Emotional eating across different eating disorders and the role of body mass, restriction, and binge eating

Abstract Objective Different subtypes of eating disorders (ED) show dysfunctional eating behaviors such as overeating and/or restriction in response to emotions. Yet, systematic comparisons of all major EDs on emotional eating patterns are lacking. Furthermore, emotional eating correlates with body mass index (BMI), which also differs between EDs and thus confounds this comparison. Method Interview‐diagnosed female ED patients (n = 204) with restrictive (AN‐R) or binge‐purge anorexia nervosa (AN‐BP), bulimia nervosa (BN), or binge‐eating disorder (BED) completed a questionnaire assessing “negative emotional eating” (sadness, anger, anxiety) and “happiness eating.” ED groups were compared to BMI‐matched healthy controls (HCs; n = 172 ranging from underweight to obesity) to exclude BMI as a confound. Results Within HCs, higher BMI was associated with higher negative emotional eating and lower happiness eating. AN‐R reported the lowest degree of negative emotional eating relative to other EDs and BMI‐matched HCs, and the highest degree of happiness eating relative to other EDs. The BN and BED groups showed higher negative emotional eating compared to BMI‐matched HCs. Patients with AN‐BP occupied an intermediate position between AN‐R and BN/BED and reported less happiness eating compared to BMI‐matched HCs. Discussion Negative emotional and happiness eating patterns differ across EDs. BMI‐independent emotional eating patterns distinguish ED subgroups and might be related to the occurrence of binge eating versus restriction. Hence, different types of emotional eating can represent fruitful targets for tailored psychotherapeutic interventions. While BN and BED might be treated with similar approaches, AN‐BP and AN‐R would need specific treatment modules.


| Eating disorders and their relation with body mass
Three main eating disorders (ED), namely anorexia nervosa (AN), further separable into a restrictive (AN-R) and a binge-purge (AN-BP) subtype, bulimia nervosa (BN) as well as binge-eating disorder (BED) can be distinguished according to the fifth version of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5;American Psychiatric Association, 2013). Severe underweight due to energy intake restriction, is a defining feature of AN but also a physical comorbidity that likely influences the more psychological diagnostic symptoms (e.g., body image disturbance). Although the maintenance of selfstarvation behaviors in AN is currently poorly understood, theoretical models emphasize the relationship with emotions (Haynos & Fruzzetti, 2011). Similarly, binge eating (i.e., eating an unusual large amount of food and experiencing a loss of control) which characterizes BN and BED, is thought to be maintained by its effects on mood and emotions (e.g., De Young et al., 2013;De Young, Zander, & Anderson, 2014). Possibly as a result of binge eating, BN and particularly BED are associated with elevated body mass index (BMI) (Hudson, Hiripi, Pope Jr, & Kessler, 2007). Thus, the relationship of disordered eating behaviors like binge eating vs. restriction on the one hand and emotions on the other seem to play a key role for symptom presentation among the EDs. Moreover, BMI and EDs seem to be intrinsically linked, raising the question whether some of the disordered eating behaviors in ED patients might be related to their body weight.
In contrast to binge eating/overeating, links of emotional eating with restricted eating and self-starvation are less well understood.
Several authors have argued that anorectic self-starvation may serve to facilitate avoidance of negative emotions (Schmidt & Treasure, 2006). Consistent with this idea, individuals with AN showed increased symptoms after negative mood induction (Wildes, Marcus, Bright, & Dapelo, 2012), reported that restrictive eating helps them managing their negative emotions (Espeset, Gulliksen, Nordbø, Skårderud, & Holte, 2012;Nordbø, Espeset, Gulliksen, Skårderud, & Holte, 2006), and exhibited greater dietary restriction subsequent to days marked by negative affect in daily life . This suggests that patients with AN also resort to altered eating to deal with upsetting emotions, just that they decrease food intake instead of the increase as seen in binge eating. In contrast, positive emotions compared to a neutral state have been shown to be associated with greater food consumption in AN in a laboratory task (Cardi, Esposito, et al., 2015a). Thus, it might be useful to look into how emotions affect food intake (increase, decrease) in EDs, to understand differences in the clinical presentation of various subtypes of EDs.
Thus far, despite the high relevance of potential maintaining factors of self-starvation (in AN) and binge eating (in BN and BED) and the obvious role of emotions in this, research investigating emotional eating across various EDs is surprisingly scarce. Moreover, the contribution of mere over−/underweight to emotional eating of the different ED subgroups has not been sufficiently addressed thus far, despite clear evidence that BMI and emotional eating are correlated (e.g., Varela, Andrés, & Saldaña, 2020) and BMI thus represents a potential confound in any comparison between ED groups and HCs.
Previous research quite consistently shows that patients with BED score higher on negative emotional eating scales compared to weightmatched controls (Escandón-Nagel, Peró, Grau, Soriano, & Feixas, 2018;Pinaquy, Chabrol, Simon, Louvet, & Barbe, 2003;Schulz & Laessle, 2010). Similarly, AN-BP and BN groups mainly exhibited higher emotional eating scores than non-weight-matched HCs (Fioravanti et al., 2014;Ricca et al., 2012;Wardle, 1987), but also no significant differences between AN-BP and non-weight-matched HCs have been found (Baños et al., 2014). However, compared to non-weight-matched HCs, individuals with AN-R exhibited lower (e.g., Danner, Evers, Stok, van Elburg, & de Ridder, 2012), higher (e.g., Ricca et al., 2012), or comparable emotional eating (Baños et al., 2014;Fioravanti et al., 2014). Furthermore, patients with BN scored higher on emotional eating compared to individuals with overweight, (Wardle, 1987). Even subtypes of AN might show specific emotional eating patterns: individuals with AN-BP exhibited higher emotional eating scores than individuals with AN-R (Kiezebrink, Campbell, Mann, & Blundell, 2009;Vervaet, van Heeringen, & Audenaert, 2004), potentially attributable to the occurrence of binge eating in the AN-BP group. Another psychometric study showed that individuals with AN-R reported lower negative emotional eating scores compared to a binge-purge group consisting of AN-BP and BN (Danner et al., 2012). However, Ricca et al. (2012) found no significant differences in emotional eating scores between AN-R, AN-BP and BN. Baños et al. (2014) showed higher emotional eating scores in individuals with obesity compared to AN. Thus, there seem to be clear differences in emotional eating across different EDs and weightrelated groups. Yet, previous research has not studied emotional eating, within one study, in individuals with various ED diagnoses while also considering the potential role of BMI in emotional eating.
An examination of emotional eating (both in response to negative and positive emotions) requires a measurement instrument that taps into both "directions" of eating changes, that is, under-and overeating. We have recently developed the Salzburg Emotional Eating Scale (SEES) that differentiates between increased eating (i.e., a potential correlate for binge eating and overeating) and decreased eating (i.e., a potential correlate for self-starvation) in response to emotions (Meule, Reichenberger, & Blechert, 2018). It further differentiates the negative emotions sadness, anxiety and anger and includes a subscale for positive emotions as previous research showed that specific emotions might differ in their impact on eating behavior (e.g., Braden et al., 2018;Macht, 2008). Healthy individuals indeed revealed characteristic emotional overeating patterns in response to sadness, undereating in response to anxiety and anger and unchanged eating behavior in response to happiness, with negative emotional eating subscales and happiness showing divergent correlations with dimensional ED symptom measures (Meule et al., 2018). In a pre-analysis of 42% of the current sample (see also Supplement A), we examined the SEES scores (among measures of depression, eating psychopathology and emotion regulation) of patients with AN-R and BN, showing that patients with AN-R decreased their food intake in response to negative but increased their food intake in response to positive emotions compared to HCs and BN, who showed the opposite pattern ). Yet, the role of positive emotions (compared to negative emotions) on eating behavior is generally less clear (Evers et al., 2018;Nicholls et al., 2016) with only few studies investigating positive emotional eating, especially among samples with EDs or overweight and obesity.
The current study thus follows up on this research and investigates the shared and unique characteristics of emotional eating across all major EDs (AN-R AN-BP, BN, and BED), based on interview-based diagnostics. In addition, we considered the role of BMI by also investigating weight-related groups (i.e., overweight and obesity) and by selecting BMI-matched healthy control groups for each ED subgroup.
We separately examined the differences between ED groups on the four emotional eating subscales (happiness, sadness, anger and anxiety eating) of the SEES. Based on previous research, we hypothesized higher negative emotional eating scores in the BN and BED groups compared to the AN-R group (Baños et al., 2014;Meule et al., 2019;Wardle, 1987), but because of a lack of previous literature, we had no hypotheses as to whether BN and BED groups differ from each other.
Although previous research showed that in BN, binge eating was associated with anger, whereas in BED it was related to depression , which would suggest differences between both groups on specific SEES subscales, one could also speculate that the typically higher BMI in the BED group may result in higher negative emotional scores. Regarding positive emotional eating, we expected higher scores in AN-R, and lower scores in BN. Regarding BMI, higher negative emotional eating and lower positive emotional eating in healthy (i.e., with regard to EDs) individuals with elevated BMI was expected based on previous research (Geliebter & Aversa, 2003;Meule et al., 2018;Nolan, Halperin, & Geliebter, 2010).
As we aimed at disentangling the effect of ED symptomatology (e.g., binge eating, restriction) from BMI differences, we compared each ED subgroup with BMI-matched control groups.

| Participants
In total, 376 female participants were included in the present study, which was part of a larger project: healthy controls with regard to EDs (HCs; with underweight n = 28, normal weight n = 84, overweight n = 29, and obesity n = 31), patients with restrictive subtype AN (AN-R; n = 64), binge-purge subtype AN (AN-BP; n = 33), BN (n = 71), and BED (n = 36). All groups were partially recruited at the University of Salzburg, Austria (mostly HCs) as well as before and during inpatient treatment at the Schoen Clinic Roseneck, Germany (mostly ED groups). All participants were tested with a structured clinical interview (Saß, Wittchen, Zaudig, & Houben, 2003) and the second version of the Eating Disorder Examination (Hilbert & Tuschen-Caffier, 2016) to confirm or exclude an ED diagnosis. Individuals in the ED groups met the respective DSM-5 criteria, while individuals in the HC groups were classified according to BMI ranges (e.g., for underweight < 18.5 kg/m 2 , normal weight 18.5-24.99 kg/m 2 , overweight 25.00-29.99 kg/m 2 and obesity ≥ 30.00 kg/m 2 ) and were examined using continuous BMI scores. Exclusion criteria for HCs were a current or lifetime ED as revealed with the abovementioned interview.
As a result, 44 individuals from originally 420 individuals who were screened for study participation were excluded because of several reasons (n = 16 HCs because of lifetime or subclinical EDs, n = 24 subclinical ED patients, n = 2 because of missing data on the SEES, n = 2 because of study withdrawal after the screening).

| Salzburg emotional eating scale (SEES)
The SEES (Meule et al., 2018) is a self-report measure that assesses the extent to which individuals perceive their food intake to be altered in response to emotional experiences. The scale consists of 20 items with answers ranging from 1 (= I eat much less than usual) to 5 (= I eat much more than usual) with a score of 3 indicating unchanged food intake. The scale differentiates between four emotional states, namely happiness (e.g., "When I am optimistic, …"), sadness (e.g., "When I am depressed, …"), anger (e.g., "When I am irritated, …) and anxiety (e.g., "When I am worried, …), derived from previous factor analyses (Meule et al., 2018). Internal consistency in the present study was Cronbach's α = .932 for happiness, α = .900 for sadness, α = .924 for anger, and α = .877 for anxiety.

| Procedure
In general, the data of the current study were part of a larger project including psychometric, experimental and naturalistic measures. Data were collected in different waves with slightly different study announcement (see also Supplement B) and different exclusion criteria for the project parts (e.g., left-handed for the labora-

| Data analysis
ED groups were compared regarding SEES scores with univariate analysis of variance as well as nonparametric Kruskal-Wallis tests as either normality of dependent variables within groups or variance homogeneity was violated. As we were particularly interested in T A B L E 1 Descriptive data with regard to body mass index, age, and years of education

| RESULTS
Descriptive data from the groups (i.e., HCs and four ED groups) can be seen in Table 1.    F I G U R E 1 Mean scores of the Salzburg Emotional Eating Scale (SEES) separately for eating disorder groups. AN-R, Anorexia Nervosa, restrictive subtype; AN-BP, Anorexia Nervosa, binge-purge subtype; BN, Bulimia Nervosa; BED, Binge-Eating Disorder. The scale ranges from "eating much less than usual" (=1) to "eating much more than usual" (=5) with 3 marking the middle point of "eating as much as usual" (=3); solid line. Error bars indicate one standard error of the mean [Color figure can be viewed at wileyonlinelibrary.com]

| Analysis of the association of BMI and SEES scores in healthy individuals
In HCs higher BMI correlated negatively with happiness eating, r

| AN-Auxiliary analyses of emotional eating with BMI-matched healthy control group
As BMI confounds the above reported comparisons of ED groups, each of the ED groups was additionally compared to weightmatched HCs. Only individuals with mild and moderate AN were used in the auxiliary analyses. In these subgroups, BMI did not F I G U R E 2 Spearman's rank correlation of the subscales of the Salzburg Emotional Eating Scale with body mass index (BMI) in healthy individuals including individuals with underweight, normal weight, overweight and obesity T A B L E 2 Descriptive body mass index (BMI) data for auxiliary analyses of eating disorders (EDs) with their BMI-matched healthy controls (HCs) differ significantly between HC ANmm_matched and AN-R mm or AN-BP mm (see Table 2).
As can be seen in Figure 3 and

| BN and BED-auxiliary analyses of emotional eating with weight-matched healthy control group
The BN and the BED groups were separately matched with individuals of the HCs group and neither BN and HC BN_matched nor BED and HC BED_matched significantly differed with regard to BMI ( Table 2).
As can be seen in Figure 4 and

| DISCUSSION
The aims of the present study were to characterize all relevant ED subgroups with regard to their emotional eating patterns (a), to examine whether and how BMI is related to various types of emotional eating (b), and to refine emotional eating by controlling for BMI Patients with BN and BED did not differ with regard to negative emotional eating suggesting that the similarities of the two groups with regard to binge eating arise on a shared negative emotional eating background, while the compensation-seen only in BN-does not alter the emotional eating expression. This contrasts with research suggesting that specific emotions might be more important for triggering emotional eating in patients with purging behavior (Rotella et al., 2018) or in patients with BN in contrast to BED . However, in general, the BN and BED groups rather report overeating in response to negative emotions, supporting the emotion regulation model of overeating (or in extreme forms binge eating) in BN and BED stating that eating might be used to cope with and alleviate any kind of negative emotion (e.g., Leehr et al., 2015). reported less frequent overeating in response to happiness compared to other emotions (Masheb & Grilo, 2006). Our results mirror that and suggest that BED patients compensate for their "negative overeating"

| The relationship of BMI and emotional eating in healthy individuals
In line with previous research (e.g., Meule et al., 2018;Nolan et al., 2010), higher BMI was significantly associated with lower happy (over)eating as well as higher sadness, anxiety and anger eating in our HCs. However, results are in contrast to van Strien, Donker, and Ouwens (2016)  Additionally, the present results further support the notion that at least in HCs, a style of "happy overeating" (or the flip side: "unhappy undereating") may be more functional with regard to weight management compared to a style of "unhappy overeating" (Meule et al., 2018). Indeed, previous research showed that negative emotional eating might be a risk factor for longitudinal weight gain (Frayn & Knäuper, 2018;van Strien, Herman, & Verheijden, 2012 (Lewinsohn, Seeley, Moerk, & Striegel-Moore, 2002), EDs in male individuals should not be underestimated (Limbers, Cohen, & Gray, 2018;Murray et al., 2017). The present study excluded male participants because of generally higher emotional (binge) eating in females compared to males (Tanofsky, Wilfley, Spurrell, Welch, & Brownell, 1997) and potentially different symptom profiles in males, which conflicted with our aim for homogenous, reasonably sized ED subgroups. Fifth, the participants (especially AN and BN) who completed questionnaires during their in-patient treatment might have already gained a better understanding of their eating behavior. Future research might circumvent this variability by fixed assessment at the beginning of the in-patient stay. Fifth, HCs did not exhibit current or lifetime EDs, however, participants might still have fulfilled criteria for current other psychological disorders. However, we refrained from recruiting a sample without any psychological disorders to avoid confounding our results by a "general psychopathology" factor.

| Clinical implications and outlook
It has previously been argued that psychotherapeutic treatment should be tailored to the degree of emotional eating: Individuals with high emotional eating might especially profit from interventions improving emotion regulation skills or mindfulness (van Strien, 2018).
Indeed, these approaches have shown promising results in reducing emotional eating meanwhile facilitating weight loss (Frayn & Knäuper, 2018). Our results support this and further identify more