Specialized inpatient treatment of adult anorexia nervosa: effectiveness and clinical significance of changes

We investigated a sample of female inpatients who were hospitalised at the Schoen Klinik Roseneck in Prien, Germany between January 2009 and January 2014. Inclusion criteria were a diagnosis of anorexia nervosa according to ICD-10 (F50.00, F50.01) including intense fear of gaining weight, a distorted body image and a BMI less than or equal to 17.5. Patients were diagnosed by experienced clinicians from the highly specialized eating disorder unit (all with a minimum master’s degree in medicine or psychology) during a standard intake interview. Furthermore, a minimum age of 18 years was mandatory. Exclusion criteria were drug/alcohol/medication abuse, acute suicidal tendencies, psychotic symptoms or a severe life-threatening somatic disorder.

Patients were `regularly’ discharged if they achieved the target weight (BMI > 18 in accordance with eating disorder guidelines [9]). Patients who did not achieve the target weight (mostly those with a very low BMI at admission) were `regularly’ discharged if follow-up treatment was assured (e.g. a re-admission was planned). Patients were `prematurely’ discharged if they had insufficient therapy motivation or for disciplinary reasons.

All participants gave written and signed informed consent. The study was approved by the responsible Medical Ethics Committee of the University of Munich.

All patients received a multimodal inpatient treatment program based on cognitive-behavioural therapy with group and individual psychotherapy. The manualised eating disorder-specific group therapy consisted of nine therapy sessions, each lasting 100 minutes. The main elements were psycho-education, behavioural and functional analysis, acceptance of one’s own body, dealing with emotions and needs, and relapse prevention. The general group psychotherapy took place up to three times a week and each session lasted 90 to 100 minutes. Patients received individual therapy once or twice a week for one hour. Additional therapy elements included social skills training, art therapy, attending classes in the teaching kitchen and exercise therapy. Patients were required to gain 700 g of body weight per week. Co-therapists weighed the patients twice a week in the morning and weight gain was visualised on charts. If patients with anorexia nervosa failed to gain weight, further steps were taken: increase of food intake and monitoring during meal times, administration of high caloric fluids or feeding through a nasal tube.

A standard set of self-rating questionnaires were used to assess eating disorder symptomatology, general psychopathology and depression at admission and discharge.

The Eating Disorder Inventory-2 (EDI-2)[23],[24] was used for the multidimensional assessment of the specific psychopathology of patients with eating disorders. It consists of 11 scales with 91 items that can be answered on a six-point scale from 1 (never) to 6 (always). Cronbach’s Alpha for the EDI-2 Global score for this sample was .96 at admission. Cronbach’s Alphas for the subscales are given in online supplemental materials (see Additional file 1).

The Brief Symptom Inventory (BSI)[25],[26] assesses current general psychological distress of patients throughout the last week on the basis of 53 items belonging to nine subscales. Answers are given on a five-point scale, ranging from 0 (not at all) to 4 (extremely). Three global parameters can be calculated. In the present study, the global severity index (GSI) was used. Cronbach’s Alpha for the GSI for this sample was .96.

The Beck Depression Inventory-2 (BDI-2)[27],[28] is a self-rating instrument to assess the severity of depressive symptoms. Patients can rate each item on a four-point scale from 0 to 3 in terms of its occurrence and its intensity during the last seven days. The cut-off for clinically relevant depressive symptoms is 20. Cronbach’s Alpha for the BDI-2 was .92.

Higher scores in these three instruments indicate greater psychopathology.

Additionally, sociodemographic and clinical characteristics (e.g. BMI ad admission and at discharge, duration of inpatient treatment, duration of the eating disorder, comorbidity) were available from each patient’s clinical records.

In order to examine the effectiveness of inpatient treatment for anorexia nervosa, results were analysed in three different ways:

Statistical group comparisons (repeated measures analyses of variance (rANOVAs) with treatment duration as covariate).

Calculation of effect sizes.

Assessment of treatment outcome on the basis of individual changes according to the criteria of clinical significance [14].

Effect sizes were calculated using the formula (M_pre-M_post)/SD_pre. Interpretation of the effect size was corrected by the effects of an untreated control group, i.e. ES = 0.10, as proposed for single group pre-post study designs by Grawe et al. [29]. Therefore, an ES > 0.30 is considered a small effect, an ES > 0.60 a medium effect and an ES > 0.90 a large effect.

The concept of clinical significance by Jacobson and colleagues [14],[20] is a construct used to evaluate clinically meaningful changes resulting from therapy. It consists of a two-part criterion: To qualify as clinically significantly improved after treatment, a patient has to i) show a reliable change, i.e. statistically significant improvement and ii) cross the cut-off point for a clinically significant change.

As shown in the participant flow chart (Figure 1) between January 2009 and January 2014 748 adult female patients with anorexia nervosa (F50.00/F50.01) were treated as inpatients in the Schoen Klinik Roseneck. Due to administrative regulations, EDI-2 data were not available from n = 313 patients. There were no significant differences between those for whom EDI-2 data were available and those for whom data were unavailable in terms of BMI at admission, age, treatment duration, duration of illness, previous inpatient and outpatient treatments. Furthermore, there were no differences in BMI change during inpatient treatment between these two groups. A total of 435 inpatients with anorexia nervosa were included in the analyses of whom n = 294 regularly completed treatment and n = 141 were premature discharges (see Table 1 for more detail). Patients had a mean age of M = 26.36 (SD = 9.02). BMI at admission was 14.56 kg/m² (SD = 1.74). Individual length of admission varied considerably (6-260 days) and mean duration of inpatient treatment was 91.79 days (SD = 44.26). The sociodemographic and clinical characteristics of the sample are presented in Table 1. For participants who were discharged prematurely, there were no differences in sociodemographic and clinical variables between those with available EDI-2 discharge data and those without. However, differences emerged for the patients who were discharged regularly: those with no discharge data had a lower BMI at admission, more depressive symptoms, more general psychopathology, and more previous inpatient treatments than those with discharge data.

BMI rose on average from 14.56 kg/m² (SD = 1.74) to 17.18 kg/m² (SD = 1.86) (ES = 1.51) during inpatient treatment. Treatment duration significantly contributed to pre-post change in body weight. The total mean weight increase was 7.33 kg (SD = 4.31). The mean weight increase per week was 0.84 kg (SD = 0.56). 52.2% of the patients remained in the anorectic BMI range (≤ 17.5) at discharge. Of 435 patients 108 (24.8%) reached a BMI ≥ 18.5 at discharge. For patients who were discharged regularly, percentages were 38.3% (BMI ≥ 17.5) and 33.1% (BMI ≥ 18.5).

As shown in Table 2, all scales of the EDI-2 indicated statistically significant changes after treatment. The largest effect sizes were found for Drive for Thinness, the EDI-2 Global score and Interoceptive Awareness, whereas the lowest were achieved for Body Dissatisfaction, Interpersonal Distrust and Perfectionism. Treatment duration only had a significant effect on pre-post change in Interoceptive Awareness.

Additional file 1 shows mean and standard deviations of community norms, the patients’ pre-treatment scores and the reliabilities used to calculate cut-offs for a clinically significant change. Furthermore, numbers of patients in the normative range for each scale are given in Additional file 1.

Results show that 27.7% of the patients improved reliably on the Global score of the EDI-2. 35.6% fulfilled the criteria for a clinically significant change. In contrast, 34.5% remained unchanged and 2.2% showed worsened symptoms (see Figure 2). Additional file 2 presents the classification of treatment outcome based on the criteria of Jacobson and Truax [14] for all subscales of the EDI-2.

Change on the GSI of the BSI was statistically significant. Treatment duration had a significant effect on pre-post change in regard to GSI. The effect size was medium (ES = 0.82) (see Table 2). 32.0% of the patients showed reliable changes during inpatient treatment according to the GSI, 34.1% were clinically significantly improved, 30.8% unchanged and 3.0% worse (see Figure 2).

Depressive symptoms significantly improved after treatment with a large effect size of ES = 1.26 (see Table 2). Regarding the BDI-2, the distribution of treatment outcome groups was as follows: 20.0% of patients were reliably improved, 55.3% were clinically significantly improved and 23.0% remained unchanged. 1.7% of the patients deteriorated (see Figure 2).

Since the number of deteriorated patients was very low, these were combined with the unchanged patients. Table 3 shows that clinically significantly improved patients had fewer symptoms of both general and eating disorder psychopathology at admission, and a lower rate of comorbid recurrent severe depression and of posttraumatic stress disorder. They were more motivated for treatment and had a higher rate of moderate depression. Deteriorated/unchanged patients showed a higher rate of comorbid recurrent severe depression and posttraumatic stress disorder.

A higher motivation for change as well as more depressive symptoms turned out to be significant predictors for a clinically significant change whereas body dissatisfaction, impulse regulation, social insecurity and education were negative predictors (see Table 4). These factors explained 35.0% of the variance in scores (Nagelkerke’s R²).

To the best of our knowledge, this study includes the largest sample size of anorexia nervosa patients treated as inpatients. Furthermore, strength of the study is that it also evaluates individual treatment outcomes. This enables a more realistic view of treatment outcomes than comparisons of means and effect sizes do [47]. A further advantage of the study is the naturalistic sample. In comparison to data from randomised controlled trials, data from naturalistic settings may yield more clinically useful findings [48] because they represent the heterogeneous nature of the cases, settings, referral routes, and therapists that characterise everyday clinical practice [49]. These factors enable more valid conclusions to be drawn about treatment improvement.

The strength of the clinical significance concept lies in its applicability to individual cases. Furthermore, evaluation of each individual patient with regard to normative functioning is of great value for clinical practice [30].

There are several limitations. First, the data do not represent all patients with anorexia nervosa admitted in the defined period but rather those for whom baseline and discharge questionnaires were available (about 50%). There were no differences between those with and without missing baseline data in terms of sociodemographic and clinical variables. Furthermore, there were no differences in BMI change during inpatient treatment. However, there were differences between those with and without discharge questionnaires. Since patients who did not complete discharge assessment were more severely ill, results may have overestimated treatment effects. On the other hand, we included outcome data from patients who were prematurely discharged so that our results may after all give a conservative evaluation of treatment outcome. Analyses of only those patients who were discharged regularly might have revealed more positive outcomes. Second, all data except those for the BMI, were based on self-ratings, and no structured clinical interview was conducted. Diagnoses were finally fixed at the end of treatment, giving the expert rater a broad vision of the patient’s behaviour. Third, no follow-up data could be collected to provide data on the long-term outcome of inpatient treatment for anorexia nervosa. Fourth, although one item rated therapy motivation as a predictor of treatment outcome, this has neither been validated against other measures that assess motivation, nor has its reliability been established. One study assessing readiness to change (University of Rhode Island Change Assessment, [50]) in inpatients with anorexia nervosa found that BMI change was not predicted by the stage of change of the patients [51]. A further limitation of the patient’s motivation being rated by the therapist is that the rating data may reflect the motivation of the therapist to some extent, and may influence the motivation of the therapist and consequently treatment quality. Fifth, the method of clinical significance has weaknesses. A statistical limitation of this approach is regression to the mean. Patients with higher scores at admission are those most likely to make huge improvements [52]. Also cut-off point C as a cut-off point between normal and dysfunctional distribution can vary considerably among studies as this criterion depends on the symptom severity in the study sample at baseline. Furthermore, the use of the two criteria to evaluate clinically significant change represents a more stringent measure than most of the previously used criteria. Moreover, one needs to take into account that treatment is, in most cases, not completed when inpatient care concludes. In most cases, inpatient psychotherapy represents a specific phase of a long-term treatment process. Consequently, additional outpatient therapy is required to ensure further improvements of symptoms and a long-term stabilization of achieved effects. There is one caveat, however: patients with anorexia nervosa are often chronically ill and becoming completely symptom-free might be unrealistic, even in the long run.

The paradigm of `clinically relevant treatment effects’ is worthy of some discussion in general. Self-ratings that focus on the assessment of symptoms are certainly important in this context, but it is still only one aspect that must be complemented by other perspectives (e.g. expert ratings). Other factors may equally well indicate clinically relevant treatment effects (e.g. acceptance of disease, improved quality of life, increased knowledge and capability to deal with one’s own illness, provision of a basis for improvement of symptoms in the ensuing outpatient therapy).

Although inpatient treatment for anorexia nervosa is effective, a certain percentage of patients do not benefit from it - at least in certain symptom areas. There are several reasons for the absence of significant or clinically relevant treatment effects. For example, a lack of motivation may reduce response to treatment. Additionally, a lack in quality of the therapy (e.g. a poor therapeutic relationship, overestimation of the patient’s capabilities) may contribute to non-response. Moreover, adverse external events that occur during inpatient treatment such as relationship breakups, job loss, emerging conflicts, may result in non-response or even a deterioration of symptoms in some patients.

Finally, there is the question of how the concept of clinical significance can be integrated in other definitions of treatment outcome. In outcome research there are many different terms used to describe treatment outcome, e.g. response, (partial) remission, recovery and relapse that are often differently defined [53]. In eating disorders research there are also a variety of definitions of outcome success that are used [54]–[57]. Couturier et al. [54] reviewed various conceptualizations of remission in adolescent anorexia nervosa and reached the conclusion that remission is best defined as the attainment of a certain weight and/or the attainment of normal EDE scores (within 1 or 2 SD of the normal range, depending on the distribution). Bardone-Cone et al. [57] defined recovery as no longer meeting diagnostic criteria for anorexia nervosa, showing a BMI ≥ 18.5, no binging or purging, and EDE scores within 1 SD of community norm for each subscale. All these definitions compare individuals’ outcomes with a norm at a predefined assessment time. However, they do not consider changes resulting from treatment. As Frank et al. [53] stated, remission and recovery are an assessment of outcome at a single point and are not entirely dependent on treatment. However, response is a change due to treatment that requires baseline and post-treatment scores. In this sense, the first criterion of the concept of clinical significance (reliable change) assesses response to treatment. A clinically significant change requires both response to treatment as well as achieving a functional range at endpoint assessment. A clinical significant change may therefore be seen as remission, but considers response to treatment, too. Therefore, the criteria of clinical significance may be a more appropriate definition for evaluating treatment outcome.

In order to optimize treatment, future studies should explore the complex interplay of factors that may prevent patients with anorexia nervosa from non-responding to inpatient treatment. Then, treatment strategies for those should be developed. For those who improved during inpatient treatment, the challenge remains in maintaining therapy success following discharge. More attention should be paid to relapse prevention strategies, of which technology-based relapse prevention may be particularly useful [58],[59]. Additional research should also address clarifying the impact of depressive symptoms or of depressive comorbidity on outcome for anorexia nervosa patients. Furthermore, future research should also investigate if patients’ therapy motivation or stage of change as assessed by validated disorder-specific measures [60]–[62] are a predictor for a clinically significant change.