Executive Functioning in Obesity, Food Addiction, and Binge-Eating Disorder

This study aimed to investigate food addiction (FA) and binge-eating disorder (BED) in their association to executive dysfunctions in adults with obesity. Data on response inhibition, attention, decision-making, and impulsivity were derived from four groups of adults with obesity: obesity and FA (n = 23), obesity and BED (n = 19), obesity and FA plus BED (FA/BED, n = 23), and a body mass index-, age-, and sex-stratified control group of otherwise healthy individuals with obesity (n = 23, OB), using established computerized neuropsychological tasks. Overall, there were few group differences in neuropsychological profiles. Individuals of the FA group did not differ from the OB group regarding executive functioning. Individuals with BED presented with significantly higher variability in their reaction times and a deficient processing of feedback for performance improvement compared to individuals of the OB group. Strikingly, individuals with FA/BED did not present neuropsychological impairments, but higher levels of depression than all other groups. The results indicated the presence of a BED-specific neuropsychological profile in the obesity spectrum. The additional trait FA was not related to altered executive functioning compared to the OB or BED groups. Future research is needed to discriminate FA and BED further using food-specific tasks.


Introduction
Obesity, defined as an excessive accumulation of body fat (body mass index (BMI) ≥ 30.0 kg/m 2 ) [1], presents one of the most prevalent health conditions in the Western world [2]. During the last decade, research on the causal and maintaining factors of obesity has focused on executive functioning and neural characteristics [3][4][5]. Executive functions represent a range of higher cognitive capacities enabling goal-directed behavior including inhibition, cognitive flexibility, planning, or decision-making [6,7]. Importantly, impairments in general executive functions in individuals with obesity, for example, reduced inhibition and planning, were linked to unsuccessful weight-loss related behavior, such as poor dietary quality and less success in weight loss therapies [8,9]. In order to provide targeted treatment options for executive dysfunctions, the specification of neuropsychological profiles in the obesity spectrum is warranted [10].
In this context, a recent review showed substantial similarities between executive functions in obesity and substance use disorder, including increased impulsive decision-making and an attentional bias towards disorder-related stimuli in both conditions [11]. In addition to these neuropsychological abnormalities, several neuroimaging studies documented a considerable overlap between obesity and substance use disorder regarding functional alterations in frontal brain regions and meso-corticolimbic circuits, which are involved in reward processing and decision-making [11][12][13].

Participants
Adults with obesity were recruited at the time of admission of two outpatient brain-directed psychological treatment studies at the Integrated Research and Treatment Center Adiposity Diseases in Leipzig, Germany (Cognitive remediation therapy for adults with obesity [41]; electroencephalographic neurofeedback training for adults with BED [42]), and were tested prior to the beginning of either intervention. Inclusion and exclusion criteria for both studies were similar. Relatedly, participants included in this study were required to be between 18 and 60 years of age, present with a BMI ≥ 35.0 kg/m 2 , display sufficient German language skills, and provide informed consent. Exclusion criteria were (1) serious somatic conditions (e.g., neurological disorders, stroke, head injury); (2) serious mental conditions (e.g., psychotic disorder, suicidality, substance use disorder, developmental or intellectual disability); (3) physical, mental, or other inability to participate in assessments (e.g., impediment in hearing, vision, or language); (4) previous or planned bariatric surgery; (5) use of medication that impacts weight or executive functioning (e.g., antipsychotics, sedatives, hypnotics); (6) current psychotherapy regarding weight or eating behavior; (7) current participation in other interventional studies; (8) lack of compliance; and (9) pregnancy or lactation. Individuals eligible for the OB group were required not to display any FA symptoms according to the YFAS 2.0 [22,43] and not to report any objective binge-eating episode during the last three months as determined via clinical interview [44,45]. Individuals eligible for the FA group needed to present with trait FA, according to the YFAS 2.0 [22,43], and were required not to report any objective binge-eating episode during the last three months as determined via clinical interview [44,45]. Individuals eligible for the BED group were required to present with the DSM-5 diagnosis of BED assessed with a clinical interview [44,45] and not to present with trait FA according to the YFAS 2.0 [22,43]. Individuals eligible for the FA/BED group had to present with trait FA, according to the YFAS 2.0 [22,43], and with the DSM-5 diagnosis of BED assessed with a clinical interview [44,45].

Ethics and Procedure
Both studies were approved by the Ethical Committee of Leipzig University Medical Center (256- 15-13072015, 143-15-20042015) and carried out in accordance with the Declaration of Helsinki. Within the two mentioned treatment studies, all participants were screened for eligibility using a standardized telephone interview. All eligible participants were invited to a baseline in-person diagnostic session. At the beginning of the diagnostic session, informed consent was obtained, followed by a series of neuropsychological tasks (see below). BMI was calculated based on participants' weight (in kg) and height (in cm) which were objectively measured using calibrated instruments at the end of the diagnostic session. The diagnostic sessions were standardized and conducted by trained staff. All participants were asked to refrain from eating at least 2 h prior to testing, in order to avoid any uncontrolled confounding effects of hunger or satiety on executive functioning.

Sample
Between November 2015 and April 2018, a total of n = 1082 individuals registered with an interest in one of the two brain-directed treatment studies, of whom a total of n = 258 were eligible and included in one of these brain-directed treatment studies. All n = 258 individuals participated in the baseline diagnostic session. For the present study, a convenience sample out of this study population was used, consisting of n = 23 individuals with trait FA based on the YFAS 2.0 [22,43], n = 19 with the DSM-5 diagnosis of BED based on clinical interview [44,45], and n = 23 individuals characterized by trait FA and BED (FA/BED). Out of n = 85 individuals eligible for the OB group, a total of n = 23 otherwise healthy individuals with obesity were stratified to the mean scores of the other groups regarding age, BMI, sex, and education.

Eating Disorder Examination
The diagnosis of BED according to DSM-5 [25] was derived from the diagnostic items of the validated German version of the Eating Disorder Examination interview (EDE 17.0D) [44,45]. According to DSM-5, the following severity levels of BED were determined: mild (1-3 objective binge-eating episodes per week), moderate (4-7 objective binge-eating episodes per week), severe (8-13 objective binge-eating episodes per week), and extreme (≥14 objective binge-eating episodes per week). The EDE has high interrater reliability (r ≥ 0.90) [46].

Yale Food Addiction Scale
The validated German version of the 35-item self-report Yale Food Addiction Scale 2.0 (YFAS 2.0; α = 0.914) [22,43] was used to determine FA. Participants were instructed to indicate how often they experienced addictive-like behavior towards food over the last year, with responses ranging from 0 (never) to 7 (every day). A total score was calculated by adding up all DSM-5 symptoms for addiction, ranging from 0 to 11. Scores ≥ 2 in combination with clinically significant distress indicate the presence of FA, with higher scores representing more severe FA: mild (2-3 symptom scores), moderate (4-5 symptom scores), and severe (≥6 symptom scores).

Eating Disorder Examination-Questionnaire
The validated German version of the short 8-item version of the Eating Disorder Examination-Questionnaire (EDE-Q8; α = 0.766) [47] was used to evaluate participants' global eating disorder psychopathology. A mean score, ranging from 0 to 6, was calculated, with higher scores indicating greater eating disorder psychopathology.

Patient Health Questionnaire-Depression Scale
The validated German version of the 9-item Patient Health Questionnaire-Depression Scale (PHQ-9; α = 0.813) [48,49] was administered to assess the level of depression based on the DSM-5 criteria for depression. Items were scored on a 0 (not at all) to 3 (almost every day) scale. Participants' sum scores were computed, ranging from 0 to 27, with higher scores representing more severe depression. Scores < 5 indicate no depressive symptoms, scores between 5 and 10 a mild depression, and scores > 10 a major depression [48,49].

Iowa Gambling Task
The computerized version of the Iowa Gambling Task (IGT) [50] was used to assess decision-making under uncertainty and complexity. The IGT requires participants to choose 100 times from four possible card decks (A, B, C, and D), aiming at winning the highest possible amount of virtual money. Each card deck contains profits and losses of different amounts. In the long term, choosing from card decks A and B will result in overall long-term loss (disadvantageous decks), whereas choosing cards from card decks C and D will result in overall long-term gain (advantageous decks). Performance in the IGT was determined by the total net score, calculated by subtracting the total number of disadvantageous choices from the total number of advantageous choices. Additionally, a learning effect across the net score of the five consecutive blocks of 20 cards was considered.

Delay Discounting Task
The computerized version of the Delay Discounting Task (DDT) [51], provided by the test software Millisecond [52] was used to determine the individual discounting rate of delayed rewards, which represents a measure of impulsive decision-making [51]. In this paradigm participants were instructed to choose between a standard amount of money (10 EUR) with different temporal delays (0, 2, 30, 180, and 365 days) or a variable amount (0-10 EUR) with no delay, until an indifference point for each delay was found or until the maximum number of trials (30) has been run for each delay. Five indifference points were calculated, where the immediate reward and delayed reward were equal in terms of subjective value. To determine performance in the DDT, the five indifference points were used to calculate the area under the curve (AUC) [53]. The AUC may range from 0 to 1 with greater AUC values being associated with less discounting of delayed rewards, i.e., less impulsivity [53].

Go/No Go
The computerized version of the visual Go/No Go paradigm [54], provided by the Vienna Test System [55], was used to assess inhibitory control. The visual Go/No Go paradigm required participants to determine whether a stimulus required a reaction or an inhibition. Therefore, 202 triangles were shown on the screen, which required a fast response, while a total of 48 circles appeared randomly in-between, indicating that no response should be given. Performance in the Go/No Go paradigm was determined by the number of commission errors (false positive responses towards a No Go trial), with more commission errors suggesting diminished inhibitory control [56].

Wisconsin Card Sorting Test
The computerized version of the Wisconsin Card Sorting Test (WCST) [57] was used as a measure of cognitive flexibility. The WCST required participants to sort each of maximally 128 cards according to three possible principles or rules (e.g., shape, color, or number), which were unknown to the participant, but would become apparent via feedback during the trials. After 10 consecutively and correctly sorted cards the rule changed suddenly and without warning. The participant then had to generate solutions and use the feedback to adapt to the new rule. Performance on the WCST was measured with the percentage of perseverative errors, which represent a tendency to perseverate on the previous rule, and the learning to learn score, which represents the average growth of conceptual efficiency during the test (average difference in percent errors between successive categories). The learning to learn score can only be calculated if participants have completed at least three categories or completed two categories and attempted a third. The learning to learn score can obtain positive and negative values, with positive scores indicating enhanced efficiency across consecutive categories, presumably because of learning [57].

Alertness
Alertness and the variability in early attentional processes was assessed with a computerized perception and attention functions battery (WAFA) [58], provided by the Vienna Test System [55]. To assess visual intrinsic alertness, participants were required to look at a fixation cross at the center of the screen and press a button as fast as possible when a target stimulus, a circle, appeared. For the assessment of phasic alertness, an auditory cue appeared shortly before the visual stimulus, preparing the participant for next execution of the reaction. Performance in intrinsic and phasic alertness was reported as log-transformed mean reaction time, with higher scores indicating longer reaction times and therefore less alertness [58]. Variability in performance in intrinsic and phasic alertness was reported as a measure of dispersion, with higher scores indicating greater variability in intrinsic or phasic alertness.

Statistical Procedure
Statistical analyses were performed using IBM SPSS Statistics version 24 (IBM Corp. Released 2016. IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp). All tests were two-tailed and considered significant when p values were <0.05. Since the domains of executive functioning are distinct and based on single measures, adjusting for multiple testing was not necessary, except for the computerized alertness task (WAFA) where four outcome measures were used. The p value for testing group differences in the WAFA was therefore adjusted to p = 0.05/4 = 0.0125. For the comparison of groups regarding socio-demographic and clinical variables, univariate analyses of variance (ANOVAs) were calculated for continuous variables and Chi-square tests (X 2 ) for categorical variables. The prerequisites for performing the tests (normal distribution and homogeneity of variance) were tested with the Shapiro-Wilk and Levene tests. When prerequisites were not met, Welch ANOVA was used, because of its robustness against violations of the assumption of homogeneity of variances [59]. In case of significance and homogeneity of variances, the Bonferroni post-hoc test was used to determine pair-wise differences; in case of violation of homogeneity of variances the Games-Howell post-hoc test was used. For Welch ANOVA, estimated omega squared (est. ω 2 ) was calculated for which ω 2 ≥ 0.01 is considered a small effect, ω 2 ≥ 0.06 a moderate, and ω 2 ≥ 0.14 a large effect [60]. For mean differences in post-hoc tests Cohen's d was calculated, with d ≥ 0.2 being considered a small effect, d ≥ 0.5 a moderate, and d ≥ 0.8 a large effect [61]. In the present study, a power of ≥ 0.75 was observed to detect moderate and ≥ 0.95 to detect large group differences in ANOVAs (two-tailed p < 0.05).

Sample
As depicted in Table 1, groups did not differ in socio-demographic characteristics. Significant differences between groups were found in eating disorder psychopathology and depression. For eating disorder psychopathology, post-hoc tests did not reach significance. The FA/BED group showed significantly more depressive symptoms compared to the OB group (5.17, 95% Confidence interval (CI) (1.86, 8.49), p ≤ 0.001). Significantly more individuals of the FA/BED group presented with severe FA compared to the FA group, where there were more cases of mild FA. The FA/BED group did not differ significantly regarding BED severity from the BED group.

Neuropsychological Assessment
In the WAFA, groups differed significantly in the variability to react to a cued target (measure of dispersion phasic alertness), see Table 2, with individuals with BED showing significantly more variability in reaction times than individuals with OB (0.88, 95% CI [0.13, 1.63], p = 0.018, d = 1.12). Significant group differences were furthermore observed in the growth of conceptual efficiency in the WCST (learning to learn score), indicating that individuals with BED showed significantly less growth of conceptual efficiency in learning compared to individuals with OB (−3.47, 95% CI [−6.92, −0.01], p = 0.049, d = −0.85). Regarding decision-making, all groups improved their net score over the five consecutive blocks of the IGT, as depicted in Figure 1. Although descriptive data of the IGT net score seem to differ substantially, no significant group differences were found and only a small effect size was observed (see Table 2), suggesting that trait FA and BED diagnosis were not associated with altered decision-making in individuals with obesity. In addition, groups did not differ in their mean reaction time (intrinsic and phasic) and variability to react to un-cued visual targets (measure of dispersion intrinsic alertness) in the WAFA, the amount of perseverative errors in the WCST, and the amount of commission errors in the Go/No Go task (see Table 2), indicating that trait FA and BED were not associated with altered attention, cognitive flexibility, and inhibition in individuals with obesity.

Neuropsychological assessment
In the WAFA, groups differed significantly in the variability to react to a cued target (measure of dispersion phasic alertness), see Table 2, with individuals with BED showing significantly more variability in reaction times than individuals with OB (0.88, 95% CI [0.13, 1.63], p = 0.018, d = 1.12). Significant group differences were furthermore observed in the growth of conceptual efficiency in the WCST (learning to learn score), indicating that individuals with BED showed significantly less growth of conceptual efficiency in learning compared to individuals with OB (−3.47, 95% CI [−6.92, −0.01], p = 0.049, d = −0.85). Regarding decision-making, all groups improved their net score over the five consecutive blocks of the IGT, as depicted in Figure 1. Although descriptive data of the IGT net score seem to differ substantially, no significant group differences were found and only a small effect size was observed (see Table 2), suggesting that trait FA and BED diagnosis were not associated with altered decision-making in individuals with obesity. In addition, groups did not differ in their mean reaction time (intrinsic and phasic) and variability to react to un-cued visual targets (measure of dispersion intrinsic alertness) in the WAFA, the amount of perseverative errors in the WCST, and the amount of commission errors in the Go/No Go task (see Table 2), indicating that trait FA and BED were not associated with altered attention, cognitive flexibility, and inhibition in individuals with obesity.

Discussion
The present study aimed at investigating the associations of trait FA and the DSM-5 diagnosis of BED regarding executive functions in adults with obesity. It was found that individuals with BED showed lower executive functioning, characterized by a higher variability in reaction times and less feedback integration, compared to otherwise healthy individuals with obesity. Contrary to expectations, there were no differences in executive functioning among individuals with FA, FA/BED, and obesity only. The results thus suggest a BED-specific, but not a FA-specific, neuropsychological profile in individuals with obesity. The combined presentation of trait FA and BED was not associated with greater impairments in executive functioning compared to their single presentations.
The fact that individuals with obesity and BED displayed a significantly higher variability in their reaction times compared to the OB group is similar to findings showing a high variability in reaction times in ADHD [62]. ADHD is highly comorbid to obesity and BED and is associated with comparable neuronal processes [62][63][64]. In ADHD, it is usually believed that the variability of response times reflects occasional gaps in attention [62], which could be related to high distractibility, making an attentional shift and the detection of task-irrelevant stimuli, for example, food cues, more likely. Another result was the significant and medium-sized difference between the BED and the OB group in the learning to learn score in the WCST, in accordance with recent research [37]. The negative learning to learn score indicated that individuals with BED became less rather than more efficient on consecutive trials of the WCST [57]. Hypothetically, this could be caused by an inability to appropriately integrate the feedback in order to improve performance, as previously evidenced in BED [29]. Transferred into everyday life, the deficient processing of feedback to improve performance could be related to a lack of feedback integration in daily learning, such as feedback given by one's own body after an objective binge-eating episode in order to prevent such an episode from occurring again, leaving an individual without control [65]. In this context, it is worth noting that the BED group did not display heightened depressive symptoms compared to the OB and FA group. Observed differences between the BED and OB group can therefore not be attributed to increased general psychopathology, but were related to BED diagnosis.
Contrary to the two previous studies investigating executive functions in individuals with trait FA [23,24], demonstrating that higher versus lower FA severity in obesity was related to poorer decision-making, inhibition, and cognitive flexibility, no other performance score reached significance in the present study. The hypothesis that individuals with obesity and FA may present with a specific neuropsychological profile in the obesity spectrum was therefore not supported in this clinical treatment-seeking sample. It is of note, however, that the FA group mostly presented with mild and moderate FA symptomatology only, allowing to speculate that FA severity was not high enough to be accompanied by cognitive impairments in this sample. On the other hand, this result could also reflect the assumption that BED, as a clinical eating disorder diagnosis, is associated with more cognitive impairments than subjective trait FA.
The hypothesis of cumulative effects on executive dysfunctions in individuals with FA plus BED was not confirmed in the present study, since the FA/BED group did not display lower performance compared to the OB, FA, or BED group in any of the tasks under investigation. Notably, the FA/BED group reported considerably more depressive symptoms compared to the OB group, with an average in the range of a major depression [48,49], and thus contradicts neuropsychological evidence of impaired executive functioning in individuals with depression [66]. However, the underlying mechanisms of this finding cannot currently be identified. Therefore, this result needs replication in larger, independent samples of individuals with FA/BED.

Methodological Considerations
When interpreting the results several strength and limitations have to be mentioned. First, using well-established measures, the diagnosis of BED was derived from an expert interview, the EDE 17.0D [44,45], and trait FA was assessed via the established self-questionnaire YFAS 2.0 [22,43]. Second, the OB group was stratified to the other groups and groups were mutually exclusive regarding trait FA and objective binge-eating episodes. Third, executive functioning was measured with established, standardized, and computerized tasks. Critically, the relatively small group sizes have to be mentioned, as only medium-to-large-sized effects could be detected with adequate power. Additionally, there might be an assessment bias across diagnostic groups, as FA was assessed with a self-administered questionnaire, while BED diagnosis was made by clinical expert interview, suggesting that trait FA and BED mirror different levels of clinical severity. Furthermore, no food-specific tasks were used. Various studies have shown a stronger association between executive functions in food-related tasks in relation to binge-eating behavior [30,38]. Lastly it is important to note that the design of this study was cross-sectional; therefore, causal conclusions about the direction of observed effects are not indicated.

Implications and Future Directions
The current study identified greater executive dysfunctions in individuals with obesity and comorbid BED compared to those with obesity only. Clinically, the assessment of eating disorder symptomatology, specifically a diagnosis of BED, seems therefore indicated in individuals with obesity. In these individuals, an increase in neuropsychological impairments is likely to interfere with standard treatment options [8,9] and may thus constitute a specific target for intervention, for example, via cognitive remediation therapy [41]. Based on this study's results, an in-depth investigation in the variability of reaction times (alertness tasks) is indicated, especially considering the effects of food stimuli, as current research points to an increased attentional bias towards food cues in individuals with BED e.g., [66,67].
Regarding trait FA, this study revealed no impairments in executive functioning other than a tendency towards higher choice-impulsivity, which could be related to the high proportion of individuals characterized by low FA severity in the FA group. Studies investigating executive functions in individuals with obesity and more severe FA are therefore needed. Alternatively, it could be argued that similarities in executive functioning between individuals with obesity plus trait FA and individuals with substance use disorder as recently shown [12] only manifest on a neurobiological [12] and not on a neuropsychological level. To further differentiate these aspects of addictive-like eating behavior, future research is warranted to specifically investigate neuronal and neuropsychological overlaps and distinctive features between trait FA and substance use disorder, especially in the context of disorder-related stimuli (e.g., food stimuli).
It was surprising that the observed large differences in attentional variability and conceptual learning between the BED and OB group did not manifest in the FA/BED group, particularly as this group presented as the most impaired group concerning mental health, with more depressive symptoms than the OB group. The assessment of trait FA in individuals with versus without BED may therefore be helpful to determine identify individuals at risk for greater mental health impairment and in need of more intensive treatment. Further research into the neuropsychological characteristics of individuals with FA plus BED is however warranted, preferably using larger samples and a longitudinal study design in order to evaluate the effects of trait FA plus BED on weight trajectories and treatment outcome in individuals with obesity.