psychology Review 34 (2014) 5472prevalent in white compared to non-white populations (e.g. Wildes,Emery, & Simons, 2001), the absence of such information raisesquestions regarding how representative these samples are.

Diagnostic inconsistencies across the reviewed studies makecomparisons difcult. The studies reviewed sampled individualswith varying duration and severity of AN, some including subthresh-old cases in their samples; some using modied DSM criteria; someincluding weight-restored patients with a higher BMI than would nor-mally meet AN criteria. Studies varied as to whether amenorrhea wasa required criterion for inclusion (though this forms part of DSM-IVcriteria for diagnosis) (American Psychiatric Association, 1994). How-ever, the forthcoming DSM-V excludes amenorrhea as a requirementof diagnosis, which may mean more individuals will meet a diagnosisof AN.

CBT programmes were not directly comparable across studies(see Table 4, Appendix A). Although all shared similar aims (address-ing eating disordered cognitions and behaviours alongside weightstabilisation), apparent inconsistencies in the effectiveness of CBTmay in fact be manifestations of differing programme aims, lengthand focus. For example, programmes varied in the emphasis onpsychoeducation, and in the order in which cognitive and behaviour-al strategies were introduced. Factors such as whether programmeswere delivered on an inpatient or outpatient basis, and in what for-mat (individual, group-based, or a combination of both) introducefurther inconsistency. Despite several studies utilising the samemanualised programme (CBT-E), authors noted some modicationsto this programme. Not all studies described therapist qualications/experience which may have also affected variability of interventions.Prior exposure to therapy was not controlled for in most studies,with many reporting previous hospital stays and previous failedtreatments.

Inconsistencies in results may also reect variation in the outcomesof interest (and the specic measures) selected by researchers. Thismade it difcult to synthesise results, and suggested a lack of consisten-cy inwhat researchers consider to be themost relevant outcome indica-tors in AN.

Due to the physical risks of AN, psychological therapy needs to occuralongside (and actively promote) weight restoration. This makes itdifcult to separate CBT-specic gains from those gains that nutritionalrehabilitation, and subsequent improvements on both physical andmental functioning, achieves (Goldbloom & Kennedy, 1995). In mostof the reviewed studies, weight restoration was a key aim of the treat-ment programmes, however others (Carter et al., 2009; Pike et al.,2003) required a stabilised weight prior to CBT treatment. This variabil-ity in physical/weight status at the point of beginning CBT treatmentmight impact on the apparent effectiveness of CBT treatment. This haspotential implications in terms of the most appropriate timing of CBTin a patients' recovery. More recently, CBT-E approaches have explicitlyincluded a focus onweight restoration alongside CBT in the initial phaseof treatment.

4.5. Clinical implications & suggestions for further research

Structured CBT approaches may be useful for increasing adherenceto treatment in AN. This is an important given that dropout from ANtreatment is typically 50% (Lowe et al., 2001). Further studies investigat-ing the specic factors which may be responsible for this improved ad-herence would be worthwhile.

Secondly, although the evidence reviewed does not suggest superi-ority over other treatments, CBT does show positive results in terms ofphysical, eating disorder-specic, and wider psychological outcomes,and also for relapse prevention. A more longitudinal approach withextended periods of follow-up would be useful to further evaluate themaintenance of treatment gains.

Differences between AN subtype were little explored in the re-

viewed studies. While post-hoc analysis in study 2 suggested nohighlight potential differences in response to treatment in AN, and it isimportant that future studies report this information to further explorewhat works for whom.

Measurement of outcome in the reviewed studies was entirelyquantitative. The inclusion of more qualitative measures (e.g. interper-sonal, social and systemic factors) in future may highlight interestingdifferences between patients and professionals in terms of what isimportant in recovery from AN.

A number of hypotheses have been proposed for the apparentmismatch between theory and effectiveness in practice of CBT forAN. Johnson, Tsoh, and Varnado (1996) discussed the limiting effectsof physical symptoms (and cognitive sequelae) associated with lowbody weight. Similarly, McIntosh et al. (2005) suggested that theamount of psychoeducational material and extensive skills acquisitionrequired in CBT may be difcult due to the cognitive rigidity of ANpatients. Such cognitive factors may account for the relatively slow im-provement of AN patients in therapy, and for the limited improvementsobserved in short-term interventions. An implication for treatmenttrials may be to extend the measurement and follow-up periods.Clinical implications include lengthening treatment to take thisslow progress into account. While Fairburn, Cooper, and Shafran(2003) suggested extending programme length for AN, treatmentsmight also need to aim the content and material more appropriatelydepending on patients' stage of treatment to ensure maximumeffectiveness.

Various formats of CBT were utilised in the reviewed studies.This afrms what Goldstein et al. (2011) refer to as a lack of com-mitment to a single treatment modality for AN (p.29). This makesit difcult to conclude whether individual CBT (alone or as part of awider treatment package) or group CBT is more effective and fur-ther research is needed on the most effective treatment formatfor AN. More recent transdiagnostic theories (Fairburn et al.,2003), and subsequently developed enhanced forms of CBT foreating disorders (CBT-E) have received increasing attention in theliterature, and benet from being manualised and replicable acrosssites/studies. Several of the reviewed studies utilised this approachwith generally positive outcomes, although it remains unclearwhether CBT-E is more effective than traditional CBT approaches inAN.

4.6. Conclusion

In line with previous reviews, this review has demonstrated thatthere is inconsistent information on the effectiveness of CBT for AN.Despite the lack of clear best practice standards for AN, profes-sionals in the eld believe that CBT remains a defensible candidatein treatment trials given the obvious risks of deferring treatmentuntil the evidence base is strong enough to draw more solidconclusions (Vitousek, 2002). The current review suggested largevariability in CBT format and content, and further research needsto be conducted in order to establish which variations of CBT aredifferences in outcome for AN subtype, study 13 found differences be-tween restricting and binge/purge subtypes (albeit in a combined clin-ical and sub-clinical AN sample) There was different treatmentresponses in these groups, including a higher rate of treatment resis-tance in restricting compared to AN binge/purge subtype. There maybe different underlying factors in the development and maintenanceof anorexic subtypes, which may have implications for their treatment.Indeed, research suggests those who have AN with bulimic features areless likely to engage in treatment and have poorer prognosis (Hsu, Crisp,& Harding, 1979; Steiner, Mazer, & Litt, 1990). Such differences warrantfurther investigation.most effective and useful in treating AN.

Appendix A

Table 4CBT programmes & outcomes of interest.

Study & location Format Number/lengthof sessions

Duration of therapy Aims/goals of therapy Outcome Measurement

Ball and Mitchell (2004)Sydney, Australia

Outpatient Individual Manualised Based on Garner & Bemis (1982) Modied to address core beliefsakin to Young's schema approach

25 1 h sessions 1 weekly for 3 months,1 fortnight for 3 months,1 month for nal 6 months

12 months in total Normalising eating behaviours Working with maladaptive core beliefs

Measured pre, post, follow-upPhysical measures: Primary outcome: weight and menstrual functioningclassied as good, intermediate, poor Weight gain BMIED-specic: EDEBD, IA EDI-2 ABOS MRSBroader psychopathology: BDI STAIOther: SSES IBCEC

Serfaty et al. (1999)Shefeld, U.K.

Outpatient Individual Manualised

20 1 h sessions 1 weekly

6 months Engagement/assessment Promoting understanding of model Collaborative case formulation Weight gain targets Dietary plans/binge reduction strategies Address cognitions & body image distortion Correct affect misidentication Work on self-esteem Schema-level work Techniques to reduce guilt/anxiety Relapse prevention

Measures taken at initial assessment and6 month follow-upPhysical measures: BMIED-specic: EDIBroader psychopathology: BDIOther: DAS LCB

McIntosh et al. (2005)Christchurch,New Zealand

Outpatient Individual Manualised

20 1 h sessions 1 weekly

Min. 20 weeks Phase one: Introduction to CBT, rationale, core techniques(self-monitoring, homework) Addressing motivation/ambivalence Normalisation of eating Weight range goals negotiatedPhase two: Specic skills (thoughts challenge/restructuring) PsychoeducationPhase three: Preparation for termination Relapse prevention

Measurement pre-treatment, at 10 sessions andpost-treatment Primary outcome measure: global AN rating (ordinalscale devised by authors based on extent to whichpp meets AN criteria)Physical measures: BMIED-specic: EDE EDI-2General psychopathology: GAF HDRS

Gowers et al. (2007)Multiple sites acrossNorth-West England,U.K.

Outpatient Individual Manualised Programme devised specicallyfor trial

12 sessions Length and frequency ofsessions not described

6 months in total Aimed to demonstrate association betweenweight gain and reduced psychopathology Motivate patient to take the next steps to recovery

Measures taken at baseline and follow-up (1 and2 years)Physical measures: BMIED-specic: EDI-2Broader psychopathology: HoNOSCA & HoNOSCA-SR MFQOther: MRAOS (adjusted for adolescents) FAD

(continued on next page)

67L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

Table 4 (continued)

Study & location Format Number/lengthof sessions

Duration of therapy Aims/goals of therapy Outcome Measurement

Pike et al. (2003)New York, U.S.A.

Outpatient Individual Manualised

50 sessions 1 weekly Length of sessions not described

One year Maintenance and consolidation of gains Continued improvement/recovery Focus on cognitive and behavioral featuresassociated with the maintenance of eating pathology Schema-based approach to address self-esteem,self-schema, interpersonal functioning Relapse prevention

EDE at pre-randomisation and end of therapy Modied SCID at start and end of therapy Following rst session: 4 self-report questions re:treatment credibility and expectancy Height and weight measured at pre-randomisation,andweight calculated weekly for BMI

Fernndez et al. (1995)Barcelona, Spain andBad Pyrmont,Germany

Inpatient basis Combination of individual andgroup formats Multimodal CBT with additionalbody therapy (psychomotortherapy and video confrontation)

Not described Varied; mean length20.5 months(SD 5.4 months)

Restoration of weight Reintroduction of normal eating Change eating-disordered thinking Improve body image/reduce body dissatisfaction

Measured at pre- and post-treatment, plus 1-yearfollow-upPhysical measures: BMIED-specic: EAT, EDIBroader psychopathology:BDIOther: Time to reach target weight Categorical outcome based on EAT score

Carter et al. (2009)Toronto, Canada

Outpatient Individual Manualised Based on that used in Pikeet al. (2003)

Up to 50 45 minute sessions Average number ofsessions = 38

One year Phase one: Strategies to address behavioral dysfunctionpertaining to eating and weightPhase two: Cognitive restructuring techniquesPhase three: Application of schema-based approach to addressa broad range of relevant issues (interpersonalproblems, developmental issues, self-esteem)

Assessments before and after initial weightrestoration,and at 3 month intervals during maintenance period Main outcome: time to relapse (DSM criteria: BMI orresumption of bingeing/purging)Secondary outcomes: EDE, EDI (at baseline) BMI (at baseline) BDI (at baseline) RSES (at baseline)

Leung et al. (1999)Birmingham, U.K.

Outpatient Group Manualised Based on existing CBT modelsfor AN (Freeman, 1995;Garner & Bemis, 1982)

10 weekly sessions plus4 follow-up sessions

4 months Detailed description of session by session themesprovidedOverall aims: Develop motivation for change Increase knowledge Teach behavioral techniques to overcome anorecticbehaviours Equip with cognitive skills to challenge maladaptivethoughts Provide opportunity for mutual support andunderstanding

YSQ pre-therapy (used as predictors of changes inanorectic cognitions/symptoms)Measures taken pre and post therapy: EAT-26 MAC

68L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

Bowers and Ansher(2000)Iowa, U.S.A.

Inpatient Involves group psychoeducation,individual cognitive therapy, groupcognitive therapy and cognitivefamily therapy Combines CBT and nutritionalrehabilitation

Not stated Not stated Increase understanding of AN Identifying, understanding, challenging and alteringautomatic thoughts/cognitive distortions Working with family communication and schemas Not directed at specic symptoms of AN

Psychometric measures administered within 3 days ofadmission and approx. 7 days prior to dischargePhysical measures: BMIED-specic: MACBroader psychopathology: Not measuredOther: ATQ, YSQ

Dalle Grave et al. (2007)& Brambilla et al.(2010)Garda, Italy

Inpatient plus residential dayhospital Mixed individual and group/familywork (under 18s) Transdiagnostic protocol cf. Fairburn& Harrison (2003) adapted forinpatienttreatment cf. Dalle Grave (2005),named CBT-MS Distinguishable from CBT-E bystepped-care approach, multidisci-plinaryteam (vs. single therapist), treatmentofmore severe cases not manageable byoutpatient treatment alone Multi-step programme dependentonrequired level of care, from outpatientCBT, intensive outpatient CBT, dayhospital CBT, inpatient CBT, post-inpatientoutpatient CBTsame theory/proce-duresat each level

Number of sessions variable Length of sessions not reported

20 weeks total 13 weeks inpatienttreatment plus7 weeks day hospitaltreatment

3 phases:1. (Weeks 14)Engaging, educating; initiation ofweight regain; formulation2. (Weeks 517)Content dictated by formulation,plus specic modules for self-esteem, perfectionism,mood intolerance, interpersonal difculties. AdditionalCB family therapy module for patients under 18 years3. (Weeks 1820)Focus on maintenance andorganising outpatient follow-up

Physical measures:BMIPrimary outcome measure: TCI on rst day of admission and last day of treatmentSecondary outcomesED-specic: EDE 12.0DBroader psychopathology: BDIAdditional measures in 2010 study: Blood plasma and platelet tests STAI BIS-11 RSES

Bowers and Ansher(2008)Iowa, U.S.A.

Inpatient Individual CT based on Beck, modi-edfor inpatients

Not reported Total duration ofinpatient treatment61 days

Aims: Weight restoration Focus on change in thoughts, feelings, behaviours(distortions, schemas, core beliefs)

Measures taken at 3 timepoints: Within 3 days of admission Approximately 7 days prior to discharge At one-year post-discharge.

(continued on next page)

69L.G

alsworthy-Francis,S.A

llan/ClinicalPsychology

Review34

(2014)5472

Table 4 (continued)

Study & location Format Number/lengthof sessions

Duration of therapy Aims/goals of therapy Outcome Measurement

Group didactic psychoeducation,andgroup CT based on Beck within aprocess-oriented framework Cognitive family work, focusing oncommunication and schemas

Facilitate emotional expression/communication Increase understanding of how interpersonalinteractions contribute to disorder

ED-specic: EAT-26, EDI-2Broader psychopathology: MMPI-2, BDI, HRSD BMI also measured at admission and discharge

Ricca et al. (2010)Florence, Italy

Outpatient Individual Manualised (Garner, Vitousek &Pike, 1997)

40 hour-long Minimum 40 weeks Focus on egosyntonic nature of ANPhase 1: Introduction of model, rationale, techniques,homework, motivation/ambivalence,normalized eating, negotiation of weight goalPhase 2: Development of skills to challenge/restructure thinkingPhase 3:Relapse prevention

Measures taken at the beginning and end of treatment,and three-year follow upPhysical measures: BMIED-specic: EDE-Q-12 BUTGeneral psychopathology: BDI STAI SCL-90-R

Byrne et al. (2011)Perth, Australia

Outpatient Individual Manualised CBT-E cf. Fairburnet al. (2003, 2008) Transdiagnostic, althoughlow-weight received longertreatment period formotivation/weight gain

Approx. 40 50-minutesessionsfor low weight; treatment guideallows exibility re: number ofsessions required for each stage

Variable Stage 1: Engaging and educating, creating individualformulation, beginning behavioral changeStage 2: Review of progress, identifying barriersto change (forming remainder of treatment)Stage 3: Modication of maintenance processesStage 4: Maintenance of gains and relapse prevention

Categorical measure of recovery based on BMI, EDbehaviours and DSM criteria) EDE-Q-12 (outcome positive if post-treatment globalscore b 1 SD above community norm) EDE-Q criteria above plus BMI 18.5Broader psychopathology: RSE EDI-Perfectionism DTS IIP-32 DASS QLESQ-SFOther: CEQ

Fairburn et al. (2013)Oxfordshire &Leicestershire, U.K. &Garda, Italy

Outpatient Individual Manualised CBT-E cf.Fairburn et al. (2003, 2008)

40 weekly sessions of50 minute duration

40 weeks therapy Closed follow-upof 60 weeks

Phase 1: Increasing motivation to changePhase 2: Weight regain alongside tackling EDpsychopatholog