Cognitive behaviour therapy response and dropout rate across purging and nonpurging bulimia nervosa and binge eating disorder: DSM-5 implications

The aims of the present study are twofold: 1) To determine the rate of response to CBT across BN-P, BN-NP and BED and 2) To describe the differences in the rate of dropout across the three diagnostic groups.

We hypothesized the finding of dimensional differences across the three diagnoses with BN-P representing the most severe and BED the least. Therefore, we expect the BED group to exhibit the most clinical improvement, followed by the BN-NP group and the BN-P group to show the least recovery. Likewise, we expect to find the highest risk of dropout in BN-P patients.

Every female patient over the age of 18 years who was diagnosed with BNP, BN-NP or BED according to DSM-IV-TR [28] criteria was invited to participate in this study. All participants were consecutively admitted for group-based outpatient treatment with no other psychotherapy at the time at the Eating Disorders Unit (University Hospital of Bellvitge), between 1998 and 2009. Patients were excluded if they presented with severe co-morbid psychopathological symptoms (e.g., suicidal attempts) requiring individual and/or inpatient therapy.

For the assessment, commonly applied questionnaires in the field of EDs, comprising the Eating Disorders Inventory-2 (EDI-2) [29] and the Symptom Checklist-Revised (SCL-90-R) [30] were employed.

Experienced psychologists and psychiatrists diagnosed all participants according to the DSM-IV-TR [28] criteria using a semi-structured face to face clinical interview (SCID-I) [33]. Additional sociodemographic-clinical information was collected including age, weight, marital status, education and occupation and clinical-psychopathological variables. As a standard procedure of clinical assessment in the ED unit of our hospital, all the participants completed the questionnaires individually and voluntarily before starting the treatment. The same assessment was repeated at the end of the treatment. Throughout the duration of the treatment, patients kept a daily food and purging diary [34]. These food diaries also collected information about daily frequency of binging, purging and exercise. The information of food diaries was used as a therapeutic tool during the treatment sessions, i.e., this information was discussed with the therapist and the rest of members of the group in every session in order to increase awareness about bulimic symptoms. Weekly binge-eating and purging frequency was determined by examining these food diaries and calculating their mean values.

The study was approved by the Ethics Committee of our institution (Ethics Committee of Clinical Research of the University Hospital of Bellvitge) and written informed consent was obtained from all participants.

Treatment consisted of a 22 outpatient 90-minute weekly sessions. There were a total of 8–10 patients per group. These comprised 6 initial sessions of psychoeducational brief group therapy [35] followed by 16 weekly outpatients sessions of CBT [32]. This program and its complementary material have already been manualized and published in Spanish [34] with demonstrated effectiveness [36]. BN and BED patients were placed in separated treatment groups, but both treatment groups were based on the same CBT program. Patients who completed treatment were assessed at end of CBT therapy and classified into three categories “full remission”, “partial remission” or “non-remission” group, which was based on treatment outcomes. Primary outcome was based on the food and purge diary and the response of some clinical questionnaires in the field of ED. The working definition of a “full remission” outcome required the absence of binging and purging (laxatives and/or vomiting) behaviors for at least 4 (consecutive) weeks and psychological improvement measured by clinical questionnaires. “Partial remission” was defined as substantial symptomatic improvement but still presence of residual symptoms (reduction of at least 50% of bulimic symptoms), and the patients who presented bad outcome was defined as “non-remission”.

The statistical analysis was carried out with SPSS 20 for Windows. Logistic regressions, adjusted by age, compared the criteria (dependent variables) risk of remission (full-partial-no remission) and the risk of therapy dropout (present vs absent) between the three diagnostic subtypes. Analysis of variance procedures (ANOVA, also adjusted by the covariate age), compared the quantitative outcomes analyzed in this study between the different diagnostic conditions. Survival analyses through Cox’s regressions adjusted by age compared the time to dropout of therapy. Survival analyses involve the modeling of time to event data whereby “death” (or failure) is considered an event (in this study the register of the dropout), allowing censored values (in this study right censored data identified patients who didn’t dropout, that is, those who stayed for the entire treatment). The models adjusted with survival in this work attempts to answer the next two questions: a) what is the fraction of sample which will survive (in this study, survive is equivalent to not dropout) past a certain time? and b) of those that survive, at what rate will they present the event (fail)? The statistical procedure with survival included all the participants at the beginning, since it considers as outcome the “time to the presence of a dropout” (in the case of non dropout, survival time is defined as time of follow-up for the participant). Due to the multiple comparisons, Bonferroni-Holm’s correction was used to prevent increase in Type I error (the total alpha level was established at 0.05). This method for adjusting global α-level is included into the closed-test-procedures and it controls the family-wise error rate, operating in a more powerful way than the usual Bonferroni’s-adjustment.

Data of this work correspond to a total sample of 454 eating disorder patients (327 BN-P, 40 BN-NP and 87 BED). There were statistical significant differences between BED and BN groups in several variables. A lower number of patients in the BED group was single. Patients in the BED group were also older and showed a significantly higher current, maximum and minimum Body Mass Index (BMI), developed their disorder at a later age of onset and suffered from it longer than the other two groups. See Table 1.

For subjects who completed therapy, the percentage of patients who were considered in full or partial remission differed according to diagnoses subtypes (Table 2) with a statistically higher number of patients in the “full remission” group among the BED diagnosis when compared to BN-P (p = .004) and BN-NP (p < .001). The study found no statistically significant differences in the number of patients who remitted between BN-P and BN-NP (p = .687). The remission rates for the total sample (including the participants who drop-out, as an intent-to-treat analysis) showed similar results: the probability of full remissions were 47.1% for BED (95% CI: 36.6% to 57.6%), 30.0% for BN-NP (95% CI: 15.8% to 44.2%) and 27.2% for BN-P (95% CI: 22.4% to 32.0%). The risk of dropout also differed between groups, and post-hoc comparisons indicated that the incidence ratio was statistically equal for BN-P and BN-NP (p = .657), but BED showed higher rates of dropout than BN-P (p < .001) and BN-NP (p = .035) Table 2.

Comparisons between participants who dropout and non-dropout into each diagnostic condition showed no statistical differences in civil status (p = .773) and employment status (p = .069), patients’ age (p = .248), onset of eating disorder (p = .197) or evolution of eating disorder (p = .590). Mean scores of EDI-2 and SCL-90-R at baseline (Table 3) did not achieve significant results, except for the EDI-2 “Maturity fears” into BN-NP patients: dropouts obtained higher mean than non-dropouts (11.83 vs 7.28; p = .045) Table 3.

Considering dropouts across diagnostic subtypes (BN-P, BN-NP and BED), no statistical differences emerged by civil status (p = .133), employment status (p = .271), onset of eating disorder (p = .167) and evolution of disease (p = .125), but patients who dropout were older in BED cohort (mean = 32.2, SD = 9.5) compared to BN-P (mean = 25.5, SD = 6.7) and BN-NP (mean = 22.8, SD = 3.7). Mean scores of EDI-2 and SCL-90-R at baseline were also statistically equal between diagnostic subtypes, with the exception of EDI-2 “Body dissatisfaction”, which achieved the higher mean for BED (23.0) compared to BN-P (18.4) and BN-NP (17.0) Table 4.

Figure 1 shows the plots for the survival function (at mean of covariate age), representing in the X-axis the number of sessions to the dropout of the treatment and in the Y-axis the cumulate survival probability (%). These curves represent the probability that a patient “survives dropout-free” for at least a specific time or longer, and they can be interpreted as a measure of the rate (velocity) of dropouts in each diagnostic condition. The best clinical result (low rate of dropout) corresponded to BN-P patients, followed by BN-NP. BED patients achieved a very different result, with more relevant slopes in the survival function. For BN patients (purgative and non-purgative), the high risk of dropout corresponds to the first two sessions of therapy (approximately 8% of the patients dropped out during this time). For BED patients, 12% of participants dropped out at session 2, and this risk was clearly higher than for BN until session 13 (when the last dropout was registered). Cox’s regression adjusted by age obtained significant differences in the functions for the three diagnostic subtypes (χ²(Wald) = 20.78, df = 2, p < .001); The comparison of the three diagnosis showed no statistical differences between BN-P and BN-NP (p = .652), while BED differed from BN-P (p < .001; OR = 3.37, 95% CI: 1.99 to 5.71) and BN-NP (p = .027; OR = 2.77, 95% CI: 1.12 to 6.85).