Guided self-help as the first step for bulimic symptoms: Implementation of a stepped-care model within specialized psychiatry


  • supported by 421-2004-2614 from Swedish Research Council and by the Söderström-Königska foundation and the Märta and Nicke Nasvell Foundation.



This study describes the implementation and effectiveness of the first step, guided self-help (GSH), in a clinical setting, of a stepped-care model of cognitive behavior therapy for patients with bulimic symptoms.


Eighty-nine patients participated.


In the intent to-treat analyses, the effect sizes were small to moderate (0.25–0.66). However, the effect sizes were substantially larger (0.44–1.66) for the patients who completed all the GSH sessions (45%). The majority of noncompleters stayed within psychiatric services. Supplementary treatments were mostly directed towards comorbid conditions, especially depression. Those discontinuing treatment in advance where characterized by more lifetime diagnoses and higher ratings on the restraint subscale of the EDE-Q.


GSH within specialized psychiatry might be effective for about 30% of the patients. There is no indication of patients losing their confidence in psychiatric services by being offered GSH as the first treatment. © 2011 by Wiley Periodicals, Inc. (Int J Eat Disord 2012)


Eating disorders, e.g., bulimia nervosa (BN), binge eating disorder (BED) and eating disorder not otherwise specified (EDNOS), are disabling conditions.1 Although cognitive behavioral therapy (CBT) is an efficient form of psychotherapy for these disorders [for review see Ref.2], there are limitations to its widespread implementation mainly because of a shortage of qualified therapists. To increase its availability, CBT has been simplified and delivered as self-help.3 A number of self-help books are available. One of which is entitled “Overcoming binge eating” by Christopher Fairburn4 and has been translated into Swedish. The efficacy of CBT-based self-help for eating disorders has been investigated in several studies, both as pure self-help (PSH) or guided self-help (GSH), where a limited number of brief visits, phone calls, or support through e-mail by a trained supporter is provided with the aim of encouraging the patient to follow the treatment program.5–7 Reviews of efficacy studies of PSH and GSH for eating disorders8–10 suggest that self-help may have some utility, even within specialized services, as a first step in treatment using a stepped care approach (i.e., programs that begin with the most cost-effective and least intensive treatment and move upward incrementally).11, 12 Using guided self-help in specialist unitsfacilitates high quality implementation of evidence-based treatment according to national guidelines for the treatment of ED that suggest self-help as a viable first step for BN and BED. However, data exploring the implementation, attrition, outcome, cost-effectiveness and potential negative consequences of self-help within a stepped care model for BN, BED, and related residual categories of ED are scarce.

Differences in research methodology and samples contribute to difficulties in interpreting results of the self-help studies. Patients presenting for treatment within specialized psychiatry are likely to be more burdened by comorbid psychiatric disorders13, 14 compared with participants recruited by advertisements in newspapers. Fairburn's self-help book for patients has been used in several studies,5, 15, 16 of which some were conducted in specialist settings.17, 18 However, it is not clear if the supporter's manual by Fairburn was used in these studies. GSH delivered according to a manual for supporters guarantees conformity regarding intensity and length. It also helps the patients focus on the methods and strategies suggested in the self-help book. The short supporter's manual by Fairburn19 helps to standardize the guidance when working with the corresponding self-help manual.4 However, patients in many studies are recruited through advertisements, which limits the generalization of the findings.

This study describes the implementation of the first step of a stepped-care model of CBT for patients with BN, BED, or related EDNOS diagnoses within specialized psychiatry. In this model, the first step was guided self-help using Fairburn's self-help book. Professionals provided support according to Fairburn's manual for supporters. The second and third steps were constituted by CBT group therapy and CBT individual therapy respectively. This model was implemented in a specialized eating disorder unit.

The first aim of this prospective clinical study was to examine effectiveness of GSH for bulimic symptoms delivered in a clinical psychiatric setting. The second aim was to analyze the clinical characteristics of those for whom self-help was useful and of those for whom it was insufficient. Another concern about stepped care models within psychiatric services is that they might lead some patients who could have responded to full treatment to dropout before they have had an opportunity to receive it. The third aim was, therefore, to explore what happened to the dropouts. Did they receive other treatments or not?



This study was conducted at the Eating Disorder Unit at the Uppsala Department of Psychiatry between May 1, 2005 and July 13, 2007. The Eating Disorder Unit carries out specialized services for eating disordered citizens of all ages and both sexes living in and near Uppsala, a city with a population of nearly 200,000. Patients can either refer themselves or be referred by general practitioners or by other psychiatric units. The Eating Disorder Unit is the only public specialized unit for treatment of eating disorders within Uppsala. All patients were consecutively included during the 26 months of study enrollment.

Patients met with a psychiatrist for an initial clinical evaluation. Before starting treatment, they were diagnosed using structured diagnostic interviews and answered to questionnaires (see assessments below). CBT for bulimic symptoms was delivered in the stepped care model. Pharmacological treatment was allowed during the study, but for antidepressants, which is also an efficient treatment for bulimic symptoms [for review see Ref.20 ], a distinction was recorded between those treated for at least six weeks ahead of CBT and those who received their prescription at their first assessment.

During the five initial months, the psychiatric staff (four nurses, five nursing assistants, one physiotherapist, and one psychologist) received supervision from the last author to act as program-led supporters. The support given followed the therapists' manual developed by Fairburn.19 In addition to receiving the self-help book “Overcoming binge eating,”4 which is translated into Swedish by the last author,21 patients met individually with the specially trained psychiatric staff for six to eight support sessions lasting for an average of 30 min.

Efficacy of new methods is threatened if the method is delivered in an inappropriate way. One and a half years after the last patient was included, a questionnaire was sent to all “supporters.” Respondents had to report on a scale from 1 to 5 (1 = does not agree at all, 2 = partly disagree, 3 = neither disagree nor agree, 4 = partly agree, 5 = fully agree) on questions related to the implementation process. Ten of eleven professionals responded. All reported that they had received appropriate education (4.5 ± 0.7). All, except one, reported that they got enough guidance (3.9 ± 1.0). All reported that they understood the difference between support and therapy (4.8 ± 0.4). Eight of the ten respondents considered that they had succeeded in behaving as a supporter (4.0 ± 0.7) and all responded that they had worked according to the therapists' manual. The majority believed that the GSH was an appropriate method (4.3 ± 0.8).

At posttreatment, the patient filled in the same questionnaires as before the treatment. Based on the Eating Disorder Questionnaire, question18 (Q 18), number of binge eating episodes experienced as out of control, were reported. The staff reported the number of sessions used for guidance, the patients' current weight and made a final GAF rating. We defined treatment completers as patients who attended all support sessions they were provided, ranging between six and eight.


Of the 297 eating disordered patients who attended an assessment session at the specialized clinic for eating disorders (ED), 161 patients had bulimic symptoms (see Fig. 1).

Figure 1..

Flowchart illustrating the recruitment and attrition during the study.

Among the 161 bulimic patients, 107 patients were recommended CBT-based GSH, while 44 were recommended other treatments. Patients with psychotic symptoms, patients judged as suicidal, or patients with complex comorbid disorders, such as severe substance abuse/dependence, were recommended other treatments instead of guided self-help CBT. As usual, decisions about treatments were made by the clinicians. Comparing those not recommended treatment with CBT GSH with those treated revealed significant differences according to eating disorder diagnoses (χ2 = 11.35: df = 3; p = .01) (assessed by SCID-I-RV interviews: see assessments). A total of 4.5% of those not recommended GSH had BED compared with 21.5% with those recommended. Corresponding numbers for eating disorder not otherwise specified (EDNOS) Type 3 were 40.9% vs. 19.6%. Ratings of depression according to MADRS-S (see assessments) differed significantly with those recommended guided self-help, being less depressed; 26.0 ± 8.7 vs. 21.3 ± 9.3 (t = 3.00; df = 149; p = .003; Cohen's d effect size = 0.52). Accordingly, those recommended self-help differed by having significantly higher GAF scores 49.8 ± 5.5 vs. 52.2 ± 5.5 (t = 2.41; df = 149; p = .02, Cohen's d effect size = 0.43).

Ten patients among the 161 recommended CBT-based GSH did not receive any treatment at all. They came for 1.2 (0.4) visits. The reasons for refraining treatment were not systematically recorded, but since they were diagnosed with an eating disorder, they were not excluded from treatment by the professional.

Eighty-nine patients (83.2%) engaged in GSH upon the specific recommendation to start their treatment with GSH as the first step. Among them, 47 patients (52.8%) were diagnosed with Bulimia Nervosa Purging type (BNP), 5 patients (5.6%) were diagnosed with Bulimia Nervosa non-Purging type (BNnonP), 17 patients (19.1%) were diagnosed with EDNOS Type 3 and finally, 20 patients were diagnosed (22.5%) with BED. Mean age was 25.0 ± 7.2 (range; 18–58) years. Gender distribution was extremely skewed, with 86 women (96.6%) and 3 men (3.4%). Comorbidity was common, with 48 (53.9%) fulfilling criteria for any mood disorders, 59 (66.3%) fulfilling criteria for any anxiety disorder and 5 (5.6%) fulfilling criteria for any substance related disorder (assessed by SCID-I-CV interviews: see assessments). Mean number of concurrent diagnoses was 3.0 ± 1.5 (range, 1–7). The mean Body Mass Index (BMI) was 23.9 ± 5.0 (range, 18–43) and mean rating on the Global Assessment of Functioning Scale (GAF) was 52.0 ± 5.6 (range, 35–71). A total of 30 patients (33.7%) were treated with antidepressants at the time of the first assessment. They maintained their medication treatment throughout the study. Six patients (6.7%) were prescribed antidepressants at the time of the first assessment.

After receiving a detailed description about the study, written informed consent was obtained from all 89 participants.

Dropout Analysis

A drop-out analyses was performed comparing participants (n = 89) and nonparticipants (n = 18) within the group who were recommended guided self-help (n = 107).

No significant differences were found between the groups according to eating disorder diagnoses; 58% versus 61% with BN and 42% versus 39% with ED NOS bulimic type (χ2 = 0.045; df = 1; p = .83, ω = 0.056). Neither were there any significant group differences according to the proportion of patients with any mood disorder; 54% vs. 61% (χ2 = 0.312; df = 1; p = .577, ω = 0.146), any anxiety disorder; 66% vs. 56% (χ2 = 0.754; df = 1; p = .385, ω = 0.223). Neither were there any differences according to total number of current diagnoses; 3.0 (1.5) vs. 2.5 (1.0) diagnoses (t = 1.73; df = 105; p = 0.092). No statistical analyses were performed due to small sample size and there was no occurrence of substance related disorders among the nonparticipants.


For assessment of ED we used the specific ED sectionfrom the Structured Clinical Interview for DSM, Research Version for Axis I Disorders (SCID-I-RV).22 Bulimic disorders were diagnosed within the following categories according to DSM-IV23: BNP, BNnonP, and EDNOS. This category was subdivided, in accordance with suggested examples in DSM-IV, into the following two categories: EDNOS Type 3—all criteria for BN are met except that the binge eating and inappropriate compensatory mechanisms occur at a frequency of less than twice a week or for a duration of less than 2 months, and EDNOS Type 6: BED—recurrent episodes of binge eating in the absence of the regular use of inappropriate compensatory behaviors characteristic of BN.

For assessment of other axis I diagnoses we used the Clinician Version (SCID-CV), a streamlined version of the SCID-I-RV. The SCID-I CV provides comprehensive assessment of most diagnoses, but for some there is only screening questions included (i.e., agoraphobia, social phobia, specific phobia, generalized anxiety disorder, and the somatoform disorders). If screening questions were assented, all criteria for that diagnosis were reviewed using Mini DSM-IV as a checklist. The first and second author performed the SCID interviews, except for six done by three residency/fellowship psychiatrists, all trained by the first author. The overall Kappa coefficient between the first and second author was 0.79 for categorical axis I diagnoses based on four randomly selected SCID-I interviews. Diagnoses were grouped according to DSM-IV into the following:

  • Any substance related disorder: alcohol dependence and abuse.

  • Any mood disorder: bipolar I, II and NOS, major depressive disorder, dysthymic disorder, depressive disorder NOS, and substance induced mood disorder.

  • Any anxiety disorder: panic disorder with and without agoraphobia, agoraphobia, specific phobia, social phobia, obsessive compulsive disorder, posttraumatic stress disorder, generalized anxiety disorder, substance induced anxiety disorder, and anxiety disorder NOS.

Length and weight were evaluated by an examination performed by a psychiatrist during the first appointment. BMI was calculated by weight (kg)/length (m).2

The Eating Disorder Examination Questionnaire (EDE-Q) is a 36-item measure adapted from the Eating Disorder Examination (EDE),24 which is a semistructured clinical interview assessing the key behavioral features and associated psychopathology of ED. The self-report version, EDE-Q, contains four subscales: restraint, eating concern, weight concern, and shape concern. The self-report questionnaire generates higher scores than the interview when assessing more complex features, such as binge eating in some studies,24, 25 but the agreement between the interview and the questionnaire on the above mentioned subscales is fairly good.25–27

The Body Shape Questionnaire (BSQ) is a self-report questionnaire28 measuring the extent of psychopathology of concerns about body shape, particularly the experience of “feeling fat”. We used the validated Swedish translation of the questionnaire.29

The Quality of Life Inventory (QOLI) consists of 16 items covering areas supposedly important to quality of life, such as health, self-respect and economy. Participants rate the importance of, and satisfaction with, each area. The QOLI is extensively evaluated for reliability, internal consistency and construct validity against a number of measures, both self-report and external nonself-report criteria, and has been found to meet the requirements for a useful measure of quality of life.30

The Rosenberg Self-Esteem Scale (RSE) scale is a widely used instrument for measuring global self-esteem. It consists of 10 items with a four-point response scale, from strongly agree to strongly disagree. The RSE has excellent psychometric properties.31, 32

Montgomery Åsberg Depression Rating Scale (MADRS-S) is designed to be particularly sensitive to treatment effects.33 The MADRS-S is the self-report version, shown to be equivalent to the Beck Depression Inventory, BDI,34 as a self-assessment instrument for depression, but the MADRS-S focuses on core depressive symptoms, and is less influenced by maladaptive personality traits.35

The GAF is a scale from 1 to 100.23 A single rating on the GAF scale integrates three different dimensions of functioning: psychological, social and occupational. The scale is recommended as a global severity measure in the assessment of outcome in routine mental health care.36–38 All participating professionals were previously trained in, and had long experience of, rating on the GAF-scale. However, 10 written case-vignettes of patients with ED symptoms were constructed and rated independently by the professionals, blind to each other. Intra-class correlation coefficient calculated from eight of twelve participating raters, the three first authors included, was 0.97.


Missing items were replaced by the median value of that item calculated from the whole group. If more than three items were missing within one questionnaire, the patient's response on that specific questionnaire was excluded. Overall, three EDE-Q from before treatment and 14 EDE-Q from after treatment were excluded because of missing data, but no other questionnaires were excluded, neither from before nor from after treatment.

Kappa statistics were used for inter-rater reliability concerning categorical axis I diagnoses. GAF ratings were analyzed for inter-rater reliability by using the intra-class correlation coefficient (ICC), two-way random effects model.39 Intra-class correlation coefficients are considered excellent if greater than 0.74, good if ranging from 0.60 to 0.74, and fair if ranging from 0.40 to 0.59.39 Group differences concerning continuous data were investigated by Student's t-test. For effect sizes, Cohen's d was calculated.40 Group differences according to categorical data were analyzed by Chi-square test with Fishers exact test when at least two cells had an expected frequency of less than five cases. For effect sizes, coefficient gamma was calculated. Change over time, from pretreatment to posttreatment, was explored using t-test or analysis of variance (ANOVA). Analyses of the treatment outcome were intent-to-treat analyses using the last observation carried forward (LOCF).41 An alpha level of 0.05 was used as a significance criterion for all the analyses. SPSS v. 13 was used for all analyses.

The study was approved by the ethics committee of the Uppsala University Hospital.



Forty patients (44.9%) received six to eight support sessions (8.0 ± 1.1), completing all the GSH sessions provided. Three patients decided not to start treatment and 46 quit treatment in advance. Noncompleters received a mean number of 3.2 ± 2.0 support sessions. According to 21 individuals, reasons for quitting were related to difficulties with the method, either that they did not improve (19 people) or were critical towards the method (two people). For 11 individuals, the reasons for quitting are unknown since they didn't show up at all after a few sessions (mean 2.6 ± 1.9). In addition, four individuals moved away from Uppsala and 10 individuals considered themselves sufficiently improved, even though their therapists recommended that they complete their treatment.

Treatment Outcome

For those who failed to comply with the final evaluation, pretreatment data was used as termination data in the intent-to-treat analyses presented in Table1. Because of missing data, the effect sizes are small to almost moderate (0.25–0.66) given the within subject nature of the comparisons.

Table 1. Effectiveness of guided self-help CBT for bulimic symptoms delivered in a clinical psychiatric setting
 Pre–Posttreatment Effects; ITT Analysis (n = 89)Perprotocol Treatment Effects (n = 39)
nPretreatment M (SD)Posttreatment M (SD)FpdnPretreatment M (SD)Posttreatment M (SD)Fpd
  1. Intent-to-treat-analyses (ITT) are presented for all 89 patients. Perprotocol analyses are presented for those 39 who finalized treatment and responded to posttreatment questionnaires.

No. of binges out of control/month EDE-Q188613.6 (9.9)8.7 (10.3)29.7<.0010.483811.9 (9.5)0.9 (1.6)52.3<.0011.61
EDE-Q: global score864.4 (0.9)3.7 (1.3)31.6<.0010.63374.2 (0.8)2.8 (1.2)61.4<.0011.37
EDE-Q: restraint864.1 (1.2)3.4 (1.6)34.1<.0010.49374.0 (1.1)2.2 (1.4)71.7<.0011.43
EDE-Q: eating concern863.8 (1.1)3.1 (1.5)34.3<.0010.53373.6 (1.1)2.0 (1.2)72.7<.0011.39
EDE-Q: shape concern865.1 (0.8)4.5 (1.4)20.8<.0010.53375.1 (0.7)3.8 (1.5)30.2<.0011.11
EDE-Q: weight concern864.5 (1.7)3.9 (1.9)19.8<.0010.33374.3 (1.0)3.0 (1.2)28.1<.0011.18
MADRS-S: depression8921.9 (9.5)18.0 (10.8)24.7<.0010.383120.6 (7.3)13.0 (8.6)25.2<.0010.95
BSQ: Body dissatisfaction8261.6 (11.5)55.9 (14.2)19.1<.0010.443861.1 (11.3)50.7 (15.2)18.8<.0010.78
RSE: Self-esteem854.5 (1.4)4.0 (1.9)14.4<.0010.30324.4 (1.4)3.2 (2.0)15.4<.0010.70
Quality of life800.4 (1.6)0.8 (1.6)9.4.0030.25300.4 (1.5)1.1 (1.6)8.8.0060.44
GAF8952.0 (5.6)57.3 (9.8)33.6<.0010.663952.6 (5.3)62.5 (10.4)38.4<.0011.20

Among the completers, 39 individuals had responded to some of the questionnaires at posttreatment. As seen in Table1, the effect sizes were larger for this group (0.44–1.66).

Treatment Completers vs. Noncompleters

Treatment completers were compared with noncompleters on clinical characteristics. Those discontinuing treatment in advance where characterized by more lifetime diagnoses and higher ratings on the restraint subscale of the EDE-Q. Overall, differences were small (Table2), and there were no significant differences according to eating disorder diagnosis comparing treatment completers vs. treatment noncompleters: BN 52.6% vs. 65.3%; BED 32.5% vs. 14.3%; ED UNS 17.5% vs. 20.4% (χ2 = 4.24: df = 3: p = .237: Effect size gamma = 0.29). Neither were there any significant differences, between completers and noncompleters, according to whether they fulfilled criteria or not for any mood, any anxiety or any substance related disorder, nor to sex.

Table 2. Comparison of pretreatment assessments between patients with bulimia nervosa symptoms completing treatment (n= 40) with guided self-help CBT and those terminating in advance (n = 49)
Clinical VariablesTreatment Completers M (SD)Treatment Noncompleters M (SD)Statistics
  1. d: measure of effect size (Cohen's d)

Age24.5 (7.2)25.4 (7.2)0.5787.5730.12
Age of ED onset16.5 (7.2)16.2 (3.8)0.2787.7880.05
Duration of ED8.0 (5.3)9.2 (7.2)0.9187.3640.19
EDE-Q: global score4.2 (0.7)4.5 (1.0)1.9884.0510.35
EDE-Q: restraint3.7 (1.1)4.5 (1.1)3.6384<.0010.73
EDE-Q: eating concern3.7 (1.1)3.9 (1.1)0.9084.3700.18
EDE-Q: shape concern5.0 (0.8)5.2 (0.9)0.7784.4420.23
EDE-Q: weight concern4.3 (0.8)4.6 (2.2)0.7884.4420.18
BSQ: Body dissatisfaction61.2 (11.3)61.9 (11.7)0.3080.7690.06
No. of binge-eating episodes with experienced loss of control during the last month (EDE-Q: 18)12.4 (9.1)14.6 (10.4)1.0484.3030.23
RSE: Self-esteem4.4 (1.4)4.6 (1.5)0.6483.5240.14
Quality of life0.6 (1.5)0.3 (1.6)1.0778.2900.19
Total number of current diagnoses2.9 (1.4)3.2 (1.6)0.8187.4180.20
Total number of life-time diagnoses4.0 (1.5)5.0 (2.4)2.4287.0180.50
MADRS-S: depression20.1 (8.0)23.3 (10.5)1.7087.1000.34
GAF52.6 ± 5.351.4 ± 5.81.0287.3130.22
BMI25.0 ± 5.023.0 ± 4.91.8375.0720.40

Antidepressants are effective treatment for bulimic symptoms. Among completers, 60% had no antidepressants during the study while 20% were treated with antidepressants at least 6 weeks before and during the study and 5% received their prescription during the initial evaluation. Among completers, 15% received antidepressive medication after the GSH treatment. Among noncompleters, the corresponding figures were 41% without medication, 45% medicated before and during the study, 8% starting medication at the study start and finally, 6% receiving antidepressants after GSH. There was a significant difference, with more noncompleters being on medication during the GSH treatment and with an earlier onset of such medication (i.e., treated for a longer period of time with antidepressants).


In total, patients attended a mean number of 15.8 appointments (visits) of any psychiatric services (SD = 14.3), ranging from 1 to 82. Within the whole group (n = 89), 19 (21%) attended five or less visits, all of whom were noncompleters. These 19 abandoned not only GSH, but also all other psychiatric services.

The 40 completers attended a mean number of 18.9 (12.4) visits, noncompleters attended a mean number of 13.4 (15.4) visits (t = 1.87; df = 87; p = .065). Among the 49 noncompleters, 30 patients (61%) received other treatments and among them, five patients received CBT for bulimic symptoms (Step 2 and 3, CBT group therapy and individual therapy). Consequently, the majority of noncompleters stayed within psychiatric services. These 30 patients attended a mean number of 19.9 (16.6) visits, ranging between 6 and 82.

Seventeen (42%) treatment completers and 14 (29%) noncompleters received no additional treatment than CBT GSH (see Table3). The majority of supplementary treatments were directed towards comorbid conditions, especially depression. A minority, 12 patients (14%), received Step 2 or 3 of CBT treatments towards eating disorder symptoms. Among those 12 patients, missing data resulted in only three patients having completed all assessments. In these three patients, EDE-Q total decreased from 4.7 (0.6) to 3.7 (0.6)

Table 3. Delivered treatment in patients with bulimia nervosa (n = 89) within specialized psychiatry, comparisons between GSH completers (n = 40) and noncompleters (n = 49)
Clinical VariablesTreatment Completers n (%)Treatment Noncompleters n (%)Statistics
Test StatisticsdfpEffect Size Gamma
  • a

    Fisher's exact test was used.

  • b

    One patient among noncompleters received both step 2 and 3.

  • c

    Mann-Whitney Test was used.

Only CBT GSH, no other treatment17 (42)14 (29)χ2 = 1.8821.190.30
CBT GSH with antidepressive medication16 (40)31 (63)χ2 = 4,781.030.44
Any pharmacological treatment16 (40)32 (65)χ2= 5.681.020.48
CBT step 2 (group therapy)a1 (2)3 (6)bχ2 =0.671.620.44
CBT step 3 (individual therapy)a6 (15)3 (6)bχ2 = 1.911.290.46
M(SD) of out-patient visits (CBT GSH included)c18.9 (12.4)13.4 (15.4)Z = 3.5187<.0010.40
0–5019 (39)   
6–2026 (65)19 (39)   
21–4012 (30)8 (16)   
>412 (5)3 (6)   


For the majority of the bulimic patients at this specialized psychiatric clinic, GSH was considered appropriate by the clinician to recommend as the first treatment since only one-fourth was excluded from this treatment.

Among patients who finalized GSH, a significant decrease in bulimic symptoms with effect sizes as high as 1.66 was recorded. The extent of decrease in bulimic symptoms in this clinical group is in accordance with previous studies.5, 15, 42

However, drop-out rate was very high and, as a consequence, the intention to treat analysis resulted in much lower effects. Most commonly, the patient terminated their treatment outside the eating disorder unit, but remained at the clinic. The professionals outside the eating disorder unit did not comply with our assessment procedures. However, for one of three patients, GSH was a sufficient treatment.

The rate of attrition, across studies of GSH for ED, varies widely, ranging from 0 to 69% (for review see Ref.9). The drop-out rate of 55% in the present study is higher than rates (23–29%) reported from other clinical trials performed within specialized psychiatry.6, 18, 43, 44 The patients in this study were informed about the stepped care model when offered GSH. This might have influenced those who wanted more support to quit prematurely to receive more intensive CBT. There were five drop-outs who did receive more CBT for bulimic symptoms after quitting GSH. However, even higher drop-out rates have been reported from primary care.16 Most important, the majority of those who quit GSH stayed within psychiatric services and received other treatments.

An important point of concern with stepped care models is that they might lead some patients who could have responded to full treatment (e.g., individual CBT-BN) to drop out before they had the opportunity to receive it. This concern is partly elucidated. In this naturalistic study, 14 (16%) did not receive any other treatments after prematurely finalizing GSH. We do not know if these patients would have responded to full CBT treatment with a therapist if this had been delivered immediately. However, the majority of patients quitting GSH received other treatments. Unfortunately, we do not have sufficient data to report on their eating disorder symptoms when they finalized their contacts. For those who proceeded in the stepped care model there is an indication of benefits from the initial treatment since there is a progressive decrease in symptoms. Since noncompleters where characterized by more lifetime diagnoses and higher ratings on the restraint subscale of the EDE-Q, these characteristics could be indicators for offering initial full CBT treatment with a therapist instead of GSH. Additionally, it is also likely to believe that patients with less restricted eating behaviors and less burdened by comorbidity should be the ones best suited for treatment based on GSH.

Strengths of this study are several. Most obvious is the high ecological validity. Examination took place in an ordinary clinical setting and no patients were recruited through ads or other formal or informal ways. Patients had no prior knowledge about the on-going study, which could have influenced the constitution of the sample. Therefore, the examined group is likely to be representative of other help-seeking, bulimic patients referred to a specialist clinic. Dropout analysis included all clinical variables evaluated in the same way as with the completers (i.e., through structured interviews). Participants in this study seem to be very similar to other clinical groups. For example, in the study by Palmer et al.,18 the distribution between different diagnostic subgroups was almost identical to this group with 59% vs. 58% having bulimia nervosa, 23% vs. 22% having BED and 18% vs. 19% in our study having eating disorder NOS. Another strength of this study is the comprehensive evaluation of psychopathology. Using structured interviews together with all other sources of information increases diagnostic reliability.45

Another strength of this study is data supporting high quality of the implementation process.46 The supporters experienced their education and guidance as appropriate. They reported that they understood the difference between support and therapy and that they had worked according to the supporters' manual.

The most obvious weakness of this study is the lack of knowledge about the course of eating disorder symptoms in the majority of both noncompleters and completers, due to a significant amount of missing data. Among both noncompleters and completers who received other treatments in addition to GSH, only a small minority wanted more treatment targeting eating disorder symptoms. The majority wanted and received other treatments, such as CBT for depression. We do not know how their bulimic symptoms developed, since no ratings of eating disorder symptoms were performed after other treatments. Data were also missing for those 12 who received more CBT, group or individual therapy for bulimic symptoms since questionnaires were inappropriately filled in or missing. This also illustrates the difficulties with performing clinical studies. We need to simplify the procedures to obtain data from the patients at assessment points and make it more accessible. Using Internet combined with reminders by means of SMS might be a viable option to traditional assessments. That would increase the possibility to track all participants for a longer period and make it easier for them to comply with the assessments. Another weakness is that we did not examine the results from treatment, in the same structured way, before the study started. We know that more patients received treatment after implementation of GSH (107 patients compared to around 60 patients during the same period) but we do not know if it was more or less efficient.

Finally, in an ultimate design to test the efficacy of self-help as a first step in a stepped care model, the patients should be randomized to either a stepped care model and followed-up throughout the entire treatment or to treatment as usual (e.g., CBT-BN with a therapist for 20 session) to calculate efficacy and cost-effectiveness.


The results from this study indicate that CBT-based GSH delivered in a stepped care model within specialized psychiatry is effective for about 30% of the patients. For those who receive more CBT it seems to offer a base for further interventions. There is no indication of patients losing their confidence in psychiatric services by being offered GSH as the first treatment.

Earn CE credit for this article!

Visit: for additional information. There may be a delay in the posting of the article, so continue to check back and look for the section on Eating Disorders. Additional information about the program is available at