Introduction

Clinically, it is widely acknowledged that having an ED not only seriously affects the individual psychologically and physiologically, but also has a severe impact on functioning in other life domains, such as work, family and social life. Thus, most people with an ED are apprehensive of disclosing the ED to others because of fears of negative reactions from their surroundings. The disorder itself has prevailing characteristics of secrecy and concealment since the core symptoms of bulimia nervosa (BN), such as binging and purging, are most often hidden behaviors [38, 44, 47]. Avoiding meals and/or eating with others is a common strategy for concealing the eating disorder [47], and since meals play an important role in the social culture, social avoidance and isolation are often consequences of the ED. Combined with negative thoughts and feelings about shape and body, these characteristics typically interfere with close, intimate and longer-term relationships. Conversely, better interpersonal functioning and fewer family problems, which are closely linked to the concept of social functioning, have been associated with positive treatment outcomes [46].

The importance of functional impairment related to eating disorders has not, however, led to an inclusion of functional impairment into the DSM criteria for ED, as it is for obsessive compulsive disorder (OCD) and many other mental disorders [1, 2]. Most ED outcome research also focuses solely on behavioral symptoms as outcome measures. In a review of 71 published studies, Williams et al. [48], e.g., found only one study that included quality of life (combined with ED symptoms) as their primary outcome measure. The remaining 70 studies measured behavioral symptoms only, in spite of the substantial amount of literature indicating that quality of life among individuals with ED is poorer than for the healthy population.

Recently, however, the statistical association between ED severity and functional impairment has been established in a number of empirical studies. In general, individuals with EDs have been found to have poor quality of life compared to individuals without EDs [4, 14, 26], and functional impairment seems to increase with illness severity [5, 14]. It has been shown that patients with binge eating disorder (BED) or BN have more days with functional impairment than individuals without EDs [27] and that binging and purging in particular seem associated with decreased functionality [14, 43]. Other studies have highlighted the negative impact of obesity on functional impairment, especially when the individual also has BED [33, 43].

One appropriate approach to a better understanding of the functional impairment associated with EDs is to clarify the complex interplay of specific factors contributing to the extent of functional impairment. However, while it is likely that the relationship between ED severity and functional impairment is moderated and mediated by a number of other pre-treatment characteristics associated with ED, we are not aware of previous studies that have investigated the impact of such moderators and mediators. The goals of the present study were to investigate two conditions (duration of illness and body weight) that may predict the strength of association between severity and impairment (moderator variables), and to examine psychological distress as a possible intervening variable (mediator) that explains how severity has its effect on impairment.

Duration of illness

A number of studies have demonstrated that a short duration of illness is related to a more favorable outcome of treatment for EDs [17, 32, 35, 40, 46]. Thus, Padierna et al. [32] found that participants who had suffered from an ED for more than 5 years were less likely to profit from treatment than those with a shorter duration of illness [25]. It is conceivable that while individuals with relatively mild ED symptoms may gradually learn to live with the symptoms in a way that have a limited impact on their daily life, individuals whose ED symptoms remain severe over time may gradually enter a life course characterized by a more marked withdrawal from social activities and decreased functioning. We therefore hypothesized that the association between the severity of ED-specific symptomatology and functional impairment, defined as impairment in social, occupational, and/or other important areas of functioning, would be stronger for patients with a longer duration of illness (Hypothesis 1a).

Body weight

The absence of regular compensatory weight-control behavior often, but not always, leads to overweight among patients with BED, which can put an extra strain on these patients and leave them more distressed [3, 13]. Spitzer and colleagues [39] found a higher degree of impairment in work and social functioning of people with BED compared to non-binging obese people; the association was due to distress about weight and eating. In a more recent study, obese individuals with BED consistently reported the poorest quality of life, compared to non-obese with BED, non-obese without BED, and obese without BED [33]. Since patients with EDs are generally characterized by an elevated focus on body weight it is likely that overweight patients with BN or EDNOS will experience distress related to weight similar to that found in patients with BED. Within the theoretical framework of Pachankis [31], an ED can be viewed as a ‘concealable stigma’, a part of the individual´s identity, that is discrediting in a social context, and that the individual tries to conceal. In his cognitive-affective-behavioral model on the effects of a concealable stigma [31], Pachankis, inspired by Goffman’s work on stigmatization [22], suggests that people concealing a stigma (e.g., a mental disorder or sexual orientation) may face unique challenges compared to people with visible stigmas (e.g., physical disability or overweight). Pachankis points out that concealable stigma may have considerable psychological and social implications for the stigmatized individual, due to the effort used to conceal the stigma. Within Pachankis’ theoretical framework, overweight related to an ED is associated with both a concealable stigma (the ED), and a visible stigma (overweight) [31]. The presence of both types of stigma might imply that increased severity of ED symptoms would have a particularly detrimental impact on overweight ED patients since it could be associated with an increased strain from both the concealable and the visible stigma. We therefore hypothesized that the association between ED severity and functional impairment would be stronger for patients with a higher body mass index than for patients with a lower body mass index (Hypothesis 1b).

Psychological distress

Psychological distress and/or co-morbid psychiatric disorders sucζh as depression and anxiety are common for patients suffering from an ED [11, 19]. The distress is likely to be caused by the constant struggle with restriction of food, and dietary restraint, the negative emotions related to binges, and (for patients with BN and many patients with EDNOS) compensatory behavior such as purging [29]. According to Pachankis’ model [31], the fact that the ED is kept hidden may lead to a further increase in psychological distress as evidenced in previous studies demonstrating higher rates of psychopathology and psychological distress for individuals with concealable stigmas compared to individuals with visible or no stigmas [8, 20]. The combined psychologically distressing effect of the core symptoms per se in combination with the strain from having to hide the symptoms is likely to affect social, occupational, and academic functioning. We therefore hypothesized that the relationship between ED severity and functional impairment would be partly mediated by psychological distress (Hypothesis 2; illustrated in Fig. 1).

Fig. 1
figure 1

A priori mediator model

Full size image

Aims of the study

We hypothesized that ED severity would predict functional impairment in a clinical sample of treatment-seeking individuals with a diagnosis of BN, BED, or EDNOS. We hypothesized a moderating effect of duration of illness (Hypothesis 1a: Effects of ED severity were predicted to be stronger for those who had a longer course of illness) and also of body weight (assessed via body mass index; Hypothesis 1b: Effects of ED severity were predicted to be stronger for more overweight patients). We also examined psychological distress as a mediator of the association between ED severity and functional impairment (Hypothesis 2; see Fig. 1). Because Hypotheses 1a and 2 were supported, we also conducted a post hoc moderated mediation analysis [6] in which we examined the moderating effect of illness duration on each of paths a and b in the mediator model in Fig. 1.

Methods

Participants and procedures

The participants were recruited between August 2012 and February 2014 for a randomized, clinical trial examining the effect of client feedback on treatment attendance and outcome. The trial is registered in http://www.clinicaltrials.gov (NCT01693237). It is approved by the regional ethics committee for the Capital Region of Denmark (journal number H-3-2011-151) and by the Danish Data Protection Agency (journal number 2007-58-0015). A detailed description of the trial design is published elsewhere [12].

The participants were 159 treatment-seeking adults referred to treatment for BN, BED or EDNOS at the outpatient clinic for EDs at Stolpegaard Psychotherapy Centre, Mental Health Services, Capital Region of Denmark. The BN and EDNOS diagnoses were set according to the clinical DSM-IV manual; BED was diagnosed according to the research criteria defined in DSM-IV [1]. Inclusion and exclusion criteria followed the ones set up for the trial. Inclusion criteria were: aged 18 years or older, body mass index (BMI, kg/m2) ≥20,Footnote 1 diagnosed with BN, BED or EDNOS as the primary diagnosis, and informed consent. Exclusion criteria were acute suicidal risk, psychosis, severe depression, severe or non-regulated physical co-morbidity, pregnancy, abuse of alcohol or medicine, use of cannabis more than once monthly, concomitant psychotherapeutic or psychiatric treatment outside PCS, inability to understand Danish, previous participation in the current trial, patients considered unable to attend treatment sessions as planned. The majority of the participants were female (98.1 %), single (64.8 %) with at least 10 years of schooling (78.6 %) and had struggled with the ED for 5 years or more (65.4 %). Nearly half were currently undergoing education (44.7 %). The mean age was 28.0 (SD = 9.7). Seventy-three participants were diagnosed with BN, 57 with EDNOS, and 29 were diagnosed with BED. The mean BMI was 26.3 (SD = 7.9). Average BMI for patients diagnosed with BED was 38.8 (SD 10.0), for BN 23.6 (SD 3.0) and for patients with EDNOS 23.8 (SD 5.0).

The data were collected at two different pre-treatment time-points: (a) routine assessment interview: ED severity and psychological distress, BMI and duration of illness; (b) trial assessment interview (2–4 weeks after routine assessment): functional impairment.

Measures

The Eating Disorder Examination Questionnaire (EDE-Q)

The patients rated ED severity using the authorized Danish version of the EDE-Q. The 38-item EDE-Q is a self-report measure based on the interviewer-administered version, which is a widely used interview in ED research and is considered ‘the gold standard’ in diagnosing EDs [10, 18]. Psychometric analyses of the EDE-Q have shown an acceptable test–retest reliability (of subscale scores) and internal consistency [7]. The EDE-Q had a good internal consistency in our sample, Cronbach’s alpha = 0.90.

The Symptom Checklist (SCL-90-R)

Psychological distress was measured with the authorized Danish version of the SCL-90-R [15, 16]. The SCL-90-R is a multi-dimensional patient-reported questionnaire for measuring psychological distress [15, 16, 30]. We used the Global Severity Index (GSI), which is the global score covering all 90 items. Internal validity has been tested and found acceptable [30]. The SCL-90-R had good internal consistency in our sample, Cronbach’s alpha = 0.96.

Sheehan Disability Scale (SDS)

Functional impairment was measured with the Danish version of the SDS, which is a widely used self-report measure of condition-specific disability. The SDS assesses functional impairment in three domains: Work/school, social life/leisure activities, and family life/home responsibilities [36]. We used the patient-rated version, i.e., the researcher asked the questions in the trial assessment interview, and the patient was asked to put a mark on a 1–10 scale. The time frame was the previous 2 weeks. A global score was calculated by adding the scores from the three domains, and thus ranged from 0 to 30; the higher the score, the more impaired functioning. Studies of the psychometric properties of the SDS have demonstrated an acceptable reliability, sensitivity to change and construct validity [37]. Cronbach’s alpha in the present sample was 0.62.

Moderator variables

Duration of illness was addressed in the routine assessment, i.e., the therapist interviewed the patient about the emergence of the current ED as part of the clinical interview. The variable was measured categorically with four possible categories: <0.5, 0.5–2, 3–5 and >5 years (M = 3.55, SD = 0.74). Because the categories were ordered, we treated the variable as continuous. In order to create a meaningful zero point, we centered the variable by subtracting the mean from each observed score [9, 21].

BMI was a continuous variable; BMI = kg/m2. In order to create a meaningful zero point, we also centered this variable by subtracting the mean from each observed score [9, 21].

Statistical analyses

Moderator tests were conducted via simultaneous regression analyses as recommended by Cohen et al. [9]. Mediator tests were conducted via a series of regression models, following the recommendation of Baron and Kenny [6]. Data were analyzed using the Statistical Package for the Social Sciences (SPSS 19.0) [23] and R [34].

Results

Hypotheses 1a and 1b: moderator analyses

Table 1 summarizes the regression coefficients for the predictor (ED severity), moderators (duration of illness and BMI) and their product terms for Hypothesis 1a and 1b.

Table 1 Unstandardized regression coefficients for the predictor, moderators and product terms
Full size table

Duration of illness

Duration of illness was found to moderate the relation between ED severity and functional impairment. The statistical interaction between duration and severity is tested by the regression coefficient for the product term, which was B = −1.446, 95 % CI (−2.484, −0.409), p = .000 (see Table 1). This indicates a significant moderator effect of duration, but not in the predicted direction. Figure 2 depicts the effect of ED severity on functional impairment at three representative levels of duration (we used the sample mean for duration as the medium duration, and 1 SD below and above this mean as the low and high duration values, respectively). The effect of ED severity (represented by the slope of the regression line) grows weaker as duration increases, such that severity is a strong predictor of functioning for patients with a relatively recent diagnosis of ED, but a rather weak predictor for those with longer durations of illness in our sample.

Fig. 2
figure 2

Duration of illness as moderator of the relationship between eating disorder severity and functional impairment. Asterisk indicates that the partial regression coefficient (slope) differs from zero (p < 0.05)

Full size image

Body weight

Body mass index did not significantly predict functional impairment and did not significantly moderate the relationship between ED severity and functional impairment. The unstandardized regression coefficient for the interaction term was B = −0.009, 95 % CIs (−0.045, 0.027), p = .622. Thus, there was no support for our hypothesis that the association between severity and functional impairment would be stronger for overweight patients.

Hypothesis 2: mediator analysis

We first established that ED severity was related to functional impairment (path c, Fig. 1); the unstandardized regression coefficient was significant, B = 1.715, 95 % CIs (0.951, 2.479), p = .000.

Next, psychological distress was regressed on to ED severity (path a). The unstandardized regression coefficient was significant, B = 0.221, 95 % CIs (0.149, 0.293), p = .000.

To test whether psychological distress was related to functional impairment (path b), functional impairment was regressed simultaneously on psychological distress and ED severity. The unstandardized regression coefficient associated with the relation between distress and functional impairment (when controlling for ED severity) was significant, B = 6.252, 95 % CIs (4.812, 7.692), p = .000.

The mediated effect (product of paths a and b in Fig. 1) divided by its standard error yielded a z score of 4.99 using the Sobel test, confirming that the indirect effect differs significantly from zero [6]. Thus, the hypothesis that psychological distress partly mediates the relation between ED severity and functional impairment was supported.

Exploratory analyses: moderated mediation

Because duration of illness moderates the overall effect of severity on impairment, and because this overall effect is mediated by psychological distress, it is of interest to ask how the two paths (see Fig. 1) representing this indirect effect may be moderated by duration. Table 2 summarizes the regression coefficients for the predictors (ED severity and psychological distress), moderators (duration of illness) and product terms for the exploratory analyses of moderated mediation.

Table 2 Unstandardized regression coefficients for moderator tests on (path a) the relationship between eating disorder severity and psychological distress, and (path b) the relationship between psychological distress and functional impairment
Full size table

Duration of illness did not significantly moderate the relationship between ED severity and psychological distress (path a), B = −0.071, 95 % CI (−0.175, 0.033), p = .174.

The moderator analyses on path b showed that duration of illness significantly moderated the relationship between psychological distress and functional impairment, when controlling for ED severity, B = −2.669, 95 % CI (−4.346, −0.993), p = .002. As illustrated in Fig. 3, the relationship was stronger for patients with a shorter duration of the ED than for patients with a longer duration.

Fig. 3
figure 3

Duration of illness as moderator of the relationship between psychological distress and functional impairment. Asterisk indicates that the partial regression coefficient (slope) differs from zero (p < 0.05)

Full size image

Discussion

There is only limited research on the relationship between ED severity and functional impairment. In this study, we analyzed the relationship between ED severity and functional impairment by examining the potentially moderating effects of duration of illness and BMI, and the potentially mediating effect of psychological distress on this relationship.

ED severity more strongly predicted functional impairment for patients who had the ED for a shorter time versus those patients who had the disorder for a longer time; we expected the opposite relationship (Hypothesis 1a). We propose two interpretations of these results. First, struggling with an ED is a radical interference in an individual’s daily life and may be experienced more impairing in the early stages. After some time, the patients could be inured to their ED and feel less impaired than in the earlier course of illness. Second, patients with a short duration of illness have not yet developed coping strategies such as routines for meals that minimize likelihood of exposure or negotiation of social relationships with others who are less curious or non-confrontative. Patients with a longer duration of the ED have had time to develop these strategies, which could decrease the association between ED severity and functioning.

Contrary to our hypothesis, having an ED and being overweight (i.e., having both a concealable and a visible stigma) did not alter the direction or strength of the relationship between ED severity and functional impairment. The results imply that the association between ED severity and functioning is constant across body weight, i.e., the same for patients with low and high body weight. This means that an elevated level of ED symptoms leads to decreased functioning just as much for patients with a low vs high body weight.

We found the relation between ED severity and functional impairment to be partly mediated by psychological distress. In other words, patients with high levels of ED severity reported high functional impairment due to a high level of psychological distress. This result supports two hypotheses. First, severe ED is associated with higher levels of psychological distress, which may be explained as a result of both the strain associated with restriction of food and dietary restraint, the negative emotions related to binging and purging, and (as suggested in Pachankis’ theory of concealable stigmas [31]) the efforts involved in hiding the ED (path a in Fig. 1). Second, the psychological distress related to high levels of symptoms may lead to increased functional impairment (path b in Fig. 1). Previous research in the effect of dealing with a concealable stigma also supports this interpretation as the researchers found higher rates of psychopathology and psychological distress for individuals with concealable stigmas compared to individuals with visible or no stigmas [8, 20]. The longitudinal design of our study strengthens the basis for causal inference because measurement of ED severity preceded that of functional impairment in time [21]. Because ED severity and psychological distress were measured at the same time point, it cannot, however, be ruled out that ED severity increases as patients experience more psychological symptoms, i.e., that the direction of causation between psychological symptoms and ED severity is the opposite than proposed above.

Since psychological distress partly mediated the association between ED severity and functional impairment (Hypothesis 2), we revisited the moderating effect of ED duration on this relation, examining the moderating effect of duration on each of the components of the indirect path in Fig. 1. The relationship between ED severity and psychological distress (path a) was not moderated by duration of illness, which indicates that there is no support for the idea that the link between ED severity and psychological distress is attenuated with time. Duration did significantly moderate path b; however, such that psychological distress more strongly predicted functional impairment for patients with a short duration of illness compared to patients with a longer duration of illness. The results imply that path a is constant across duration, i.e., that ED symptoms lead to distress just as much for patients with a long duration of ED as for patients with a shorter duration of ED. This seems to indicate that patients with a longer history of ED pathology did not develop coping mechanisms, which would lead to an attenuation of the link between ED severity and distress. However, patients with a long duration of illness show less of a tendency for distress to lead to impairment (path b). They cope better with work/family/social roles and are better able to maintain functioning in spite of distress, relative to patients with a shorter history of ED. This indicates that functioning is likely to be a good indicator of distress (and thus of possibly severe ED symptoms) for ED patients with a briefer duration of illness, whereas longer term ED sufferers may be experiencing higher levels of distress even while less functionally impaired.

The participants in our study had a mean score of 17.9 on the measure of functional impairment (Sheehan Disability Scale). This is only marginally lower than the mean scores of treatment-seeking individuals with obsessive–compulsive disorder (OCD) [24, 41, 42], (M = 19.3 and 19.4, respectively). As OCD is considered a severely disabling mental disorder, these results indicate that an ED can be experienced as overly functionally impairing.

Clinical and research implications

Overall, the results from the current study show that treatment-seeking patients with an ED are functionally impaired in a degree that resembles other socially disabling mental disorders such as OCD. It is therefore important to acknowledge social withdrawal, work or family problems as part of early symptom recognition of an ED. Some authors have recently proposed a staging model to map the course of ED from high risk to severe enduring illness, which would be helpful to match interventions to the specific stages [45]. We also recommend prioritizing treatment efforts focused on specifically targeting the social aspects of the ED (e.g., the effect of the ED on children, spouses and professional relationships) instead of solely treating ED symptoms. The results further suggest a difference between patients with a short duration of ED and patients with a long duration of illness as both ED severity and psychological distress more strongly predicted functional impairment for patients with a short duration than for patients with a long duration of ED. We therefore emphasize the importance of individualizing treatment efforts, which would make it possible to plan different treatment courses for patients with short versus long duration of illness. For patients with short duration of illness, ED symptom severity as well as psychological distress are closely tied to functioning, so improvement in these areas is likely to have a substantial impact on work, social, and family functioning, something that would not necessarily be the case for patients with a longer duration of ED.

As for ED research, more studies are needed to look at the complex relationship between ED severity and functioning, for patients with a longer duration of ED in particular. Furthermore, we suggest to broaden the concept and definition of ED outcome by including measures of functioning; this could contribute to a better understanding and treatment of EDs.

Limitations

There are some limitations to this study. In the trial design, only few trial-specific exclusion criteria were added, in order to keep the sample as naturalistic as possible. The participants were treatment-seeking, Danish females, and those with anorexia nervosa were not included; this limits the generalizability of the findings. We used the EDE-Q global score to measure severity, which does not necessarily reflect binge eating frequency, which is a core symptom in BN and BED, and which contributes to the severity of an ED. The low reliability of the SDS in this sample (α = 0.618) reduces correlations with predictors, lowering the overall R 2 value and the power of the tests [21]. The fact that both mediator and moderator findings were significant despite this handicap attests to the importance of studying functional impairment in ED populations, and development of an improved measure of functioning is an important area for further research. Finally, our mediation results should be interpreted with caution; severity (predictor in these models) and psychological distress were measured concurrently at the initial assessment point, functional impairment were assessed later. To make a stronger case for the direction of causation between these two variables and to improve estimates of the mediated paths, future studies should ideally assess the putative mediator variable at an earlier time point than the putative outcome [28].

Conclusions

This study established a link between ED severity, psychological distress and functional impairment showing that impaired role functioning appears to be especially a risk for newer ED patients. More work is needed to understand why and how more experienced ED sufferers, although they still find symptoms distressing, experience less functional impairment as a function of the severity of the ED. Simultaneously, the study draws attention to the potentially severe social consequences of an ED; a topic that has not been much studied.