Disgust and Self-Disgust in Eating Disorders: A Systematic Review and Meta-Analysis

Disgust and self-disgust are aversive emotions which are often encountered in people with eating disorders. We conducted a systematic review and meta-analysis of disgust and self-disgust in people with eating disorders using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The systematic review of the literature revealed 52 original research papers. There was substantial heterogeneity regarding the research question and outcomes. However, we found 5 articles on disgust elicited by food images, 10 studies on generic disgust sensitivity, and 4 studies on self-disgust, and we proceeded to a meta-analytic approach on these studies. We found that women with eating disorders have significantly higher momentary disgust feelings in response to food images (1.32; 95% CI 1.05, 1.59), higher generic disgust sensitivity (0.49; 95% CI 0.24, 0.71), and higher self-disgust (1.90; 95% CI 1.51, 2.29) compared with healthy controls. These findings indicate the potential clinical relevance of disgust and self-disgust in the treatment of eating disorders.


Introduction
Eating disorders (EDs) are complex psychiatric disorders characterized by preoccupation with weight, shape, and food [1]. The current treatment outcome for EDs is still unsatisfactory, and these disorders may persist for many years. Therefore, there is an interest in elucidating the mechanisms that underpin eating psychopathology so that more targeted and personalised treatments can be developed [2].
Anxiety, fear, and disgust have been thought to be involved in the problematic avoidance and/or compensatory/safety behaviours often observed in EDs [3][4][5]. Interview and self-report measures [6] have been developed to define the form of fear in eating disorders, and exposure-based treatments [7][8][9] have been designed to target these fears. However, disgust has been less extensively studied, although it is strongly linked to avoiding oral ingestion, according to Rozin and Fallon [10]. For instance, the ingestion of foods containing a potential high-pathogen load, such as rotten foods, can be a critical stimulus to elicit disgust and aversion through the evaluation of the stimulus as being contaminated. The association between disgust and a real or perceived threat from food ingestion may cause the development of unhealthy eating behaviours [11]. Furthermore, individuals may experience overwhelming feelings of disgust in response to highly palatable or calorically dense foods due to overestimating the likelihood of experiencing future weight stigma.

Study Selection and Data Extraction
The results of the literature search were first imported into Endnote, and duplicate records were removed using the web tool for systematic reviews called Rayyan [17]. Two independent reviewers (S.B. and H.M.) screened the titles and abstracts of eligible studies using Rayyan software. Disagreements over the suitability of studies for inclusion were resolved by consensus and further by discussing with other research team members (J.T., H.H., J.L.K., and L.M.A.).
Candidate studies were reviewed as full texts and relevant data were extracted using an online Microsoft Excel spreadsheet. Extracted data included the following: title, abstract, authors, contact details, country of origin study and participant characteristics, disgust measure used (questionnaires and/or experimental tasks), and disgust outcomes with means and standard deviations. For additional data that were not reported, the first reviewer (S.B.) contacted the authors.

Synthesis of Results
For the systematic review, studies that met the broader criteria of investigating disgust and self-disgust in individuals with EDs were reviewed, and the findings of these studies were qualitatively reported.
For the meta-analysis, the outcome variables of interest were disgust elicited by food images, generic disgust sensitivity, and self-disgust. For the synthesis of results, Hedges' g effect sizes were calculated for primary outcomes from the means and standard deviations. Hedges' g results in a more precise estimate of the effect size than Cohen's d [18], especially for studies with small sample sizes. Effect size values of 0.20 were considered "small," 0.50 were considered "moderate," and 0.80 were considered "large" [19].

Meta-Analyses
Five separate meta-analyses were conducted for between-group effect sizes of: (1) disgust elicited by food images in EDs compared to controls; (2) generic disgust sensitivity in EDs compared to controls; (3) generic disgust sensitivity in AN compared to controls; (4) generic disgust sensitivity in BN compared to controls; (5) self-disgust in EDs compared to controls. Heterogeneity was suspected in all data, and so a random effects meta-analysis with the DerSimonian and Laird method [20] was used to pool Hedges' g values.

Sensitivity Analyses
The between-study heterogeneity was assessed using Higgins I 2 that was considered to be high when I 2 > 75% [19]. Publication bias was assessed using Egger's test for small study effects [21] with funnel plots. The sensitivity analyses were then conducted using the trim and fill method [22] to identify and correct funnel plot asymmetry. This method also determines if the removal of smaller studies would reduce publication bias based on re-estimated g values. Effect size calculations and analyses were computed in Stata 17 software [23].

Quality Assessment
Two reviewers (S.B. and H.M.) independently evaluated the risk of bias for candidate studies using the Joanna Briggs Institute Critical Appraisal Tool [24]. This tool comprises several checklists for different research designs, except for experimental studies, and determines the extent to which a study has an adequate methodological quality in four categories: Yes, No, Unclear, or Not Applicable. Since one eligible study had an experimental design in this systematic review and JBI could not offer a tool for experimental studies, we made a minor modification to the checklist developed for quasi-experimental studies by adding an item that scrutinized whether the study used true randomization, taken from the checklist for Randomized Control Trial studies [24].

Study Selection
Three database searches yielded 524 candidate papers, and three studies were identified through hand-searching. Following the deletion of duplicate articles, 290 articles were screened and fully read. Of these, 234 studies not meeting the eligibility criteria were dismissed. Figure 1 is a PRISMA flow diagram depicting the search for eligible studies. A total of 52 studies were eligible for inclusion in the systematic review, and 19 of these studies fitted the inclusion criteria for the meta-analyses (disgust elicited by food images n = 5, generic disgust sensitivity n = 10, self-disgust n = 4). All the included studies were considered to have a low risk of bias except for one study [25] that we excluded based on frequent responses to No and Unclear, showing a high risk of bias. The quality assessment of the candidate studies is shown in Supplementary Tables S4-S8.

Systematic Review
The findings from the 52 individual studies included are detailed in Tables 1-5. The majority of studies investigated AN, BN and mixed eating disorder groups in adult females.

Methods to Investigate Disgust and Self-Disgust
The eligible studies for the systematic review used various research methods which were categorized into five main methodological approaches: questionnaires or diaries to measure disgust/self-disgust, stimuli to trigger disgust, experimental tasks, brain imaging, neurophysiology, and qualitative approaches presented in Tables 1-5.

Questionnaires
Trait disgust was measured using a range of questionnaires. Four studies applied self-report scales, such as the Differential Affect Scale [26] and the Differential Emotion Scale [27]. Two studies adapted self-report questionnaires to measure body disgust [28] in relation to pregnancy and sexuality [29]. Nine studies utilized inventories developed specifically for disgust. These inventories include the Questionnaire for Assessment of Disgust Sensitivity [30], the Questionnaire for Assessment of Disgust Proneness [31], the Disgust Propensity and Sensitivity Scale [32] or its revised form [33], the Disgust Sensitivity Questionnaire [34,35], the Disgust Scale [36,37] or its revised form [38], or the Disgust Questionnaire [39,40]. Two studies [41,42] measured a cluster of emotions of which disgust was one emotion, but they did not measure disgust specifically. The studies on self-disgust (n = 4) used the Multi-dimensional Self-Disgust Scale [43], the Questionnaire for Assessment of Self-Disgust [44], and the Self-Disgust Scale [33,38]. See Table 1.
Other experimental tasks involving facial expressions consisted of "The Visual Probe Detection Task" to investigate attentional bias [72], "The Emotional Go/No-Go Task" to yield inhibition capability [73], "The Voluntary Facial Expression Task" [74], "Facial Emotion Discrimination Task" [75], or "The Pose and Imitated Expression Task" [76]. In addition, in the study of Gagnon et al. [56], "The Temporal Bisection Task" was used to  measure subjective time perception whilst showing disgusting food pictures rather than  facial expressions. See Table 3.

Findings of Qualitative Studies
The results suggested that cognitive processes (autobiographical memories or beliefs about losing control over food; [81]), internal physical sensations, negative interpersonal experiences (i.e., judgement and criticism of others), or life events (i.e., bullying) might trigger disgust towards food or the body in people with eating disorders [54,83]. Moreover, avoidance seems to be the most relevant coping strategy [83]. For details, see Table 5.

Meta-Analysis
From the studies described in the tables above, we were able to extract data suitable for a meta-analysis in three of the domains (food images, generic disgust, and self-directed disgust.

Disgust Elicited by Food Images
A total of five studies with a case-control design reported disgust ratings towards food images salient for people with EDs. The five studies that were included in the meta-analysis with a total of 284 female participants, of which 139 had an ED with a mean age ranging from 16.5 to 31.4. In some studies, the diagnostic categories were separated: AN versus BN versus ED versus HC [52]; AN-R versus HC [49]; BN versus HC [50]; ED versus HC [46]; AN versus AAN versus BN versus BED [45]. State disgust was measured with a visual analogue scale (VAS). Figure 2 shows the meta-analysis of five studies that compared disgust elicited by food images between individuals with any ED diagnosis and HC (g = 1.32; 95% CI 1.05, 1.59; p < 0.001).

Generic Disgust Sensitivity
A total of 10 case-control studies [27,30,[32][33][34][35][36][37][38]77] investigated overall disgust sensitivity in 1329 female participants, of which 767 had an ED with a mean age ranging from 21.9 to 29.7. Disgust sensitivity data are reported for the combined ED group, and two more meta-analyses were conducted on sub-samples (AN and BN groups). Eight studies [27,[32][33][34][35][36][37][38] had relevant data for AN (AN n = 526; HC n = 454). One of these studies provided subgroup AN value (AN-R and AN-BP) separately, and therefore statistical data (means, standard deviations, and sample sizes) were combined for the analysis. Additionally, six studies that presented overall disgust sensitivity in people with BN relative to controls (BN n = 211; HC n = 385) were eligible meta-analysis [27,30,[32][33][34]77]. Figure 3 shows the meta-analysis of 10 studies which compared generic disgust sensitivity level between individuals with any ED diagnosis and HC (g = 0.49; 95% CI 0.27, 0.71; p < 0.001).   [34] showed that the ED group had a higher overall disgust sensitivity and disgust sensitivity to food, animal, and body products than HC.  (44) -Both ED groups (remitted and clinical) reported a higher level of disgust towards the human body and body products and foodstuffs of animal origin than the other three domains (invertebrate animals, gastro-enteric, sexual practices). retrospectively.

N/A
-Factor 1, indicating feeling of guilty, disgusted, and angry, was at the highest level compared to the other three factors in the period between binge and purge. Following the purge, this level decreased by reaching the same level reported in phase 1.              -Greater activation in the left medial orbito-frontal and anterior cingulate cortices and less activation in the lateral prefrontal cortex, inferior parietal lobule, and cerebellum in response to food images among patients with EDs compared to HC. N/A * -BN patients had less activation in the lateral and apical prefrontal cortex in response to food images than HC.
N/A *    The sub-group meta-analyses found that the generic disgust sensitivity score was significantly different in individuals with AN versus HC (g = 0.60; 95% CI 0.34, 0.87; p < 0.001) and those with BN versus HC (g = 0.50; 95% CI 0.20, 0.80; p = 0.004). See Figure 4 for AN meta-analysis and Figure 5 for BN meta-analysis.

Self-Disgust
A total of four comparable case-control studies investigated overall self-disgust in 1196 female participants, of which 579 had an ED with a mean age ranging from 25.36 to 32.16. Two studies presented sub-type (AN and BN; [33]) or sub-scale (personal and behavioural self-disgust; [44]) values separately, and therefore, statistical data (means, standard deviations, and sample sizes) were combined for the analysis. Figure 6 shows the meta-analysis of four studies comparing self-disgust between individuals with ED diagnosis with HC (g = 1.90; 95% CI 1.51, 2.29; p < 0.001).

Sensitivity Analyses
The Higgins I 2 heterogeneity statistics indicated small to moderate (11.4% to 66.95%) heterogeneity, aside from a meta-analysis on self-disgust with 84.19%.
For all meta-analyses, the Egger test for small-study effects was conducted to evaluate publication bias. The Egger's test results indicated that it was not necessary to remove any smaller studies from the analysis (z = 0.09, p = 0.93) for disgust elicited by food images. The trim and fill correction for missing data provided no evidence for any missing studies from the analysis.
Moreover, it was not necessary to remove any smaller studies from the analyses for generic disgust sensitivity (ED: z = 0.25, p = 0.81; AN: z = 1.51, p = 0.13; BN: z = −0.65, p = 0.51). The trim and fill correction for missing data was performed and revealed that there was no evidence for any missing studies on disgust sensitivity for the ED meta-analysis. However, one study for BN and two studies for AN were missing in the meta-analyses. The effect size was then re-calculated, which remained significant (BN: g = 0.60; 95% CI = 0.27, 0.92; AN: g = 0.51; 95% CI = 0.26, 0.76).
The Egger's test and funnel plot showed an asymmetric distribution across studies on self-disgust, suggesting significant publication bias. The trim and fill correction for missing data was also conducted, revealing that there were no missing studies from the analysis. See Supplementary Figures S1-S5 for funnel plots.

Summary of Evidence
A total of 52 studies were eligible for our systematic review. We could synthesise studies with regard to methodological approaches used to investigate disgust and selfdisgust. However, the results of studies that explored the neuroscience underpinnings of disgust were difficult to synthesise because the experimental paradigms or techniques were so varied. Our random-effects meta-analyses revealed that self-disgust (g = 1.90) and disgust in response to food images (g = 1.32) were significantly elevated in people with EDs, whereas generic disgust sensitivity was smaller (g = 0.49). The largest number of studies measured generic disgust, and it was thus possible to do a separate analysis for people with anorexia nervosa and bulimia nervosa. The results were similar across these diagnoses. However, the number of studies that measured self-disgust and food-related disgust was small, and few included people with BED, so it remains uncertain as to whether these are transdiagnostic constructs.

Strengths and Limitations
To the best of our knowledge, this is the first systematic review and meta-analysis summarizing measures, outcomes, and the role of disgust and self-disgust in the field of eating disorders both qualitatively and quantitatively. Nevertheless, the results of this study should be interpreted in the context of some limitations. First, most studies focused almost exclusively on adult females with AN and BN, limiting generalisability. Similarly, studies included in our systematic review could not provide knowledge on disgust-related constructs across other eating disorder diagnoses such as ARFID, pica, and rumination disorders. Thus, further research is needed across males, gender-diverse groups, and ED diagnoses, for example, to examine whether food-disgust and self-disgust are transdiagnostic elements. Second, the methods of measuring disgust and self-disgust varied. For example, some studies investigated disgust using complex items: "feeling disgusted/depressed/guilty" or "feeling of guilty, disgusted, and angry". More specific measures evaluating disgust may be needed using its distinctive features [84] to remove the possible confounding effect of other emotions. For example, studies measuring more objective measures such as the perception of disgust using facial EMG or facial recognition technology may hold promise [74,79]. Third, studies included in this systematic review mainly used visual stimuli, but auditory and olfactory stimuli might also be of use. Surprisingly, there are few studies on selfdisgust despite this being a key trait. Finally, studies exploring the wider disgust-related experiences through qualitative methodology were limited.
There are also potential limitations to the methodology of this systematic review and meta-analysis. First, we only included English articles, and search terms did not involve EDNOS and OSFED. Disgust is multi-faceted emotion, but our search terms did not specify terms which might include other facets of disgust, such as moral disgust [85] or sexual disgust [86]. For example, pudicity life events (i.e., events that had an element of sexual shame) were commonly found in the year that preceded the onset of anorexia nervosa [87]. Moral injury [88] or betrayal sensitivity [89] might be worth exploring.

Clinical Implications
These findings highlight the need to consider disgust in the psychopathology of eating disorders. Future studies might explore whether personality-based traits (e.g., perfectionism, cognitive, or moral rigidity, sensory sensitivity) predispose one to experiencing disgust towards food, eating and physical or behavioural characteristics of self, and possible differences across ED diagnoses.
Furthermore, disgust may need to be considered in the treatment of eating disorders. This may involve addressing disgust experienced in response to food or eating specific stimuli or using training approaches to moderate self-disgust. However, interventions directed at these emotions specifically are not yet readily available. Existing psychotherapies address difficult emotions in general, and these may include disgust and self-disgust, but they do not provide specific guidance on how to address them across ED diagnoses.
Traditional or adjunctive [90] exposure techniques that target aversive learning processes may be helpful to overcome disgust outcomes in response to food or eating. For example, future aversive expected outcomes might elicit a disgust response [6,15]. These may relate to weight gain or stigma and include negative evaluation from self and others [91].
In order to overcome future body/self-judgment expectancies that may relate to selfdisgust, a different approach may be needed. A variety of novel approaches have been developed, including virtual reality used to expose the individual to the experience of being in a larger body [10] and imaginal exposure [8]. These exposures may be embedded in approaches that encourage developing an identity that includes more self-compassion, such as MANTRA [92] and compassion-focused therapy [93].

Conclusions
Our systematic review and meta-analysis suggested that disgust and self-disgust appear to be of importance in eating disorders, but to date, they have been poorly researched across the spectrum of eating disorder diagnoses and as possible predisposing or perpetuating factors. Understanding the current evidence for the role of disgust in EDs can direct future research and the development of effective treatments targeting this emotion.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/xxx/s1, Figure S1: Funnel plot of studies of disgust elicited by food images in EDs versus HCs; Figure S2: Funnel plot of studies of generic disgust sensitivity in EDs versus HCs; Figure S3: Funnel plot of studies of generic disgust sensitivity in AN versus HCs after applying a trim-and-fill method; Figure S4: Funnel plot of studies of generic disgust sensitivity in BN versus HCs after applying a trim-and-fill method; Figure S5: Funnel plot of studies of self-disgust in EDs versus HCs; Table S1: PRISMA (2020) main checklist; Table S2: PRISMA (2020) abstract checklist; Table S3: Search terms used in the systematic search of the electronic databases; Table S4: Quality assessment of cross-sectional studies; Table S5: Quality assessment of case-control studies; Table S6: Quality assessment of qualitative studies; Table S7: Quality assessment of experimental studies; Table  S8: Quality assessment of randomized controlled trials.