behavioral therapy for eating disorders on quality of life

Received: 12 January 2017

|

Revised: 26 February 2017

|

Accepted: 16 March 2017

DOI: 10.1002/eat.22719

REVIEW

The effects of cognitive-behavioral therapy for eating disorders on quality of life: A meta-analysis Jake Linardon, BA (hons) School of Psychology, Australian Catholic University, 115 Victoria Parade/Locked Bag 4115, Melbourne, Victoria 3065, Australia Correspondence Jake Linardon, Faculty of Health Sciences, Australian Catholic University, 115 Victoria Parade/Locked Bag 4115, Melbourne, Victoria, 3065, Australia. Email: [email protected]

|

Leah Brennan, PhD Abstract Objective: Meta-analyses have documented the efficacy of cognitive-behavioral therapy (CBT) for reducing symptoms of eating disorders. However, it is not known whether CBT for eating disorders can also improve quality of life (QoL). This meta-analysis therefore examined the effects of CBT for eating disorders on subjective QoL and health-related quality of life (QoL). Method: Studies that assessed QoL before and after CBT for eating disorders were searched in the PsycInfo and Medline database. Thirty-four articles met inclusion criteria. Pooled within and between-groups Hedge’s g were calculated at post-treatment and follow-up for treatment changes on both subjective and HRQoL using a random effects model. Results: CBT led to significant and modest improvements in subjective QoL and HRQoL from pre to post-treatment and follow-up. CBT led to greater subjective QoL improvements than inactive (i.e., wait-list) and active (i.e., a combination of bona fide therapies, psychoeducation) comparisons. CBT also led to greater HRQoL improvements than inactive, but not active, comparisons. Prepost QoL improvements were larger in studies that delivered CBT individually and by a therapist or according to the cognitive maintenance model of eating disorders (CBT-BN or CBT-E); though this was not replicated at follow-up Conclusions: Findings provide preliminary evidence that CBT for eating disorders is associated with modest improvements in QOL, and that CBT may be associated with greater improvements in QOL relative to comparison conditions.

Resumen Objetivo: Los meta-analisis han documentado la eficacia de la terapia cognitivo-conductual (TCC) para reducir los síntomas de trastorno alimentario. Sin embargo, no se sabe si la TCC para trastorn puede mejorar la calidad de vida (CV). Por lo tanto, este meta-analisis nos alimentarios tambie los efectos de la TCC para trastornos alimentarios en CV subjetiva y CV relacionada con examino la salud (CVRS). todo: En las bases de datos PsycInfo y Medline se realizaron bu squedas de estudios que evalMe s de la TCC para trastornos de la conducta alimentaria. Un total de uaran la CV antes y despue n. Se agruparon intra y entre grutreinta y cuatro artículos cumplieron con los criterios de inclusio la g de Hedge en el post tratamiento y el seguimiento para evaluar los cambios pos, se calculo tanto en el tratamiento de CV subjetiva como en la CVRS utilizando un modelo de efecto aleatorio. Resultados: La TCC dio lugar a mejorías modestas y significativas en la CV subjetiva y en la CVRS mejoras subjetivas en la CV en desde el pre y post tratamiento y el seguimiento. La TCC genero n a las terapias inactivas (i.e., lista de espera) y activas (i.e., una combinacio n de terapias comparacio , psicoeducacio n). La TCC tambie n llevo a una mejoría mayor en la CVRS que las comde buena fe paraciones inactivas, pero no en las activas. Las mejorías en la CV pre y post fueron mayores en

Int J Eat Disord. 2017;1–16.

wileyonlinelibrary.com/journal/eat

C 2017 Wiley Periodicals, Inc. V

|

1

2

|

LINARDON

AND

BRENNAN

los estudios que ofrecían la TCC en forma individual y por un terapeuta o de acuerdo al modelo de mantenimiento cognitivo de los trastornos de la conducta alimentaria (TCC-BN o TCC-E); aunque esto no fue replicado en el seguimiento. Conclusiones: Los hallazgos proporcionan una evidencia preliminar que la TCC para trastornos alimentarios esta asociada con una mejoría modesta en la CV y que la TCC puede estar asociada con n con las condiciones de comparacio n. una mayor mejoría en la CV en relacio KEYWORDS

cognitive-behavioral therapy, eating disorders, quality of life

1 | INTRODUCTION

the confluence of comorbid psychopathology on QoL and are also more sensitive to change (Williams, Watts, & Wade, 2012).

Eating disorders are highly prevalent, chronic, and disabling conditions

Cognitive-behavioral therapy (CBT) is effective for reducing symp-

that negatively impacts an individual’s quality of life (QoL) (Fairburn &

toms of eating disorders. While several distinct cognitive-behavioral

Harrison, 2003). Individuals with eating disorders consistently report a

treatment protocols and formats (e.g., day-patient, inpatient, self-help)

poorer QoL than healthy controls (Jenkins, Hoste, Meyer, & Blissett,

for eating disorders exist, a specific form of therapist-led manualised

2011), and studies have reported greater QoL impairments in individu-

CBT (CBT-BN) is the leading evidence-based treatment for bulimia

als with eating disorders relative to other mental health conditions

nervosa (BN) and binge eating (Hay, Bacaltchuk, Stefano, & Kashyap,

(e.g., mood disorders) (De la Rie, Noordenbos, & Van Furth, 2005) and

2009). CBT-BN is based on a cognitive maintenance model of BN and

to several common medical disorders (e.g., angina and cystic fibrosis)

it has recently been revised so it is suitable for the treatment of all eat-

(Keilen, Treasure, Schmidt, & Treasure, 1994).

ing disorders (CBT-E) (Fairburn, Cooper, & Shafran, 2003). Given the

QoL is a multidimensional construct encompassing physical, psy-

superior effects of CBT-BN at reducing eating disorder symptoms rela-

chological and social dimensions of health (WHOQOL group, 1995).

tive to alternate psychological and pharmacological treatments, interna-

There are several approaches to measuring QoL. One approach is to

tional treatment guidelines recommend CBT-BN as the first-line of

assess QoL through objective indicators (e.g., income level, housing sta-

treatment for certain eating disorders (National Institue of Clinical

tus), generally by reference to the standing of an individual to the pop-

Excellence, 2004).

ulation (Holloway & Carson, 2002). Other approaches of assessing QoL

While eating disorder symptom reduction is critical for determining

are through self-report questionnaires. Many QoL measures assess an

CBTs success, individuals with eating disorders typically seek treatment

individual’s sense of wellbeing, satisfaction with life, and overall happi-

because of the debilitating effect their condition has on their QoL

ness (Engel, Adair, Hayas, & Abraham, 2009). This assessment of QoL

(Bohn et al., 2008). Thus, there has been a recent focus on assessing

is typically referred to as subjective QoL (Engel et al., 2009). Subjective

treatment outcomes in terms of both symptom reduction and improve-

QoL can be a global measure of wellbeing or satisfaction, or it can be

ments in QoL (Williams et al., 2012). Several studies have assessed the

broken down into distinct domains (e.g., social wellbeing). Subjective

impact of CBT for eating disorders on subjective and HRQoL, and a

QoL measures typically used within eating disorder populations include

generally consistent finding is that QoL improves immediately following

the Social Adjustment Scale, the Questionnaire on Life Satisfaction,

CBT (Ljotsson et al., 2007; Watson, Allen, Fursland, Byrne, & Nathan,

and the Quality of life Enjoyment and Satisfaction Questionnaire.

2012). However, QoL changes following CBT—and eating disorder

Unlike subjective QoL, health-related quality of life (HRQoL), which is

treatment in general—has received minimal research attention, and key

also assessed via self-report, assesses one’s life specifically in relation

questions remain unanswered. For instance, it is not known (1) what

to physical, sychological, and social health. HRQoL is composed of

the magnitude of QoL improvements immediately following CBT are;

both “generic” and “disease-specific” measures. Generic measures (e.g.,

(2) whether improvements in QoL following CBT are sustained over

Short-Form 36) can be applied to anyone and are generally used to

the long-term; (3) whether CBT is more effective at improving QoL in

make comparisons across conditions and populations (Engel et al.,

the short and long-term than both active and inactive comparisons.

2009). Generic measures can also be used as a global measure or it can

The current meta-analysis aims to address this gap by examining

also be broken down into specific domains (e.g., physical and mental

(a) if, and to what extent, CBT for eating disorders improves QoL in the

HRQoL). Conversely, disease-specific measures are designed to assess

short and long-term and (b) whether CBT is superior to alternative psy-

HRQoL in a specific populations, with the intention of assessing impair-

chological treatments at improving QoL in the short and long-term. We

ment peculiar to specific psychopathology (Engel et al., 2009). Both

also aim to test whether the effects of CBT on QoL are moderated by

generic and disease-specific measures are recommended for use in eat-

certain study characteristics, including (1) diagnosis, (2) study design, (3)

ing disorder research; generic measures allow for QoL comparisons

treatment modality (individual, group, self-help), (4) treatment format

across several populations (e.g., healthy controls, psychiatric popula-

(CBT-BN or CBT-E vs. other), (5) study quality, and (6) analysis reported

tions) while eating disorder-specific measures are designed to rule out

(completer, intention to treat).

LINARDON

AND

|

BRENNAN

2 | METHOD

3

sufficient data to calculate an effect size. A flowchart of the search strategy is presented in Figure 1.

This review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines (Moher, Liberati, Tetzlaff, & Altman, 2009).

2.1 | Search strategy

2.4 | Quality assessment The quality of included studies was assessed using the Quality Assessment Tool for Quantitative Studies developed by the Effective Public Health Practice Project (EPHPP) (Thomas, Ciliska, Dobbins, & Micucci,

The primary search strategy involved searching the PsycInfo and Med-

2004). This assessment tool was deemed suitable for systematic

line database in December 2016. The following two sets of terms were

reviews on intervention effectiveness, and can be used for RCTs and

searched simultaneously using the AND Boolean operator; (a) “eating

prepost designs (Thomas et al., 2004). Content and construct validity

disorder” OR bulimi* OR anorexi* OR binge* OR EDNOS; (b) CBT* or

has been established (Armijo-Olivo, Stiles, Hagen, Biondo, & Cum-

“cognitive-behav*” OR “cognitive behav*”. The secondary search strat-

mings, 2012). A rating of “strong,” “moderate” or “weak” methodologi-

egy involved searching the reference list of included papers and rele-

cal quality was assigned to each of the following six different domains:

vant reviews.

(1) selection bias; (2) study design; (3) confounders; (4) blinding; (5) data collection methods; (6) withdrawals and drop-outs. Then, a global qual-

2.2 | Inclusion and exclusion criteria

ity rating was made based on the ratings from the six domains. Studies

Included studies were those that (a) administered CBT that was specifi-

while those with one “weak” rating were rated as “moderate” quality,

cally designed to treat eating disorders, (b) in an adult eating disorder sample, (c) that included at least one measure of QoL at pretreatment and post-treatment or follow-up, (d) and was published in English and in a peer-review journal. Excluded studies were that that (a) administered a multidisciplinary treatment that included components of CBT in

that received no “weak” domain ratings were rated as “strong” quality, and those with two or more “weak” ratings were rated as “weak” quality. Any discrepancies were discussed among authors until a consensus was reached. Given the limited number of included studies, and consistent with previous meta-analyses, studies were not excluded based on their quality rating.

combination with other psychological or pharmacological treatments; (b) administered CBT weight loss interventions; (c) administered purely behavioral treatments or third-wave cognitive-behavioral treatments

2.5 | Data extraction

(e.g., Acceptance and Commitment Therapy, Dialectical Behaviour

The following information was extracted from the included studies:

Therapy), since these treatments distinguish themselves from CBT by

Diagnostic type; sample size; study design; CBT modality (individual

focusing on different aspects and perusing a different treatment goal.

therapist-led, group, self-help) and type (based on the cognitive mainte-

Although RCTs are the best method for testing a treatments efficacy,

nance model, CBT-BN/CBT-E, or not); comparison treatment (active or

due to the limited published studies that have assessed QoL change,

inactive); length of follow-up; QoL measure; analysis reported (com-

we included both prospective controlled and uncontrolled designs. This

pleter or ITT); any data that would permit calculation of an effect size.

allowed us to calculate both uncontrolled (i.e., prepost change) and controlled effect sizes (i.e., comparison between conditions). Because we are aware of the limitations of uncontrolled effect sizes (Cuijpers, Weitz, Cristea, & Twisk, 2016), we ensured that our controlled effect sizes were based on the available RCTs.

2.3 | Study selection

2.6 | Effect size calculation and data synthesis Primary analyses were conducted for post-treatment and follow-up subjective QoL and HRQoL changes. For studies that reported multiple follow-up points, effect sizes were calculated for the last reported follow-up. ITT data were prioritised over completer data, For within-subject designs (prepost change or pre to follow-up

The search strategy outputs from the databases were combined. Dupli-

change), the standardised mean gain was calculated by dividing the dif-

cates were removed. Titles and abstracts were screened to determine

ference between the post-treatment (or follow-up) and pretreatment

whether the study was related to the research question. To maximize

QoL mean by the pooled standard deviation (Lipsey & Wilson, 2001).

identification of relevant articles, the full-text of any study that admin-

Effect sizes were then converted to Hedge’s g to correct for biases due

istered CBT for eating disorders was read entirely to determine eligibil-

to small sample sizes (Hedges & Olkin, 1985). The standard error is

ity. This was because QoL is typically a secondary outcome reported

needed to correct for these biases. To obtain the standard error in

and is often not mentioned in the title or abstract when it is reported

repeated measures designs, the correlation between pretreatment and

in text. Articles that met inclusion criteria were screened to determine

post-treatment (or follow-up) QoL score is needed. However, r was not

eligibility for meta-analysis. Authors of articles that did not provide suf-

reported in included studies. Thus, we used the test retest reliability of

ficient data to calculate an effect size were contacted. Both authors

the relevant scale published in separate studies (Lipsey & Wilson,

discussed studies for which inclusion was uncertain. A total of 34

2001). Then, to calculate a pooled effect size, each studies overall

articles met full inclusion criteria. Thirty-three articles were included in

effect size was weighted by its inverse variance. A positive g indicates

the meta-analysis [one paper (Goldbloom et al., 1997) did not provide

QoL improvements from pre to post treatment/follow-up. A negative g

4

|

LINARDON

FIGURE 1

AND

BRENNAN

Flow chart of literature search

indicates decrements in QoL from pre to post-treatment/follow-up.

were calculated by weighing each effect size by its inverse variance.

Small (0.2), medium (0.5) and large (0.8) effects are specified.

Relative to comparison conditions, positive g indicates that CBT was

For between-subject designs (CBT versus comparisons), the stand-

associated with a greater QoL while a negative g indicates that CBT

ardised mean difference, d, was initially calculated by dividing the dif-

was associated with a lower QoL. One included study administered

ference between the post-treatment (or follow-up) CBT group mean

interpersonal psychotherapy (IPT) and behaviour therapy (BT) as com-

and the post-treatment (or follow-up) comparison group mean by the

parison treatments (Fairburn et al., 1991). For this study, we computed

pooled standard deviation (Lipsey & Wilson, 2001). D was also con-

an effect size comparing CBT to IPT, as across included studies IPT

verted to Hedge’s g to account for sample size and pooled effect sizes

was administered more often as a comparison treatment.

LINARDON

AND

|

BRENNAN

Diagnosis: A transdiagnostic, BN, anorexia nervosa (AN), or binge

At times, multiple effect sizes were calculated from the same study. This occurred when studies reported data for several subjective

eating disorder (BED) sample.

QoL or several HRQoL measures, or when studies compared multiple CBT conditions (e.g., self-help, therapist-led) to a control condition.

CBT modality: Individual therapist-led CBT, group CBT, or CBT self-help.

Including multiple effect sizes from the same study biases the overall effect size estimate (Borenstein, Hedges, Higgins, & Rothstein, 2009). To maintain statistical independence, we computed separate effect

5

CBT type: CBT that was based on the cognitive maintenance model of eating disorders (i.e., CBT-BN, Fairburn, Marcus, and Wilson, 1993 or CBT enhanced, Fairburn, 2008), or an alternative CBT approach.

sizes for each subjective or HRQoL measure or for CBT control comQuality rating: Strong, moderate, or weak quality rating.

parison, and then aggregated these estimates to produce an overall

Analysis reported: ITT or completer data reported.

effect size for that study. Although we intended on analysing the effects of CBT on specific subjective and HRQoL domains (e.g., physical and mental domains), this was not feasible because too few studies

Study design: Controlled study or a prepost design (only for within groups analyses).

used measures that assess these separate domains. For the few studies that assessed multiple domains, an aggregated effect size combining

Comparison type: Delivery of an inactive (e.g., wait-list) condition or active comparison treatment (only for between groups analyses).

these domains on either subjective or HRQoL was computed.

2.7 | Heterogeneity and moderator analyses

2.8 | Publication bias

To calculate the pooled effect sizes, the Comprehensive Meta-Analysis

The Fail-Safe N was calculated to address potential publication bias

program was used. For primary analyses, a random effects model was

(Rosenthal, 1991). The Fail-Safe N estimates how many missing studies

used over a fixed effects model. In the fixed effects model, it is

would need to be included in the meta-analysis for the effect size to

assumed that all studies in the meta-analysis are homogenous and are

become statistically nonsignificant. An effect is considered robust to

essentially replications of each other. In the random effects model,

publication bias if N is >5 K 1 10, where K is the number of studies

however, it is assumed that all included studies can be seen as a sample

included in the analysis (Rosenthal, 1991).

drawn from the population. Compared to the fixed effects model, the random effects model produces wider 95% confidence intervals, which

3 | RESULTS

means that it typically produces more conservative test statistics (Borenstein et al., 2009). Heterogeneity was assessed using the Q and I2 statistic. The Q statistic assesses the presence of heterogeneity, while the I2 statistic assesses the degree of heterogeneity, ranging from zero (complete homogeneity) to 100 (complete heterogeneity) (Higgins & Thompson, 2002). Since heterogeneity was expected, we examined whether the effect sizes were moderated by study characteristics. For each subgroup, a pooled effect size is calculated, and a test is conducted to examine whether subgroup effect sizes differ significantly from each other. A mixed effects model was used, which pools studies within subgroups using a random effects model, but tests for significant differences between subgroups using a fixed effects model (Borenstein et al., 2009). Mixed effects models are generally a conservative approach for testing moderation effects, and are widely used in meta-analytic research. However, it is important to note that the use of mixed effect models has been criticised for failing to detect true effects when the number studies contributing to an analysis is low (Cornell et al., 2014). Rather, some have suggested that estimations based on Bayesian procedures are more appropriate, as Bayesian calculations have been shown to be more stable and powerful in meta-analyses with a small number of studies. However, Bayesian procedures require knowledge of a prior distribution (the range of possible values) for s, which is often not known (Cornell et al., 2014; Liem et al., 2010). Statistically significant differences between subgroups are denoted by the Qbetween statistic. The following categorical moderators were examined for both the within and between-group analyses:

3.1 | Study characteristics Thirty-four papers met full inclusion criteria. Effect sizes could not be calculated for one paper (Goldbloom et al., 1997), so 33 papers were included in the meta-analysis. Tables 1 and 2 presents the characteristics of the included RCTs and non-RCTs, respectively. The majority of studies received a “moderate” quality rating (k 5 17) followed by “strong” (k 5 11) and then “weak” quality ratings (k 5 6). See Tables 1 and 2 for domain and global quality ratings for each study. Of the 33 studies, 23 were RCTs, one was an uncontrolled study, and 9 were single treatment prepost designs. Of the 24 studies that included a comparison condition, nine used a wait-list control while 12 administered an active comparison. The most common active comparison treatment was IPT (k 5 4); other active comparisons included Emotion and Social Mind Training, Behavioral Weight Loss, Specialist Supportive Clinical Management, Psychoeducation, Supportive Expressive Therapy, Multidisciplinary Specialist Treatment, Treatment as Usual and Short-term Focal Psychotherapy (all k’s 5 1).The most frequent mode of delivery was individual therapist-led (k 5 14) followed by guided self-help (k 5 11) and then group CBT (k 5 9).1 Majority of studies used a transdiagnostic sample (k 5 14), followed by BN (k 5 8) BED (k 5 9) and AN (k 5 2). The most commonly used subjective QoL measure was the Social Adjustment Scale (k5 10) followed by the Questionnaire on Life Satisfaction (QLS; k 5 3), Satisfaction with Life Scale (k 5 2), Work and social 1

One study administered both an individual therapist-led and a guided selfhelp CBT.

D1

M

W

M

M

M

W

M

M

M

M

M

W

M

M

W

M

W

M

Agras, Walsh, Fairburn, Wilson, and Kraemer (2000)

Banasiak, Paxton, and Hay (2005)

Carrard et al. (2011)

Chen et al. (2003)

Davis, McVey, Heinmaa, Rockert, and Kennedy (1999)

Durand and King (2003)

Fairburn et al. (2015)

Fairburn et al. (1991)

Fairburn, Kirk, O’connor, and Cooper (1986)

Fischer, Meyer, Dremmel, Schlup, and Munsch (2014)

Garner et al. (1993)

Ghaderi and Scott (2003)

Goldbloom et al. (1997)

Lavender et al. (2012)

Ljotsson et al. (2007)

Mitchell et al. (2008)

Munsch et al., (2007)

Peterson et al. (2009)

W

M

W

M

M

M

M

M

W

W

M

M

M

W

M

S

W

M

M

D4

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

S

D5

M

M

M

S

M

M

M

S

S

S

S

M

M

M

M

M

M

M

D6

M

W

M

M

S

M

M

M

M

M

S

S

W

S

S

W

M

M

G

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

RCT

Design

BED

BED

Mixed

Mixed

Mixed

BN

Mixed

BN

BED

BN

BN

Mixed

BN

BN

BN

BED

BN

BN

Sample

Individual therapist-led CBT-BN (60) Individual therapist-assisted CBT-BN (63) CBT GSH (67)

Group CBT (44)

Individual therapist-led CBT-BN (66) CBT-BN Telemedicine (66)

CBT internet GSH (35)

Group CBT (35)

Individual therapist-led CBT-BN (24)

CBT GSH (15) CBT PSH (16)

Individual therapist-led CBT (30)

Group CBT (20)



Individual therapist-led CBT-E (65)

CBT GSH (34)

Individual therapist-led CBT (39)

Group CBT-BN (30) Individual therapist-led CBT-BN (30)

CBT internet-based GSH (37)

CBT GSH (54)


CBT intervention (n)

QLS IWQOL

Wait-list (69)

SF-36

SLS

CIA

SAS

SAS

SAS

QLS

SAS

SAS

CIA

SAS

SAS

SAS

IWQOL

SAS

SAS

Measure

Group BWL (36)

-

Wait-list (34)

ESMT (35)

Fluoxetine (23) Fluoxetine 1 CBT-BN (29)

-

SET (30)

Wait-list (21)

Individual STFP (11)

Individual therapist-led IPT (25)


Specialist treatment (34)

Psychoeducation only (19)

-

Wait-list (37)

Wait-list (55)


Comparison (n)

AND

(Continues)

HRQoL (disease-speciﬁc)

Subjective

HRQoL (generic)

Subjective


Subjective

Subjective

Subjective

Subjective

Subjective

Subjective


Subjective

Subjective

Subjective


Subjective

Subjective

QoL type

LINARDON

S

S

W

S

S

W

S

S

M

W

M

S

S

M

S

W

S

W

D3

|

M

S

S

S

S

M

M

M

S

M

S

S

S

S

S

S

S

D2

Study Quality Rating

Characteristics of the RCTs that met full inclusion criteria

Study

T A B LE 1

6 BRENNAN

| Note. M 5 Moderate; S 5 Strong; W5 weak rating; D5 domain; D15 selection bias; D25 Study design; D35 Confounders; D45 Blinding; D55 Data collection method; D65 Withdrawal and dropouts; G5 Global quality rating; n 5 the number of participants allocated to treatment condition; BED 5 Binge eating disorder; AN 5 anorexia Nervosa; BN 5 Bulimia nervosa; IPT5 interpersonal psychotherapy; BWL5 behavioural weight loss; SSCM 5 Specialist supportive clinical management; SET5 Supportive expressive therapy; TAU5 treatment as usual; STFT5 short-term focal psychotherapy; PSH5 pure selfhelp; TA5 therapist-assisted; TL5 therapist-led; QLS5 questionnaire on life satisfaction; SLS5 Satisfaction with life scale; CIA5 clinical impairment assessment; SF-365 short-form 36; SWS5 satisfaction with life scale; IWQOL5 Impact of weight on quality of life; WSAS5 work and social adjustment scale.

Subjective SAS Group IPT (81) Group CBT-BN (81) BED RCT M S M S Wilﬂey et al. (2002)

W

S

S

HRQoL (generic) SF-36 SSCM (32) Individual therapist-led CBT-AN (31) AN RCT S S M S Touyz et al. (2013)

M

S

S

Subjective EQ-5D VAS Wait-list (106) CBT GSH (108) Mixed RCT M M M M ter Huurne et al. (2015)

W

S

S

Subjective WSAS TAU (64) CBT GSH (59) Mixed RCT S M M S Striegel-Moore et al. (2010)

M

S

S

SWLS Wait-list (18) Group CBT (18) BED RCT W M W W Schlup, Munsch, Meyer, Margraf, and Wilhelm (2009)

W

M

S

WHO-QOL Wait-list (38) CBT internet GSH (38) Mixed RCT S S M S S S anchez-Ortiz et al. (2011)

M

D4 D3

S

Subjective

BRENNAN

HRQoL (generic)

AND

D2 D1 Study

T A B LE 1

(Continued)

Study Quality Rating

D5

D6

G

Design

Sample

CBT intervention (n)

Comparison (n)

Measure

QoL type

LINARDON

7

Adjustment Scale (k 5 2), Quality of Life Enjoyment and Satisfaction Questionnaire – Short Form (QLESQ; k 5 2), EuroQol visual analogue scale (k 5 1), and the Quality of Life Index – Spanish Version (k 5 1). Only two generic HRQoL measures were reported across studies, with three studies reporting the Short-Form 36 and one study reporting the World Health Organisation QoL scale. Several studies assessed eating disorder-specific HRQoL measures, with the Clinical Impairment Assessment (k 5 6) being the most frequently reported measure, followed by the Impact of Weight on Quality of Life Scale (k 5 4) and the Quality of Life in Eating Disorders Scale (k 5 1). No study assessed objective indicators of QoL.

3.2 | Within-group effect size 3.2.1 | Prepost effect size For the within-group prepost analysis on subjective QoL (Ncomp 5 20, N 5 1,044 participants), the random effects model produced a statistically significant, medium effect size of g 5 0.50 (95% CI 5 0.38, 0.62), v5 .001, p