consiliere bulimia nervoasa

7/29/2019 Consiliere Bulimia Nervoasa

1/21

Journal of Counseling & Development

April 2013

Volume 91152

2013 by the American Counseling Association. All rights reserved.

Received 08/26/11Revised 10/28/11

Accepted 02/10/12DOI: 10.1002/j.1556-6676.2013.00083.x

Bulimia nervosa is an eating disorder associated with a drive

or thinness or requent episodes o binge eating and harm-

ul compensatory behaviors to avoid an increase in weight

(American Psychiatric Association, 2000). A binge eating

episode involves a lack o control over the intake o substantial

amounts o ood. Compensatory behaviors can include vomit-

ing, laxatives, diuretics, asting, and extreme exercise (Keel &

Haedt, 2008; Shapiro et al., 2007). Characteristics o bulimia

nervosa also include a severe concern with ones body shape

and weight (Steano, Bacaltchuk, Blay, & Hay, 2006) and

perceptual distortion and extreme body dissatisaction (Cash

& Deagle, 1997). Cash and Deagle (1997) ound that 73%o patients with bulimia nervosa perceived their body size to

be larger than it actually was, and their body dissatisaction

attitudes exceeded 87% o the control participants.

Hudson, Hiripi, Pope, and Kessler (2007) estimated that

bulimia nervosa aected 0.5% o males and 1.5% o emales

over their lietime, with prevalence rates increasing in recent

generations. The estimated mean age o onset was 19.7 years,

and the average duration o the disorder was 8.3 years (Hudson

et al., 2007). Hudson and colleagues reported that 94.5% o par-

ticipants with bulimia nervosa had at least one other diagnosed

mental disorder, and most participants had received mental health

assistance or other emotional issues. Seventy-eight percent o

participants reported signifcant role impairments in their home,work, personal, or social lie, and 43.9% reported a severe impair-

ment. However, only 48.3% o individuals with bulimia nervosa

sought treatment or their eating disorder.

Although the ull syndromal incidence o bulimia nervosa

has been stable over time, the report o bulimia nervosa is on the

Bradley T. Erford, Taryn Rchards, Elzabeth Peacock, Karen Voth, Heather McGar, and Brooke Mller, Education SpecialtiesDepartment, Loyola University Maryland; Kelly Dncan, Division o Counseling and Psychology in Education, University o SouthDakota; Catherne Y. Chang, Department o Counseling and Psychological Services, Georgia State University. Correspondenceconcerning this article should be addressed to Bradley T. Erord, School Counseling Program, Education Specialties Department,Loyola University Maryland, Timonium Graduate Center, 2034 Greenspring Drive, Timonium, MD 21093 (e-mail: [email protected]).

Counseling and Guided Self-HelpOutcomes for Clients WithBulimia Nervosa: A Meta-Analysis

of Clinical Trials From 1980 to 2010Bradley T. Erord, Taryn Richards, Elizabeth Peacock,

Karen Voith, Heather McGair, Brooke Muller, Kelly Duncan,

and Catherine Y. Chang

This meta-analysis included 111 clinical trials exploring the eectiveness o counseling/psychotherapy and guided

sel-help approaches in the treatment o bulimia nervosa. In general, single-group studies supported higher efcacy

o counseling/psychotherapy, whereas wait-list, treatment-as-usual, and placebo studies indicated both approaches

were equally eective at termination (posttest) and ollow-up in altering binging, purging, laxative use, and sel-reported

bulimia or body dissatisaction perceptions in nearly all comparisons.

Keywords:meta-analysis, bulimia nervosa, guided sel-help, counseling, psychotherapy

rise. With less than hal o those diagnosed seeking treatment,

it is important to determine the most eective and accessible

treatment modalities available to restore individuals to a healthy

level o unctioning. Common treatment approaches or bulimia

nervosa include counseling/psychotherapy, pharmacotherapy,

and guided sel-help. There is debate over which approach is

most eective. For example, many studies have reported the

efcacy o counseling and psychotherapy in the treatment o

bulimia nervosa (Fettes & Peters, 1992; Ghaderi & Anderson,

1999; Lewandowski, Gebing, Anthony, & OBrien, 1997;

Shapiro et al., 2007; Thompson-Brenner, Glass, & Westen,

2003; Whitbread & Mcgown, 1994; Whittal, Agras, & Gould,1999). Although medication alone produced an initial posi-

tive result, Nakash-Eisikovits, Dierberger, and Westen (2002)

concluded that the eects did not last, and better results were

obtained when medication was combined with psychotherapy.

Counseling/psychotherapy alone, as well as in combination

with pharmacotherapy, requires highly trained mental health

and/or medical proessionals working in outpatient or inpatient

acilities, and these treatments can be quite expensive.

At the same time, there have been studies on client-directed

(pure) sel-help and therapist-directed guided sel-help ap-

proaches, and clinical trials have yielded mixed results. Sha-

piro et al. (2007) reported that the guided sel-help approach

yielded smaller eects than psychotherapy, whereas Steanoet al. (2006) recommended the sel-help approach as an e-

ective and less expensive alternative to psychotherapy and

medication or the treatment o bulimia nervosa. Signifcant

advantages o the guided sel-help approach are lower cost,

convenience, and accessibility o services in rural locales.

Earn CE credit.Visit http://learning.counseling.orgto purchase and complete the test online.


2/21


April 2013

Volume 91 153

Counseling and Guided Sel-Help Outcomes or Clients With Bulimia Nervosa

Counseling/psychotherapy was the most common treat-

ment approach or clients with eating disorders, and eect

sizes o clinical trials were generally positive (i.e., d> 0).

Simultaneously, recent research also inconsistently supported

the use o guided sel-help, although ar ewer clinical trials

on this approach were reported. Proessional counselors who

treat clients with bulimia nervosa should be interested in theefcacy and staying power o these therapeutic approaches.

Our current meta-analytic study was undertaken to answer

three general questions: (a) Is counseling/psychotherapy an

eective treatment or clients with bulimia nervosa, and i

so, do the results last? (b) Is guided sel-help an eective

treatment or clients with bulimia nervosa, and i so, do the

results last? and (c) Is there a dierence between counseling/

psychotherapy and guided sel-help interventions in the treat-

ment o clients with bulimia nervosa?

The Effectiveness of

Counseling/PsychotherapyMany clinical trials support the eectiveness o counseling

and psychotherapy or the treatment o bulimia nervosa,

with a majority o studies using cognitive behavior therapy

(CBT) as a primary treatment. Many researchers stated that

CBT is the treatment o choice or bulimia nervosa (Ghaderi

& Anderson, 1999; Lewandowski et al., 1997; Shapiro et al.,

2007; Whitbread & Mcgown, 2008; Whittal et al., 1999). Le-

wandowski et al. (1997) attributed the popularity o CBT to

the availability o published, standardized treatment manuals

and CBTs ocus on clients cognitive distortions and negative

attitudes, which are common symptoms reported by clients

with bulimia nervosa.Many CBT studies ound moderate to large eect sizes

in the reduction o bulimic symptoms and body dissatisac-

tion attitudes (e.g., Agras et al., 1994; P. J. Cooper & Steere,

1995; Ghaderi, 2006a; Grifths, Hadzi-Pavlovic, & Channon-

Little, 1994; Nevonen & Broberg, 2006; Tasca et al., 2006;

Wiley & Agras, 1993). Several previous attempts have

been made to synthesize these fndings. Lewandowski et al.

(1997) conducted a fxed-eects model meta-analysis o 25

studies using behavioral outcome measures and 17 studies

using cognitive-attitudinal outcome measures. They reported

average correlations o .69 or behavioral outcomes and .67

or attitude-related outcomes. Lewandowski et al. concluded

that CBT eectively reduced behavioral symptoms and cog-nitive distortions, such as concern with body shape/weight

and depressive symptoms. These researchers ound an eect

size od= 0.27 or ollow-up results over a small number o

studies reporting results at varying lengths o time.

Using a ixed-eects model, Ghaderi and Anderson

(1999) perormed a meta-analysis on randomized controlled

trials (RCTs) to assess the eectiveness o CBT. Ghaderi

and Anderson reported large mean gain eect sizes (single-

group studies) or binge eating (d= 1.32) and purging (d

= 1.32) and small to medium mean dierence eect sizes

(comparison-group studies) or binge eating (d= 0.47) and

purging (d= 0.58). In general, CBT was superior to other

psychotherapy treatments and pharmacotherapy. However,

these researchers ound little evidence that these gains were

maintained at ollow-up.

Whittal et al. (1999) compared 39 studies o psychotherapyand pharmacology treatments or bulimia nervosa and con-

cluded that CBT was superior to pharmacotherapy. Average

eect sizes or bulimic behaviors, attitudes, and depression

were reportedly 1.22 to 1.35 or CBT and 0.39 to 0.73 or the

medication trials. The efcacy o the combination o psycho-

therapy and pharmacology was explored; however, the small

number o studies led to inconclusive results.

Whitbread and Mcgown (1994) conducted a fxed-eects

meta-analysis using a mean gain ormula (single group)

because many clinical trials lacked a control group. They

reported an average eect size od= 1.72 or CBT trials and

concluded that CBT was superior to behavior therapy (d=

1.05), short-term psychotherapy (d= 1.01), amily therapy

(d= 1.00), and pharmacotherapy (d= 0.98). Whitbread and

Mcgown attributed the success o CBT to participant train-

ing in assertiveness, communication, problem solving, and

social skills. CBT also addressed the cognitive distortions that

aected the body shape attitudes oten expressed by clients

with bulimia nervosa.

Although the majority o clinical trials supported the su-

periority o CBT over all other approaches to counseling and

psychotherapy, Thompson-Brenner et al. (2003) concluded

that behavior therapy was actually more eective than CBT.

Their meta-analytic results indicated that 44% o patients

receiving behavior therapy were ully recovered at the con-clusion o treatment compared with 39.6% o CBT patients.

Mean dierence eect sizes (comparison group) or binge

eating behaviors across three CBT studies averagedd= 0.52

compared with d= 0.83 across three behavior therapy studies.

Mean dierence eect sizes or purging behaviors were d=

0.79 across fve CBT studies compared with d= 0.90 across

fve behavior therapy treatments. Unortunately, no ollow-up

results were reported.

Another issue embedded in the question o treatment

efcacy was the eectiveness o group versus individual

approaches to counseling and psychotherapy. Fettes and

Peters (1992) concluded that group therapy was moderately

eective and was superior to individual therapy alone.Fettes and Peters reported that 25% o group therapy par-

ticipants were in remission rom symptoms at termination

and maintained abstinence at the 1-year ollow-up. Average

group therapy eect sizes were d= 0.89 or 3 to 6 months

o ollow-up andd= 1.17 or 9 to 12 months o ollow-up.

Conversely, using a random-eects model, Thompson-

Brenner et al. (2003) concluded that individual therapy

was more eective than group therapy because 45.6% o

individual therapy patients stopped displaying binging and


3/21


April 2013

Volume 91154

Erord et al.

purging symptoms at treatment conclusion compared with

only 26.7% o group therapy patients. Shapiro et al. (2007)

also concluded that individual psychotherapy resulted in a

greater percentage o participants without bulimic behav-

iors than was true or group therapy participants.

Support or CBT and behavior therapy and individual and

group approaches in the treatment o bulimia nervosa sug-gested the general efcacy o counseling and psychotherapy.

However, with a lack o inormation on the lasting eects,

it is not clear whether these treatments are eective or the

treatment o bulimia nervosa over the long term (Ghaderi &

Anderson, 1999; Lewandowski et al., 1997; Whitbread &

Mcgown, 1994). Prior to 2003, all meta-analyses o bulimia

treatment used the fxed-eects model. The most recent meta-

analyses on the eectiveness o counseling and psychotherapy

or the treatment o bulimia nervosa were published in 2003

(Thompson et al; diverse clinical trials) and 2007 (Shapiro et

al.; RCTs only), both using a random-eects model. In addi-

tion, numerous additional clinical trials have appeared in the

literature over the past decade, many with better ollow-up

procedures, which could shed light on the question o treat-

ment efcacy at both termination and ollow-up.

The Effectiveness of Guided Self-Help

Far ewer clinical trials have studied the efcacy o the guided

sel-help approach or the treatment o bulimia nervosa, and

those that have been conducted have ound mixed results.

No studies ound to date used standardized meta-analytic

procedures to produce eect sizes or sel-help procedures

on behavioral and attitudinal symptoms. Steano et al. (2006)

conducted a systematic review o sel-help RCTs and oundsignifcant reductions in binge eating requency at termina-

tion compared with the wait-list control. Unortunately, these

researchers could not locate adequate inormation about the

lasting eects o sel-help treatments because o the lack o

ollow-up studies. Steano et al. recommended the use o sel-

help or initial treatment o bulimia nervosa but emphasized

the need or additional RCTs to assess the efcacy o sel-help

and ollow-up eects.

Sysko and Walsh (2008) reviewed client-directed sel-help

trials and revealed generally positive results, with an abstinence

rate o 26.8% to 50% or bulimia symptoms. These researchers

ound that sel-help was superior to the wait-list control condi-

tion, with reduction in symptoms ranging rom 25% to 87% or

the sel-help participants compared with 6% to 19% or those

in the no-treatment condition. Sysko and Walsh concluded

that therapist-guided sel-help reduced binging and purging

requency when compared with pure client-initiated sel-help.

Still, sel-help was somewhat benefcial i no other treatment

option was available. Sysko and Walsh ound that reductions

in symptoms were maintained rom ollow-up until 3 to 18

months. However, the lack o sel-help RCTs limits confdence

in pure client-initiated sel-help efcacy.

The Differences Between Counseling/Psychotherapy and Guided Self-Help

Few studies have directly compared counseling/psycho-

therapy with sel-help trials. Steano et al. (2006) reported

no dierence between the remission rates o sel-help and

individual or group CBT, supporting the use o the sel-helpapproach as a more accessible and cost-eective preliminary

treatment. However, this conclusion should be viewed with

caution because it was based on ew trials; small sample

sizes; and diverse treatment lengths (e.g., several sessions to

multiple months o treatment), therapist qualifcations (e.g.,

proessional counselors, psychologists, graduate research

assistants), and outcome measures.

Similarly, Keel and Haedt (2008) reviewed studies that

compared psychotherapy with a CBT-based guided sel-help

program or adolescent patients with bulimia nervosa and

reported no dierences in bulimic behavior abstinence rates.

However, the guided sel-help condition had more patients

demonstrating no binge eating behaviors at termination.

Several additional guided sel-help trials have also been

published in the extant literature over the past decade, and our

current meta-analysis was aimed at answering the question o

the dierential treatment efcacy o traditional counseling/

psychotherapy and guided sel-help approaches. In the current

meta-analysis, we addressed the three main questions noted

earlier by searching the extant literature or published clini-

cal trials that used quasi-experimental or true experimental

designs o interventionsor bulimia nervosa

Method

For this meta-analysis, counselingorpsychotherapy was de-fned as any intervention or treatment perormed by a mental

health practitioner or practitioner-in-training meant to reduce

the symptomatic display o bulimia nervosa. Self-help or

guided self-help was defned as any intervention primarily

perormed by a client with or without guidance rom a mental

health practitioner or practitioner-in-training meant to reduce

the symptomatic display o bulimia nervosa.

Inclusion and Exclusion Criteria

We used nine criteria to acilitate study selection procedures

to obtain a robust set o moderate to high-quality clinical

trials on the treatment o bulimia nervosa:

1. Studies appeared in print between 1980 and 2010.

2. Studies were published in English with no limitation

on the nation or culture o origin.

3. A treatment or intervention was implemented to di-

rectly reduce the symptoms o participants diagnosed

with bulimia nervosa.

4. Treatment involved individual, group, or amily ap-

proaches to counseling or psychotherapy. Drug trials

were excluded.


4/21


April 2013

Volume 91 155


5. Symptoms o bulimia nervosa were assessed by at

least one standardized measurement procedure (e.g.,

sel-report rating scale, requency count).

6. Output data (means and standard deviations) were

available or computation o mean gain eect sizes

or mean dierence eect sizes.

7. Participants were adolescents (age 13 years and older)or adults.

8. Studies had a minimum sample size o six participants.

9. Studies included quasi-experimental or true ex-

perimental clinical trial designs using either a single

group or some control or comparison condition (i.e.,

wait-list, placebo, or treatment as usual [TAU]).

Nonexperimental or preexperimental designs were

excluded.

I multiple studies were published using results rom the

same sample, redundant studies were eliminated to preserve

the independence o results.

Search Strategies

Candidate studies were identifed through redundant com-

puterized searches, review o reerence lists rom previous

meta-analyses and clinical trials, and hand searches o the

journals most likely to publish clinical trials on the treat-

ment o bulimia nervosa. We conducted computerized

searches o PsycINFO, Academic Search Premier, and

MEDLINE rom 1980 to 2010 using key words related to

intervention (e.g., counseling, psychotherapy, self-help)

and condition (e.g., bulimia, binging, purging). Search

parameters were limited to English, age (adolescents 13

years and older and adults), peer review, and clinical tri-als.Next,reerence lists o previously published synthesis

articles and clinical trials were searched or additional

candidate art icles. Finally,journals with high requencies

o candidate studies were searched (i.e.,International Jour-

nal of Eating Disorders, Behavior Research and Therapy,

European Eating Disorders Review, Journal of Consulting

and Clinical Psychology, American Journal of Psychiatry,

andArchives of General Psychiatry).Dissertation abstracts

were not searched because we assumed that moderate- to

high-quality dissertation candidates would have been sub-

mitted to a peer-reviewed journal or publication.

The third and ourth authors provided independent

judgments while applying inclusion/exclusion criteria toinormation garnered rom the title, abstract, and ull text

(when available) o each candidate study.Disagreements

were resolved by consensus-building processes, and the frst

author adjudicated fnal selection decisions.

Coding Procedures

Coding o 25 participant (e.g., sample size, age, sex, ethnic-

ity), design (e.g., randomization, recruitment method, setting

o treatment, type o treatment, type o comparison group),

and method (e.g., blind assessment, treatment manual, in-

dividual or group method, number o sessions, duration o

sessions) characteristics was completed to acilitate later

moderator or mediator analysis should sets o eect sizes

lack homogeneity. Each article was independently coded by

two authors (rom among the second to sixth author, with di-

erent authors coding dierent articles), each o whom was agraduate counseling student who excelled in research and as-

sessment course work, completed a training session conducted

by the frst author, and underwent rigorous supervision during

the coding process. Full text versions o each selected article

were obtained, and the frst author reereed any discrepancies

among coder ratings. Peer review o selected clinical trials,

all o which used true or quasi-experimental designs, served

as a proxy or study quality.

Outcome Measures

Outcome measures were required to be direct assessments o

one o the fve dependent variables: binging, purging, laxative

usage, bulimia rating (specifc subscales rom sel-report in-

struments), and body dissatisaction. Within the 111 selected

articles, nearly all outcome measures used were standardized

sel-report measures. The Eating Disorders Inventory (Garner,

Olmstead, & Polivy, 1983) was used in 30% o the trials,

the Eating Attitudes Test (Garner & Garfnkel, 1979) was

used in 8% o the trials, the Body Shape Questionnaire (P. J.

Cooper, Taylor, Cooper, & Fairburn, 1987) was used in 7%

o the trials, and the Bulimic Investigatory Test, Edinburgh

(Henderson & Freeman, 1987) was used in 8% o the trials.

The Eating Disorder Examination (Z. Cooper & Fairburn,

1987) was a clinician-administered interview that was used

in 30% o the trials.

Statistical Methods

As per Erord, Savin-Murphy, and Butler (2010), eect sizes

(i.e., mean dierence or mean gain eect size) rom similar

study designs (i.e., all wait-list, all placebo, all TAU, or all

single-group designs separately) were combined. All eect

sizes were independent.We analyzed posttreatment eects by

combining eect sizes generated immediately ater the buli-

mia treatment. Follow-up eects were analyzed by advancing

the last (i.e., most conservative) ollow-up eect size. Cohens

dwith pooled variance was used to compute standardized

mean dierence eect sizes or comparison-group studies;

positive eect sizes indicated a positive eect o treatment.Computation o standardized mean gain eect sizes or

single-group samples (dsg

) ollowed a ormula suggested

by Lipsey and Wilson (2001) and used a deault reliability

estimate o .80 in cases where sample reliabilities were not

reported. All eect size estimates (d) were then corrected or

sample bias (Erord et al., 2010), and then these unbiased

estimates (d) were again corrected using an inverse weight-

ing procedure (Erord et al., 2010; Lipsey & Wilson, 2001),

producing the corrected eect size (d+). Finally, eect size


5/21


April 2013

Volume 91156

Erord et al.

estimates rom similar study designs and dependent variables

were combined and averaged or hypothesis and homogeneity

testing (Cochrans Q andI2; see Erord et al., 2010) using a

random-eects model (Hedges & Olkin, 1985). Within homo-

geneity studies, ip < .05 or the Q statistic, the null hypothesis

o homogeneity could be rejected and potential mediation or

moderation explored.Likewise, Higgins, Thompson, Deeks,and Altman (2003) recommendedI2 interpretations o 0%

indicating total homogeneity, 25% low, 50% moderate, 75%

high, and 100% total heterogeneity. II2 > 50%, exploration

o mediator or moderator variables may be warranted.

Finally, hypothesis testing o d+ > 0 was acilitated by

the computation o 95% confdence intervals (CIs; Erord et

al., 2010; Lipsey & Wilson, 2001). Thus, i the entire 95%

CI range was greater than zero, the null hypothesis could

be rejected. Power is a major consideration in any empirical

study using samples o data. Ordinarily, meta-analytic sum-

mary statistics based on k> 20 studies have sufcient power

(Cornwell, 1993; Cornwell & Ladd, 1993), reducing the

probability o Type II errors. In this meta-analysis o bulimia

treatment outcome research, nearly all comparison-groupd+

analyses were underpowered (k< 20). In contrast, most o

the single-group, posttest, and ollow-up d+ analyses hadk

> 20, thus demonstrating sufcient power.

Publication Bias

Unpublished manuscripts were not included in the current

meta-analysis, which may have resulted in some publication

bias. Funnel plot analysis and Rosenthals (1979) ail-saeN

procedure were conducted on each set o eect sizes to assess

or publication bias. The ail-saeNprocedure provides inor-

mation about the stability o a meta-analysis by calculatingthe number o studies needed to bring a signifcantp level to

a nonsignifcant level o .01. Few outliers were noted, because

the eect sizes basically conormed to expected graphical

confgurations, and these ew outliers tended to reect both

high and low estimates in equal proportion. As a result,

these ew outliers were retained in the analyses rather than

removed or trimmed. Fail-saeNestimates or each analysis

are included in the tables.

Results

The decision-making ow process or article selection is

outlined in Figure 1. Electronic search procedures identifed1,441 candidate articles, whereas hand searching o reer-

ence lists and target journal tables o contents identifed 65

more articles, or a total o 1,506 candidate articles. O these

candidate articles, 1,346 were excluded or violation o at

least one o the inclusion criteria on cursory inspection, and

an additional 49 were excluded ater closer scrutiny through

a ull-text review (e.g., no direct outcome measurement,

no appropriate data to compute d, duplicate study).Judges

agreed on 97.0% (= .93) o independent selection decisions

and reached consensus on the rest. Landis and Koch (1977)

provided the ollowing interpretations or kappas: .41 to .60

were moderate and sufcient or research purposes, .61 to .80were substantial, and .81 to 1.00 were almost perect. In the

reerence list, a single asterisk preceding an article indicates

the article was advanced into the meta-analysis.

Study Characteristics

O the 111 articles advanced to the coding process, 82 were

single-group pretestposttest designs, and 29 used random-

ized samples with a comparison-group design (15 wait-list,

8 placebo, and 6 TAU).The total number o participants was

FiGuRE 1

Flow Chart of inclded Stdes

Note. k= number o studies.

Potentially relevantarticles identifed throughcomputerized search oPsycINFO and MEDLINE19902008 (k= 1,441)

Potentially relevant additionalarticles identifed throughsearch o article reerencelists and hand search oprominent journals (k= 65)

Total number o relevantarticles identifed andscreened or inclusion(k= 1,506)

Articles excluded ater titleand abstract review orailure to meet all inclusioncriteria (k= 1,346)

Articles potentiallyappropriate to be includedin the review and procuredin ull text (k= 160)

Excluded articles (k= 49),

including:

No direct measure ofbulimia outcomes (k= 12)

Treatment not counseling/psychotherapy (k= 3)

Appropriate effect sizedata not available (k= 28)

Duplicate study/sample(k= 6)

Articles fnally included in the meta-analysis with usableinormation (k= 111; n= 4,926; 142 posttest comparisons;75 ollow-up comparisons), including:

Single-study groups (k= 82; n= 3,272; 102 posttest

comparisons; 65 ollow-up comparisons) Wait-list control groups (k= 15; n= 852; 23 posttestcomparisons; 2 ollow-up comparisons)

Placebo study groups (k= 8; n= 394; 11 posttestcomparisons; 7 ollow-up comparisons)

Treatment-as-usual comparison groups (k= 6; n= 408;6 posttest comparisons; 1 ollow-up comparison)


6/21


April 2013

Volume 91 157


4,926. Across all 25 coded characteristics, intercoder agree-

ment ranged rom 73% to 100%, with a median percentage

agreement o 97%. Related kappas ranged rom .47 to 1.00,

with a median kappa o .93. Erord et al. (2011) pointed out

that eect sizes vary depending on the comparison condi-

tion. For example, single-group mean gain eect sizes are

usually higher, on average, than comparison-group eectsizes. Among comparison-group eect sizes, wait-list results

are usually more positive, on average, than TAU or placebo

comparisons because the latter two control methods provide

an active comparison as opposed to a no-treatment, wait-list

comparison. Thus, dierent study conditions may yield vary-

ing results on the same outcome variable (Thompson, 2002,

2006). Caution is thereore warranted in the interpretation o

eect size magnitudes, because each must be interpreted as

embedded in the appropriate context o condition, outcome

variable, number o studies, and sample sizes. Finally, or

interpretation o eect sizes, Cohen (1988) suggested d=

0.20 (small), d= 0.50 (medium), andd= 0.80 (large). Be-

cause d+ can be converted to a percentile rank using the z

transormation, an eect size o 1.00 means that the average

treatment group participant scored at the 84th percentile o the

comparison-group distribution (or mean dierences), or or

single-group studies (mean gain), the average posttest score

alls at the 84th percentile o the pretest score distribution.

Is Counseling/Psychotherapy Eective or the

Treatment o Clients With Bulimia Nervosa?

And I So, Do the Results Last?

The results o the current meta-analysis o the eectiveness

o counseling/psychotherapy on the treatment o bulimia

nervosa were analyzed by condition (single group, wait-list,placebo, and TAU) across fve relevant bulimia outcome vari-

ables (requency o binging, requency o purging [vomiting],

requency o use o laxatives, sel-reported bulimia charac-

teristics, and sel-reported body dissatisaction). The second

part o the question involves the assessment o lasting eects

o bulimia treatments. Unortunately, only slightly more than

hal o all identifed clinical trials conducted ollow-up stud-

ies to determine the staying power o treatments or bulimia

nervosa ater the conclusion o treatment. Furthermore,

these ollow-up studies varied markedly in the time ater

termination that the ollow-up measurement was assessed.

When ollow-up eect sizes were reported or studies with

multiple ollow-up measurements, the eects were reportedat the point most distant rom termination, thus giving the

most conservative estimate o the lasting eects. We present

the results by outcome variable across comparison conditions,

where kindicates number o studies andi indicates the number

o eect sizes derived.

Frequency of binging for counseling/psychotherapy.Table

1 presents summary results at the end o treatment (posttest)

or binging behaviors across the comparison conditions.

Eect size averages across all conditions were positive,

meaning that the treatments had positive average treatment

results, except in the placebo condition. A portion o the TAU

condition 95% CI range (d+ = 0.28; 95% CI [0.18, 0.74])

was not above zero, thus the null hypothesis o no dierence

could not be rejected. This was probably due to the small

number o studies and sample size in the TAU condition (i =

3, n = 168), because the placebo condition had virtually thesame d+ (0.26), but with i = 8 andn = 282, the 95% CI [0.01,

0.51] had sufcient control o standard error to reject the null.

Still, both o these d+ analyses displayed small magnitudes.

The single-group d+ was 0.71 (medium to large eect size;

i = 70, n = 2,322), and, surprisingly, the wait-list condition

was still larger, at d+ = 0.99 (large eect; i = 14, n = 475),

indicating robust, eective treatment results. All analyses dis-

played homogeneous eect size groupings, so no moderator

or mediator analyses were conducted. Also, the ail-sae Ns

were quite robust. For example, the TAU condition with only

three studies andd+ o 0.28 still would require the location

o an additional 83 unpublished, unlocated TAU studies with

eect sizes o zero to reduce the observed eect size to a

nonsignifcant d+ o 0.01.

Regarding the staying power o the treatment or reduction

o binging behaviors (see Table 1), both single-group (d+ =

0.75; medium to large eect; i = 49, n = 1,193) and placebo

(d+ = 0.77; medium to large eect; i = 6, n = 223) conditions

yielded average eect sizes greater than zero. Wait-list (d+

= 0.56; medium eect; i = 2, n = 117) and TAU (d+ = 0.11;

small eect; i = 1, n = 71) conditions did not, but again, these

analyses involved only two and one located studies, respec-

tively. It is important to note that a ew additional ollow-up

studies with similar results would have provided enough

power or the wait-list condition (d+ = 0.56) to reject the nullhypothesis o no dierence rom zero. All group eect size

estimates were homogeneous.

Frequency of purging for counseling/psychotherapy. Ac-

cording to Table 2, the d+ was 0.63 (medium eect; i = 63, n

= 1,961) or the single-group condition, 0.98 (large eect; i =

16, n = 491) or the wait-list condition, 0.57 (medium eect;

i = 3, n = 168) or TAU, and 0.36 (small to medium eect;

i = 6, n = 235) or placebo studies. All o these conditions

displayed average eect sizes greater than zero, and all eect

size groupings were homogeneous.

Follow-up or single-group studies resulted in a signifcant

positive result (d+ = 0.71; medium to large eect; i = 36, n =

885), but the wait-list (d+ = 0.66; medium eect; i = 2, n =117), TAU (d+ = 0.18; small eect; i = 1, n = 71), and placebo

(d+ = 0.31; small eect; i = 4, n = 135) conditions were not

greater than zero, again probably because o the small number

o studies reporting ollow-up results. All eect size group-

ings were homogeneous.

Frequency of laxative use for counseling/psychotherapy.

Few articles reported on requency o laxative use, but o

those that did, all showed a signifcant eect o treatment (see

Table 3). The d+ was 0.45 (small eect; i = 17, n = 654) or


7/21

158

TABLE1

SmmaryStatstcsfo

rtheFreqencyofBngngOtcomeVarable

ComparsonGrop

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Note.Bingingisthedependent

variable.

k=numberofstudies;i=

num

berofeffectsizesderived;d+=meaneffectsizeestimate;95%

CI=95%

confidenceinterval;>0

=thed+wasgreaterthan0;SigDiff=significantdifferencebetween

thecounseling/psychotherapyandguidedself-helpconditions(Yes=significantdifference;No=no

significantdifferences);Q(df)

=thehomogeneityindexforthegivend

egreesoffreedom;I2=ahomogeneity

indexpresentedasapercentage;YesintheHomogeneous

columnmeansthed+compris

esahomogeneousgroupingofeffectsizes.Posttest=themeasurementwastak

enattheterminationoftreatment;TAU=

treatment-as-usual

condition;Follow-up=themeasurementwastakenatthelongestposttreatmentfollow-upavailable.

Tm

e

Treatment

k

i

n

d+

95%C

i

>0

SgDffFal-SafeN

Q(df)

I2

Homogeneos

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

54

15

9

4

3

0

5

1

34

9

2

0

1

0

3

1

70

17

14

6

3

8

1

49

11

2

1

6

1

2,322

472

475

263

168

282

39

1,193

337

117

71

223

39

0.71

0.62

0.99

0.70

0.28

0.26

0.50

0.75

0.67

0.56

0.11

0.77

0.13

[0.63,0.79]

[0.38,0.85]

[0.76,1.20]

[0.47,0.98]

[0.18,0.74]

[0.01,0.51]

[0.15,1.15]

[0.66,0.85]

[0.34,1.00]

[0.40,1.52]

[0.37,0.59]

[0.03,1.50]

[0.51,0.73]

YesYesYesYesNo

YesNoYesYesNo

No

YesNo

Yes

Yes

No N

oNo

4,956

1,046

1,385

421

83

206

50

3,690

737

111

11

461

13

76.99(69)

23.71(16)

11.73(13)

3.52(5)

2.02(2)

5.96(7)

46.97(48)

15.72(10)

2.35(1)

4.65(5)

10.4

32.5

0.0

0.0

1.0

0.0

0.0

36.4

0.0

0.0

YesYesYesYesYes

Yes

YesYesYes

Yes


8/21

159

TABLE2

SmmaryStatstcsforthe

FreqencyofPrgng(Vomtng)OtcomeVarable

ComparsonGrop

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Note.Purgingisthedependentvariable.


num

berofeffectsizesderived;d+=meaneffectsizeestimate;95%

CI=95%

confidenceinterval;>0

=thed+wasgreaterthan0;SigDiff=significantdifferencebetween

thecounseling/psychotherapyandguidedself-helpconditions(Yes=significantdifference;No=no

significantdifferences);Q(df)

=thehomogeneityindexforthegivend

egreesoffreedom;I2=ahomogeneity

indexpresentedasapercentage;YesintheHomogeneous




treatment-as-usual


Tm

e

Treatment

k

i

n

d+

95%C

i

>0

SgDffFal-SafeN

Q(df)

I2

Homogeneos

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

41

15

9

3

3

0

4

1

24

9

2

0

1

0

2

1

63

17

16

3

3

6

1

36

11

2

1

4

1

1,961

482

491

151

168

235

39

885

337

117

71

135

39

0.63

0.50

0.98

1.37

0.57

0.36

0.48

0.71

0.65

0.66

0.18

0.31

0.17

[0.55,0.71]

[0.28,0.71]

[0.59,1.38]

[0.36,2.37]

[0.26,0.89]

[0.09,0.63]

[0.17,1.13]

[0.61,0.80]

[0.35,0.96]

[0.00,1.31]

[0.30,0.66]

[0.05,0.68]

[0.49,0.83]

YesYesYesYesYes

YesNoYesYesNo

No

NoNo

Yes

No

No N

oNo

3,938

845

1,574

410

172

214

48

2,538

718

131

18

124

17

64.22(62)

19.75(16)

14.37(15)

1.61(2)

0.65(2)

2.72(5)

37.78(35)

13.97(10)

1.00(1)

2.38(3)

3.5

19.0

0.0

0.0

0.0

0.0

7.4

28.4

0.0

0.0

YesYesYesYesYes

Yes

YesYesYes

Yes


9/21

160

TABLE3

SmmaryStatstcsforthe

FreqencyofuseofLaxatves

OtcomeVarable

ComparsonGrop

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Note.Useoflaxativesisthedependentvariable.


numberofeffectsizesderived;d+=meaneffectsizeestimate;95%

CI=95%

confidenceinterval;

>0=thed+wasgreaterthan

0;SigDiff=significantdifferencebetweenthecounseling/psychotherapyandg

uidedself-helpconditions(Yes=significantdifference;No=

nosignificantdifferences);Q(df)=thehomogeneityindexforthegiven

degreesoffreedom;I2=ahomogeneityindexpresentedasapercentage;YesintheHomogeneous


esahomogeneousgroupingofeffectsizes.Posttest=themeasurementwastakenattheterminationoftreatment;TAU=treatment-as-usual


Tm

e

Treatment

k

i

n

d+

95%C

i

>0

SgDffFal-SafeN

Q(df)

I2

Homogeneos

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

9

7

1

0

0

0

1

0

4

4

0

0

0

0

0

0

17

9

3

1

5

6

654

217

112

50

154

163

0.45

0.26

0.68

0.58

0.24

0.54

[0.32,0.59]

[0.14,0.37]

[0.18,1.18]

[0.00,1.16]

[0.01,0.49]

[0.08,1.00]

YesYesYes

Yes

NoYes

Yes

Yes

770

230

205

58

121

324

14.43(16)

3.47(8)

0.82(2)

4.12(4)

4.57(5)

0.0

0.0

0.0

2.9

0.0

YesYesYes

YesYes


10/21


April 2013

Volume 91 161


the single-group condition, 0.68 (medium to large eect; i =

3, n = 112) or the wait-list condition, and 0.58 (medium to

large eect; i = 1, n = 50) or the placebo condition. No TAU

studies assessing or laxative use were identifed. Again, no

heterogeneity was detected.

Follow-up or treatment o bulimia with laxative use as

the outcome measure did not result in a signifcant positiveeect or single-group studies (d+ = 0.24; small eect; i =

5, n = 154). This eect size grouping was homogeneous.

Unortunately, no ollow-up studies were located and selected

on laxative use or the wait-list, TAU, or placebo conditions.

Self-report bulimia ratings for counseling/psychotherapy.

Table 4 provides summary statistics or sel-report bulimia

ratings. Again, all conditions were signifcantly higher than

zero, even though several o the analyses had ew studies and

small sample sizes. The d+ was 0.81 (large eect; i = 44, n =

1,151) or the single-group condition, 0.99 (large eect; i =

5, n = 143) or the wait-list condition, 0.93 (large eect; i =

3, n = 112) or the TAU condition, and 0.62 (medium to large

eect; i = 4, n = 158) or the placebo condition. All conditions

yielded homogeneous eect size estimates.

Follow-up or bulimia sel-ratings on single-group studies

resulted in a signifcant positive result (d+ = 0.88; large e-

ect; i = 29, n = 709), but not or the placebo condition (d+ =

0.38; small eect; i = 4, n = 158). Both eect size groupings

were homogeneous. No TAU or wait-list ollow-up studies

were located.

Self-reported body dissatisfaction for counseling/psychotherapy.

Table 5 provides summary statistics or client sel-report o

body dissatisaction. No placebo studies were located, but

all other conditions resulted in d+ greater than zero, and all

conditions displayed homogeneous eect size groupings.The d+ was 0.50 (medium eect; i = 41, n = 1,424) or the

single-group condition, 0.66 (medium eect; i = 5, n = 143)

or the wait-list condition, and 0.60 (medium eect; i = 4, n

= 218) or the TAU condition.

Follow-up or body dissatisaction sel-ratings was sig-

nifcantly greater than zero or both the single-group (d+ =

0.56; medium eect; i = 24, n = 768) and TAU (d+ = 0.58;

small eect; i = 1, n = 71) conditions, with both eect size

groupings displaying homogeneity. No wait-list or placebo

ollow-up trials were located.

Is Guided Sel-Help Eective or the Treatment

o Clients With Bulimia Nervosa?And I So, Do the Results Last?

Far ewer articles were located evaluating the eectiveness o

sel-help interventions or the treatment o clients with buli-

mia nervosa compared with counseling and psychotherapy,

and some o these trials described guided help procedures that

involved minimal interventions and educational support by

mental health or medical proessionals. These were combined

or convenience in the ollowing analyses and reerred to as

guided sel-help. The same procedures and variables were

examined to determine the eectiveness o guided sel-help

interventions with clients with bulimia nervosa, including

posttreatment and ollow-up assessment at the most distant

ollow-up point.

Frequency of binging for guided self-help. According to

Table 1, the d+ was 0.62 (medium eect; i = 17, n = 472) or

the single-group guided sel-help condition and 0.70 (mediumto large eect; i = 6, n = 263) or the wait-list condition.

Both were signifcantly higher than zero, meaning the null

hypothesis could be rejected and a conclusion made that the

treatment was eective. Homogeneity was displayed within

both eect size groupings. No TAU trials were located, and a

nonsignifcant d+ o 0.50 (medium eect; i = 1, n = 39) was

derived or the single placebo trial.

For the guided sel-help treatment, ollow-up studies us-

ing a single-group analysis resulted in a signifcant positive

result (d+ = 0.67; medium to large eect; i = 11, n = 337),

but the single placebo study (d+ = 0.13; small eect; i = 1, n

= 39) was not dierent rom zero. No TAU or wait-list guided

sel-help ollow-up trials were located. The distribution o

eect sizes or the single-group analysis was homogeneous.

Frequency of purging for guided self-help.Purging behav-

ior results or the guided sel-help condition are presented in

Table 2. The d+ was 0.50 (medium eect; i = 17, n = 482) or

the single-group guided sel-help condition and 1.37 (large

eect; i = 3, n = 151) or the wait-list condition. Both were

signifcantly higher than zero and displayed homogeneity

within both eect size groupings. No TAU trials were located,

and a nonsignifcant d+ o 0.48 (medium eect; i = 1, n = 39)

was derived or the single-placebo trial.

For the guided sel-help treatment, ollow-up studies using

single-group methodology resulted in a signifcant positiveresult (d+ = 0.65; medium to large eect; i = 11, n = 337), but

the single-placebo study (d+ = 0.17; small eect; i = 1, n =

39) was not dierent rom zero. No TAU or wait-list guided

sel-help ollow-up trials were located. The distribution o

eect sizes or the single-group analysis was homogeneous.

Frequency of laxative use for guided self-help.No wait-list,

TAU, or placebo trials were located or the posttreatment or

the ollow-up conditions or laxative use. According to Table

3, the d+ or the single-group guided sel-help condition was

0.26 (small eect; i = 9, n = 217), which was signifcantly

higher than zero and represented a homogeneous grouping o

eect sizes. On ollow-up, the single-group set o studies also

yielded a homogeneous set o eect sizes and a signifcantd+ o 0.54 (medium eect; i = 6, n = 163), indicating that

participants actually used laxatives less on ollow-up than at

the conclusion o treatment.

Bulimia self-ratings for guided self-help.No placebo trials

were located, but Table 4 indicates that the other three conditions

were signifcantly higher than zero or the bulimia sel-rating

outcome variable analysis. The d+ was 0.58 (medium eect; i =

7, n = 220) or the single-group guided sel-helpcondition, 1.25

or the wait-list condition (large eect; i = 3, n = 192), and 0.61


11/21

162

TABLE4

SmmaryStatstcsf

ortheSelf-ReportBlmaOtc

omeVarable

ComparsonGrop

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Note.Self-reportbulimiaisthed

ependentvariable.k=numberofstudies;i=numberofeffectsizesderived;d+=meaneffectsizeestimate;95%CI=95%

confidenceinterval;

>0=thed+wasgreaterthan

0;SigDiff=significantdifferencebetwe

enthecounseling/psychotherapyandg

uidedself-helpconditions(Yes=signific

antdifference;No=

nosignificantdifferences);Q(df)=thehomogeneityindexforthegiven

degreesoffreedom;I2=ahomogeneityindexpresentedasapercentage;YesintheHomogeneous




treatment-as-usual


Tm

e

Treatment

k

i

n

d+

95%C

i

>0

SgDffFal-SafeN

Q(df)

I2

Homogeneos

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

32

7

3

3

3

2

2

0

20

5

0

0

0

1

2

0

44

7

5

3

3

2

4

29

5

1

4

1,151

220

143

192

112

125

158

709

157

109

158

0.81

0.58

0.99

1.25

0.93

0.61

0.62

0.88

0.69

0.53

0.38

[0.67,0.95]

[0.42,0.75]

[0.34,1.64]

[0.86,1.64]

[0.45,1.40]

[0.23,0.99]

[0.28,0.97]

[0.76,1.01]

[0.45,0.93

[0.14,0.92]

[0.11,0.87]

YesYesYesYesYesYesYes

YesYes

YesNo

Yes

No

No

Yes

3,551

409

496

374

278

122

249

1,764

344

53

153

22.41(43)

6.18(6)

4.01(4)

2.08(2)

1.81(2)

0.04(1)

2.47(3)

27.72(28)

3.82(4)

3.21(3)

0.0

2.9

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

YesYesYesYesYesYesYes

YesYes

Yes


12/21

163

TABLE5

SmmaryStatstcsfo

rtheBodyDssatsfactonOtcomeVarable

ComparsonGrop

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Singlegroup

Singlegroup

Wait-list

Wait-list

TAUTAUPlacebo

Placebo

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Posttest

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Follow-up

Note.Bodydissatisfactionisthedependentvariable.


numberofeffectsizesderived;d+=meaneffectsizeestimate;95%C

I=95%

confidence

interval;>0=thed+wasgre

aterthan0;SigDiff=significantdiffere

ncebetweenthecounseling/psychotherapyandguidedself-helpconditions(Yes=significantdiffer-

ence;No=nosignificantdiffe

rences);Q(df)=thehomogeneityindex

forthegivendegreesoffreedom;I2=

ahomogeneityindexpresentedasape

rcentage;Yesinthe

Homogeneouscolumnmeans

thed+comprisesahomogeneousgroupingofeffectsizes.Posttest=themea

surementwastakenatthetermination

oftreatment;TAU=

treatment-as-usualcondition;Follow-up=themeasurementwastaken

atthelongestposttreatmentfollow-upavailable.

Tm

e

Treatment

k

i

n

d+

95%C

i

>0

SgDffFal-SafeN

Q(df)

I2

Homogeneos

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

Counseling

Self-help

28

11

3

3

4

1

0

0

17

8

0

0

1

1

0

0

41

11

5

4

4

1

24

8

1

1

1,424

384

143

222

218

112

768

268

71

109

0.50

0.38

0.66

0.71

0.60

0.42

0.56

0.54

0.58

0.51

[0.40,0.59]

[0.23,0.54]

[0.29,1.03]

[0.26,1.15]

[0.33,0.88]

[0.04,0.80]

[0.46,0.66]

[0.34,0.74]

[0.10,1.06]

[0.12,0.90]

YesYesYesYesYesYes

YesYes

YesYes

Yes

No

No N

oNo

2,034

420

143

282

241

42

1,349

429

58

51

35.65(40)

8.47(10)

3.00(4)

2.70(3)

0.52(3)

21.80(23)

8.22(7)

0.0

0.0

0.0

0.0

0.0

0.0

14.8

YesYesYesYesYes

YesYes


13/21


April 2013

Volume 91164

Erord et al.

(medium to large eect; i = 2, n = 125) or the TAU condition.

Homogeneity was demonstrated in all conditions.

For the guided sel-help treatment, no placebo or wait-list

ollow-up trials were located. Follow-up studies using a single

group resulted in a signifcant positive result (d+ = 0.69; me-

dium to large eect; i = 5, n = 157), as did the single TAU study

(d+ = 0.53; medium eect; i = 1, n = 109). The distribution oeect sizes or the single-group analysis was homogeneous.

Body dissatisfaction self-ratings for guided self-help. Ac-

cording to Table 5, the d+ was 0.38 (small to medium eect; i

= 11, n = 384) or the single-group guided sel-help condition,

0.71 (medium to large eect; i = 4, n = 222) or the wait-list

condition, and 0.42 (small to medium eect; i = 1, n = 112)

or the TAU condition. All three conditions were signifcantly

higher than zero, and homogeneity was displayed within each

eect size grouping. No placebo trials were located.

For the guided sel-help treatment, ollow-up studies using

a single-group design (d+ = 0.54; medium eect; i = 8, n =

268) and a single TAU study (d+ = 0.51; medium eect; i = 1,

n = 109) resulted in signifcant positive results. No placebo or

wait-list sel-help ollow-up trials were located. The distribution

o eect sizes or the single-group analysis was homogeneous.

Is There a Dierence Between Counseling/

Psychotherapy and Guided Sel-Help Interventionsin the Treatment o Clients With Bulimia Nervosa?

Dierences between combined eect sizes or counseling/

psychotherapy and guided sel-help conditions were deter-

mined by comparing the d+ o the guided sel-help condition

with the 95% CI range associated with the counseling/psycho-

therapy d+. We reasoned that the higher number o studies and

samples sizes accompanying the counseling/psychotherapy d+made that range more stable than the range associated with

the guided sel-help studies. This comparison is similar to the

null testing procedure used earlier; that is, i the d+ or the

guided sel-help condition alls outside o the 95% CI range

or the counseling/psychotherapy d+, the null hypothesis o no

dierence can be rejected and the possibility o a signifcant

dierence between mean eect sizes o the two conditions can

be considered. The summary decision or each comparison is

designated in the Sig Di column o Tables 1 to 5. I the

designation is yes, then a signifcant dierence does exist; i

the designation is no, the null hypothesis was retained.

Binging. According to the results reported in Table 1,

counseling/psychotherapy was signifcantly more eectivethan guided sel-help at treatment termination or both the

single-group (counseling/psychotherapy d+ = 0.71, i = 70;

guided sel-help d+ = 0.62, i = 17) and wait-list (counseling/

psychotherapy d+ = 0.99, i = 14; guided sel-help d+ = 0.70,

i = 6) comparisons. No dierence was noted in the placebo

comparison (counseling/psychotherapy d+ = 0.26, i = 8;

guided sel-help d+ = 0.50, i = 1). At ollow-up, there was no

signifcant dierence or either the single-group (counseling/


i = 11) or the placebo (counseling/psychotherapy d+ = 0.77,

i = 6; guided sel-help d+ = 0.13, i = 1) conditions.

Purging. The results in Table 2 indicate that counseling/

psychotherapy yielded signifcantly better results than guided

sel-help or the single-group condition (counseling/psychotherapy

d+ = 0.63, i = 63; guided sel-help d+ = 0.50, i = 17) but not

or the wait-list (counseling/psychotherapy d+ = 0.98, i = 16;guided sel-help d+ = 1.37, i = 3) or placebo (counseling/


i = 1) conditions. No dierence was noted at ollow-up or

either the single-group (counseling/psychotherapy d+ = 0.71,

i = 36; guided sel-help d+ = 0.65, i = 11) or the placebo

(counseling/psychotherapy d+ = 0.31, i = 4; guided sel-help

d+ = 0.17, i = 1) conditions.

Laxatives.The results in Table 3 indicate that counseling/

psychotherapy was signifcantly better at reducing the use o

laxatives than guided sel-help in single-group trials (counsel-

ing/psychotherapy d+ = 0.45, i = 17; guided sel-help d+ =

0.26, i = 9), but at ollow-up, guided sel-help was superior

to counseling/psychotherapy (counseling/psychotherapy d+

= 0.24, i = 5; guided sel-help d+ = 0.54, i = 6). No wait-list,

placebo, or TAU comparisons were available.

Self-report bulimia scales.As seen in Table 4, counseling/

psychotherapy produced signifcantly better sel-report bulimia

scale outcomes or the single-group comparison (counseling/

psychotherapy d+ = 0.81, i = 44; guided sel-help d+ = 0.58, i

= 7) but no signifcant dierences or the wait-list (counseling/

psychotherapy d+ = 0.99, i = 5; guided sel-help d+ = 1.25, i =

3) or TAU (counseling/psychotherapy d+ = 0.93, i = 3; guided

sel-help d+ = 0.61, i = 2) conditions. On ollow-up, the supe-

riority o counseling/psychotherapy over guided sel-help was

maintained (counseling/psychotherapy d+ = 0.88, i = 29; guidedsel-help d+ = 0.69, i = 5). No wait-list, TAU, or placebo ollow-

up comparisons were available or sel-report bulimia scales.

Body dissatisfaction.Table 5 results indicate that counseling/

psychotherapy was superior to guided sel-help interventions

in single-group studies (counseling/psychotherapy d+ = 0.50,

i = 41; guided sel-help d+ = 0.38, i = 11) but not wait-list

(counseling/psychotherapy d+ = 0.66, i = 5; guided sel-help

d+ = 0.71, i = 4) or TAU (counseling/psychotherapy d+ = 0.60,

i = 4; guided sel-help d+ = 0.42, i = 1) comparison studies.

There were no dierences between counseling/psychotherapy

and guided sel-help on ollow-up or either single-group

(counseling/psychotherapy d+ = 0.56, i = 24; guided sel-

help d+ = 0.54, i = 8) or TAU (counseling/psychotherapy d+= 0.58, i = 1; guided sel-help d+ = 0.51, i = 1) conditions.

Discussion

Is Counseling/Psychotherapy Eective or the

Treatment o Clients With Bulimia Nervosa?


Counseling/psychotherapy is quite eective in the treatment

o clients with bulimia nervosa. In nearly all the observed


14/21


April 2013

Volume 91 165


comparisons in our meta-analysis, counseling and psycho-

therapy resulted in positive average weighted eect sizes

(d+) o at least small to medium eects that were signif-

cantly higher than zero at termination o treatment (17 o 18

comparisons; 94%) across all our conditions (single group,

wait-list, TAU, and placebo) and all fve outcome measures

(binging, purging, laxative use, bulimia sel-ratings, andbody dissatisaction sel-ratings). However, only about hal

o all ollow-up study comparisons (six o 13 comparisons;

46%) across all conditions and outcome variables resulted in

eect size averages signifcantly greater than zero, although

these eects were also primarily medium in size. All tests o

homogeneity (Cochrans Q andI2) indicated signifcant homo-

geneity and no eects o moderating or mediating variables.

So counseling/psychotherapy is eective in producing

short-term positive therapeutic changes in clients with symp-

toms o bulimia but is inconsistently long-lasting and resistant

to relapse. The results o the current meta-analysis are similar

to previous meta-analyses o bulimia treatment (Ghaderi &

Anderson, 1999; Lewandoski et al., 1997; Thompson-Brenner

et al., 2003; Whitbread & Mcgown, 1994; Whittal et al.,

1999), but our meta-analysis includes ar more clinical trials

o recent publication, with more diverse study designs, and

with disaggregated outcome variables. It is also the frst to

use a random-eects model on a large sample o theoretically

diverse approaches, which tends to result in more conservative

eect size estimates.

The absence o mediator or moderator variables means

that no dierences were detected among various approaches

to counseling/psychotherapy. This means that no theoretical

approach seemed superior to any other and that individual,

group, and systemic approaches appeared to be equivalent.Both o these fndings may help clariy previous contrary

conclusions. For example, Thompson-Brenner et al. (2003)

concluded in a random-eects meta-analysis that strict behav-

ior therapy (with no cognitive component) was more eective

than CBT in reducing purging behaviors, although these

results were based on a handul o trials in each condition.

Likewise, the study o dierential eectiveness o individual

versus group interventions in bulimia treatment led Fettes and

Peters (1992) to conclude that group therapy was superior,

whereas Thompson-Brenner et al. (2003) and Shapiro et al.

(2007) reached the opposite conclusion. As the number o

clinical trials o treatment o bulimia nervosa continues to

accumulate, the power o such analyses also increases (Corn-well, 1993; Cornwell & Ladd, 1993). Thereore, as evidence

continues to accumulate, these previous conclusions will be

reexamined and refned in the aggregated context provided

by meta-analyses.

A continuing concern is the lack o substantive evidence

o the lasting eects o counseling/psychotherapy in the

treatment o clients with bulimia nervosa. Hal o the com-

parisons in the current meta-analysis indicate lasting eects

at ollow-up, whereas the other hal do not. Perhaps this is

due to the observation that ewer than hal o all clinical tri-

als conducted ollow-up components. Fewer studies reduce

the power o analyses, and some optimism can be gained

by inspecting the ollow-up categories in Tables 1 to 5 and

learning that many o these average eect sizes are medium

in magnitude despite the act that they are composed o ewer

than fve studies.

Is Guided Sel-Help Eective or the Treatmento Clients With Bulimia Nervosa?


A number o studies have been published recently explor-

ing the efcacy o sel-help and guided-help interventions.

Similar to the results or counseling/psychotherapy, in nearly

all observed instances, sel-help or guided sel-help resulted

in positive average weighted eect sizes (d+) o medium e-

ects that were signifcantly higher than zero at termination o

treatment (11 o 13 comparisons; 85%) across all our condi-

tions (single group, wait-list, TAU, and placebo) and all fve

outcome measures (binging, purging, laxative use, bulimia

sel-ratings, and body dissatisaction sel-ratings). But the

ollow-up study comparisons were a bit more positive than

or counseling/psychotherapy, because seven o nine com-

parisons (78%) resulted in eect size averages signifcantly

greater than zero. These eects were also primarily medium

in size. More important, all tests o homogeneity (Cochrans

Q andI2) indicated signifcant homogeneity and no eects

o moderating or mediating variables.

Thereore, as with counseling/psychotherapy, guided sel-

help is also eective in producing short-term positive changes

in clients with symptoms o bulimia nervosa and appears

to yield more substantial lasting results that display greaterresistance to relapse than does counseling/psychotherapy.

This is not so surprising when one considers that guided

sel-help interventions rely on client motivation or success.

Although this selection actor is potentially problematic in

all experimental research, it may be particularly problematic

when researchers recruit participants specifcally or a sel-

help study; that is, participants who are not sel-motivated

may remove themselves rom the study at higher rates than

highly motivated participants. Also, only 29 guided sel-help

articles were selected into this meta-analysis, ar ewer than

the 82 articles exploring the eectiveness o counseling/

psychotherapy, so the results o the current meta-analysis

must be viewed with caution.

Is There a Dierence Between Counseling/

Psychotherapy and Guided Sel-Help Interventions

in the Treatment o Clients With Bulimia Nervosa?

Few head-to-head studies pitting counseling/psychotherapy

against guided sel-help were available, so the comparisons

between these two broad approaches to the treatment o clients

with bulimia nervosa were based on studies contained in 111

dierent articles conducted by a multitude o researchers


15/21


April 2013

Volume 91166

Erord et al.

around the world and summarized in this meta-analysis. In

single-group study comparisons, which had the largest num-

ber o trials by ar, counseling/psychotherapy was superior

to guided sel-help interventions across all fve outcome

variables o binging, purging, use o laxatives, sel-report

bulimia ratings, and body dissatisaction. However, this

advantage was only maintained at ollow-up on the outcomemeasures o laxative use and sel-report bulimia ratings. In

all other conditions (i.e., wait-list, TAU, and placebo) and or

all other outcome variables (i.e., binging, purging, and body

dissatisaction), guided sel-help interventions were just as

eective as counseling/psychotherapy at both termination and

ollow-up. O course, ar ewer studies using wait-list, TAU,

and placebo comparison conditions are currently available in

the literature than studies using single-group designs.

The results o any meta-analysis must be interpreted with

caution because o methodological limitations. Overall,

however, these results provide positive indications or the

eectiveness o both counseling/psychotherapy and guided

sel-help or the treatment o clients with bulimia nervosa.

The results also suggest the need or head-to-head RCTs o

these two increasingly common approaches.

In addition, although cost analyses were not available, it

stands to reason that the guided sel-help approach may be

available to clients at a lower cost compared with counseling/

psychotherapy. This consideration, coupled with the impor-

tance o sel-motivation in any guided sel-help procedures,

may make sel-help or guided sel-help an efcacious frst

line o deense in helping clients with bulimia nervosa. Fur-

ther research should certainly explore the cost-eectiveness,

time-eectiveness, and overall treatment efcacy o guided

sel-help approaches to helping clients with bulimia.

Limitations of This Meta-Analysis

The current meta-analysis used rigorous methodological

procedures. We conducted exhaustive searches o published

literature and required a nine-level process or inclusion o

relevant studies, including that each use a standardized outcome

measure. When analyzing results, we used a random-eects

model to enhance generalizability and conservative statistical

assumptions, such as weighting eect sizes or inverse variance

and conducting two tests or homogeneity. We also assessed

or publication bias using both unnel plots and computation

o ail-saeNs.As a result, the aorementioned conclusions areprobably generalizable across relevant populations, treatment

variations, outcome variables, and research designs.

Despite, or because o, these selection protocols, some

study limitations may still exist. For example, the rigorous

selection criteria meant to enhance study quality may have

led to the elimination o viable studies, thus resulting in some

publication bias. Although we maintain that study quality

is important, inclusion o lower quality studies sometimes

alters the results o a meta-analysis (Whiston, Rahardja,

Eder, & Tai, 2011). O course, variations in characteristics

o clinical trials did occur; or example, ewer than hal o

all clinical trials used a standardized treatment manual, and

others ailed to provide sufcient inormation related to

some design sample or treatment procedures. Fortunately, all

random-eects analyses indicated signifcant homogeneity o

eect size estimates, so the absence o this inormation didnot aect mediator analyses.

However, perhaps the greatest limitation was the small

number o studies available or some comparisons. As

Cornwell (1993) and Cornwell and Ladd (1993) indicated,

sufcient power is generally gained in a meta-analysis when

the number o studies approaches or exceeds 20 clinical trials.

Although most o the single-group comparisons or counseling/

psychotherapy easily exceeded this criterion, most o the

analyses conducted on comparison conditions or or the

guided sel-help analyses did not. As additional studies o

the treatment o bulimia nervosa accumulate in uture years,

the power o these analyses can be expected to increase, thus

avoiding Type II errors. This was a particularly problematic

issue in analyses o ollow-up results to determine the staying

power o interventions.

Implications for Counseling Practiceand Research

The current meta-analysis represents the most recent and

largest study o treatment efcacy or bulimia nervosa. It

used a random-eects model, which yields a conservative

estimate o outcomes counselors can reasonably expect in

clinical practice. Both counseling/psychotherapy and guided

sel-help approaches appear to lead to clinically signifcant re-ductions in binging, purging, laxative use, sel-report bulimia

ratings, and body dissatisaction ratings. The eect sizes are

generally medium (d+ ~.50) and have been substantiated by

previous meta-analytic studies (Ghaderi & Anderson, 1999;

Lewandoski et al., 1997; Thompson-Brenner et al., 2003;

Whitbread & Mcgown, 1994; Whittal et al., 1999). Evidence

o long-term efcacy o bulimia treatment is inconsistent,

although guided sel-help approaches appear to hold up

better over time than counseling/psychotherapy. The lack o

consistent display o long-term efcacy could be due to the

relatively smaller numbers o ollow-up studies available in

the extant literature, which reduces the power o analyses.

This suspicion is bolstered given that most o the ollow-upcomparisons in this meta-analysis resulted in moderate eect

sizes. Additional clinical trials with ollow-up phases that

use wait-list, TAU, and placebo control designs are needed

to clariy the issue o long-term efcacy.

It is possible that booster or ollow-up sessions could also

enhance the long-term efcacy o counseling/psychotherapy

treatment, as has been suggested or the treatment o depres-

sion (Erord et al., 2011). It is unortunate that time and

resources are expended by clients with bulimia nervosa to


16/21


April 2013

Volume 91 167


gain improvement over debilitating symptoms, oten resulting

in medium to large eects o treatment, only to have those

gains reduced ater treatment terminates. Critics o mental

health care treatment and a fnancially burdened public are

right to be skeptical o such here today, gone tomorrow

treatment results. So it is incumbent on counselors and

counseling researchers to determine i certain approachesto the treatment o bulimia and implementation o booster

or ollow-up sessions lead to consistently positive outcomes

over the long term.

Cost-eectiveness is an essential consideration in the

context o the current health care debate. It is interesting

that not a single study or meta-analysis on the treatment

o bulimia nervosa measured the cost-eectiveness o the

interventions used, whether or counseling/psychotherapy or

guided sel-help. Counselors and counseling researchers must

explore the costs involved in dierent approaches to treating

bulimia nervosa and other disorders or issues that clients

present to surmise and put into practice the most time- and

cost-eective practices.

An interesting result with implications or clinical

practice was that, with only a ew exceptions, counseling/

psychotherapy and guided sel-help were equally eective,

and perhaps the latter had even better long-term efcacy.

Additional research is needed to confrm and extend these

results and the cost-eectiveness o each approach, but it

is interesting to postulate that sel-help or guided sel-help

approaches may be an eective initial intervention to help

sel-motivated individuals with bulimia nervosa signifcantly

reduce symptoms o binging, purging, and laxative use, as

well as reduce problematic cognitive displays or perceptions

related to bulimia or body dissatisaction. It will be importantthat any protocol or implementing sel-help treatment, either

in written or online ormats, should be standardized and

thoroughly evaluated to enhance eectiveness and minimize

risk to a vulnerable clientele.

At this point in the genesis o treatment or bulimia ner-

vosa, we need more RCTs that measure efcacy at termina-

tion, but especially at short- and long-term ollow-up points.

Even though wait-list and placebo studies are valuable, TAU

studies are needed most. TAU studies have the advantage

o not withholding active treatment rom control group

participants, instead providing participants with a treatment

approach they would normally receive i they presented

or treatment o bulimia at an outpatient acility. Ater all,although it is valuable to know that an intervention is better

than nothing, it is more valuable to know whether the planned

bulimia intervention is better than what the client would

have received under regular circumstances (e.g., supportive

counseling, case management). Use o a TAU comparison

group also minimizes the ethical dilemma o withholding a

viable treatment rom participants in the wait-list or placebo

condition until the completion o the control phase (Erord

et al., 2011; Weisz, McCarthy, & Valeri, 2006).

Finally, even though each study selected into the current

meta-analysis was published in a reereed journal, and even

though the studies were published over a 30-year period, the

adequacy o descriptions o study and sample characteris-

tics was highly variable. Oten, critical inormation needed

to replicate the procedures o the treatment in research or

clinical practice was absent. Such lapses render the studiesunhelpul in moving counseling practice orward. That is, i

a studys results show a particular treatment to be eective in

reducing the requency o binging and purging episodes but

practitioners reading the article cannot replicate the treatment,

what has been gained by publishing the research? Although

much progress has been made over the past several decades,

journal editors and editorial board members must redouble

their insistence that critical study and sample characteristics

be included in published articles. Relatedly, it is incumbent

on researchers to use standardized treatment protocols that

interested readers and publishers can access to better under-

stand and replicate treatment procedures.

References

*References marked with an asterisk indicate studies included

in the meta-analysis.

*Agras, W. S., & Rossiter, E. (1992). Pharmacologic and cognitive

behavioral treatment or bulimia nervosa: A controlled comparison.

American Journal of Psychiatry, 149, 8287.

*Agras, W. S., Rossiter, E. M., Arnow, B., Telch, C. F., Raeburn, S.

D., Bruce, B., & Koran, L. M. (1994). One-year ollow-up o

psychosocial and pharmacologic treatments or bulimia nervosa.

Journal of Clinical Psychiatry, 55, 179183.

*Agras, W. S., Schneider, J. A., Arnow, B., Raeburn, S. D., & Telch,C. F. (1989). Cognitivebehavioral and response-prevention treat-

ments or bulimia nervosa.Journal of Consulting and Clinical

Psychology, 57, 215221.

*Agras, W. S., Walsh, T., Fairburn, C. G., Wilson, G. T., & Kraemer,

H. C. (2000). A multicenter comparison o cognitivebehavioral

therapy and interpersonal psychotherapy or bulimia nervosa.

Archives of General Psychiatry, 57, 459466.

American Psychiatric Association. (2000).Diagnostic and statisti-

cal manual of mental disorders (4th ed., text rev.). Washington,

DC: Author.

*Bachar, E., Latzer, Y., Kreitler, S., & Berry, E. M. (1999). Empirical

comparison o two psychological therapies: Sel psychology and

cognitive orientation in the treatment o anorexia and bulimia.

Journal of Psychotherapy Practice & Research, 8, 115128.

*Bailer, U., Zwaan, M., Leisch, F., Strnad, A., Lennkh-Wolsberg,

C., El-Giamal, N., . . . Kasper, S. (2004). Guided sel-help versus

cognitivebehavioral group therapy in the treatment o bulimia

nervosa.International Journal of Eating Disorders, 35, 522537.

doi:10.1002/eat.20003

*Banasiak, S. J., Paxton, S. J., & Hay, P. (2005). Guided sel-help

or bulimia nervosa in primary care: A randomized controlled

trial.Psychological Medicine, 35, 12831294.


17/21


April 2013

Volume 91168

Erord et al.

*Bara-Carril, N., Williams, C. J., Pombo-Carril, M. G., Reid, Y.,

Murray, K., Aubin, S., . . . Schmidt, U. (2004). A preliminary

investigation into the easibility and efcacy o a CD-ROM-

based cognitivebehavioral sel-help intervention or bulimia


doi:10.1002/eat.10267

*Bauer, B. G. (1984). Bulimia: A review o a group treatmentprogram. Journal of College Student Personnel, 25, 221227.

*Becker-Stoll, F., & Gerlingho, M. (2004). The impact o a

our-month day treatment programme on alexithymia in eating

disorders.European Eating Disorders Review, 12, 159163.

*Bell, L., & Hodder, L. (2001). An evaluation o a supervised sel-

help programme or bulimic disorders. Clinical Psychology &

Psychotherapy, 8, 252262. doi:10.1002/cpp.281

*Berry, D. M., & Abramowitz, S. I. (1989). Educative/support groups

and subliminal psychodynamic activation or bulimic college

women.International Journal of Eating Disorders, 8, 7585.

*Bohanske, J., & Lemberg, R. (1987). An intensive group process-

retreat model or the treatment o bulimia. Group, 11, 112137.

*Brambilla, F., Dalle Grave, R., Calugi, S., Marchesini, G., Baroni,

S., & Marazziti, D. (2010). Eects o cognitivebehavioral

therapy on eating disorders: Neurotransmitter secretory re-

sponse totreatment.Psychoneuroendocrinology, 35, 729737.

doi:10.1016/j.psyneuen.2009.10.014

*Braun, D. L., Sunday, S. R., Fornari, V. M., & Halmi, K. A. (1999).

Bright light therapy decreases winter binge requency in women

with bulimia nervosa: A double-blind, placebo-controlled study.

Comparative Psychiatry, 40, 442448.

*Bulik, C. M., Sullivan, P. F., Carter, F. A., McIntosh, V. V., & Joyce,

P. R. (1998). The role o exposure with response prevention in

the cognitivebehavioural therapy or bulimia nervosa.Psycho-

logical Medicine: A Journal of Research in Psychiatry and theAllied Sciences, 28, 611623.

*Burton, E., & Stice, E. (2006). Evaluation o a healthy-weight

treatment program or bulimia nervosa: A preliminary random-

ized trial. Behaviour Research & Therapy, 44, 17271738.

doi:10.1016/j.brat.2005.12.008

*Carrard, I., Rouget, P., Fernndez-Aranda, F., Volkart, A.-C.,

Damoiseau, M., & Lam, T. (2006). Evaluation and deployment

o evidence based patient sel-management support program or

bulimia nervosa.International Journal of Medical Informatics,

75, 101109.

*Carter, J. C., & Fairburn, C. G. (1998). Cognitivebehavioral sel-

help or binge eating disorder: A controlled eectiveness study.

Journal of Consulting and Clinical Psychology, 66, 616623.

Cash, T. F., & Deagle, E. A. (1997). The nature and extent o body-

image disturbances in anorexia nervosa and bulimia nervosa:

A meta-analysis. International Journal of Eating Disorders,

21, 219.

*Charpentier, P., Marttunen, M., Fadjukov, S., & Huttunen, A.

(2003). CBT or adolescent eating disorder patients: Literature

review and a report o an open trial o group CBT or adoles-

cent outpatients with eating disorders. Psychiatria Fennica,

34, 8098.

*Chen, E., Touyz, S. W., Beumont, P. J. V., Fairburn, C. G., Grifths,

R., Butow, P., . . . Basten, C. (2003). Comparison o group and

individual cognitivebehavioral therapy or patients with bulimia


Cohen, J. (1988). Statistical power analysis for the behavioral sci-

ences (2nd ed.). Hillsdale, NJ: Erlbaum.

*Conners, M., Johnson, C., & Stuckey, M. (1984). Treatment obulimia with brie psychoeducational group therapy.Am

consiliere bulimia nervoasa

Documents