Impulsivity and Compulsivity in Binge Eating Disorder: A Systematic Review of Behavioral Studies

Abstract

Background:

Binge eating disorder (BED) often includes impulsive and compulsive behaviors related to eating behavior and food. Impulsivity and compulsivity generally may contribute to the etiology and maintenance of multiple psychiatric disorders including BED. This review aimed to identify and synthesize available behavioral studies of impulsivity and compulsivity among individuals with BED.

Method:

A systematic search was performed focusing on BED and specific facets of impulsivity (rapid response and choice) and compulsivity (cognitive flexibility, set-shifting, and/or habit learning). All case-control studies comparing adults with either full-threshold or subthreshold BED to individuals with normal weight, overweight, or other eating disorders (e.g., bulimia nervosa) were included.

Results:

Thirty-two studies representing 29 unique samples met inclusion criteria. Increased choice impulsivity was observed among individuals with BED relative to individuals with normal weight. There were mixed findings and/or a lack of available evidence regarding rapid response impulsivity and compulsivity. The presence of between-group differences was not dependent on sample characteristics (e.g., full or partial-threshold BED diagnosis, or treatment-seeking status). Heterogeneity relating to covariates, task methodologies, and power limited conclusions.

Conclusions:

Literature supports an association between choice impulsivity and BED. More research is needed to determine if individuals with BED demonstrate elevated levels of either rapid response impulsivity or types of compulsivity. There exists a need for continued research, with particular attention paid to important covariates, task methodologies, and issues of power.

Introduction

A hallmark feature of many psychiatric disorders, such as substance-use, gambling, obsessive-compulsive and binge-eating disorders, is diminished control over behavior^1–3. The impulsive and compulsive transdiagnostic features may represent intermediate phenotypes or endophenotypes¹. Impulsivity has been defined as, “a predisposition toward rapid, unplanned reactions to internal or external stimuli with diminished regard to the negative consequences of these reactions to the impulsive individual or to others.”^4,5 In contrast, compulsivity has been defined as involving “the performance of repetitive and functionally impairing overt or covert behavior without adaptive function, performed in a habitual or stereotyped fashion, either according to rigid rules or as a means of avoiding perceived negative consequences.”¹ An early conceptual model⁶ described an impulsive-compulsive spectrum, where impulsivity and compulsivity were viewed as opposite ends of a single dimension. However, more recent data suggest that both impulsivity and compulsivity represent complex multi-faceted constructs. The relative contribution of impulsivity and compulsivity to a clinical presentation may vary as a function of psychiatric condition, personality, or disease course⁷.

Significant efforts have been made to improve understanding of the neurobehavioral underpinnings of diminished self-control, including behavioral tests of impulsivity and compulsivity. To date, evidence using behavioral paradigms indicates significant impulsivity and compulsivity among people with obsessive-compulsive disorder¹, substance-use disorders⁸, and other addictive disorders (e.g., gambling disorder)^9,10 Less is known about how these constructs may relate to the etiology, maintenance and treatment of binge-eating disorder (BED). BED is characterized by eating large amounts of food within a discrete time period, with a concurrent sense of loss of control¹¹. BED has been described as sharing similarities with addictions, including positive associations with impulsivity^12,13. BED is associated with psychiatric comorbidities¹⁴, decreased work productivity, higher health care utilization and costs¹⁵, and increased suicidality¹⁶. BED also frequently co-occurs with obesity, a state with multiple etiologies (Keith et al., 2006). Obesity may be influenced by environmental (e.g., food availability and food policy) and individual (e.g., epigenetic processes, physical activity, and eating behavior) factors. After controlling for relevant covariates, individuals with BED showed a more than 5-fold increase in experiencing extreme obesity¹⁷ (defined by a body mass index (BMI) over 40)¹⁸. Nonetheless, most individuals with obesity do not have BED¹⁹. Multiple studies have shown increased levels of eating-disorder psychopathology among individuals with BED as compared with obesity alone^20,21. Regarding neurobiological underpinnings, corticostriatal dysfunction may underlie BED²², with other neurochemical systems (e.g., monoamines, opioids, orexins, cannabinoids, glutamate, GABA) having been implicated in preclinical and clinical studies of BED²³.

To understand how impulsivity and compulsivity may relate to BED, important measurement and conceptual issues should be considered. Significant theoretical work emphasizes two distinct types of impulsivity^1,24,25: rapid-response and choice impulsivity. Other psychological processes (e.g., reflection) are related to impulsivity, but also involve other dissociable processes, including difficulties or differences in gathering information needed to make decisions. In contrast, rapid-response and choice impulsivities are understood as exclusively forms of impulsivity and thus will be considered here. Compulsivity is also a heterogeneous construct with varying definitions/forms. Based on neurobiological evidence, we will consider cognitive flexibility, set-shifting, and habit-learning forms of compulsivity^1,26. Some authors argue in favor of attentional bias as a form of compulsivity, but due to its overlap with other processes including impulsivity, it will not be considered here. We will consider behavioral evidence in human studies. Significant research has investigated relationships between self-reported impulsivity and BED.^27–29. However, as noted in other reviews, self-reported impulsivity and compulsivity are not always highly correlated with behavioral measures of these constructs^30,31. Both self-report and behavioral measures have demonstrated prognostic significance, but most data indicate that their effects are likely independent^30,31. Overall, we aim to address 5 principal questions:

1. Do individuals with BED and individuals with obesity and without BED differ on behavioral measures of impulsivity and compulsivity?

2. Do individuals with BED and individuals with normal weight and without BED differ on behavioral measures of impulsivity and compulsivity?

3. Do individuals with BED and individuals with other types of eating disorders differ on behavioral measures of impulsivity and compulsivity?

4. What are important future directions, with a specific focus on conceptual and methodological considerations?

Methods

This systematic review was conducted following prior recommendations. The following elements are included: objectives; systematic search using explicit inclusion/exclusion criteria; description of study selection and characteristics; results of individual studies; summary of results using narrative and tabular format; and, discussion of the overall quality of the literature, direction of effects, and future directions.

Search Strategy

A systematic search was completed on 01/31/2020. All relevant articles published to this date were considered. The PRISMA flow chart is displayed in Figure 1. Searched databases included PubMed, EMBASE, and Psychinfo, with a combination of controlled vocabulary words and search terms combined using Boolean logic. Boolean logic combines keyword searches using operators such as “and”, “or”, and “not”. The goal was to identify all studies reporting associations between binge eating and impulsivity and compulsivity among adults. Included studies were further refined to reflect only those studies that used behavioral tests specific to the two impulsivity constructs (rapid-response and choice) and three compulsivity constructs (cognitive flexibility, set-shifting, and habit learning). Wherever possible, tasks that met inclusion criteria were decided upon in advance, and any tasks not discussed a priori were discussed between the primary (MMC) and senior author (MNP) before inclusion or exclusion. For both rapid-response and choice impulsivity, a consensus statement by the International Society for Research on Impulsivity was used to identify behavioral tasks that were appropriate for inclusion^24,25. These reviews also discuss at length the construct validity associated with these tasks. For compulsivity, several reviews were considered to develop the compulsivity-related assessments^22,26,32. Search terms included:

Fig. 1.

PRISMA flow chart of study identification and inclusion.

((exp Binge-Eating Disorder/ [Medline]; Binge Eating Disorder/ [EMBASE]; Binge Eating Disorder/ [Psychinfo] or (binge eating or binge-eating).tw.) and (exp Impulsive Behavior/ [Medline]; Impulsiveness / [EMBASE]; Impulsiveness / [Psychinfo]; or (impulsiv* or “rapid response impulsivity” or “go/no go” or “go no go” or “stop signal” or “continuous performance” or “stroop” or “richard’s task” or “richards task” or “monetary choice questionnaire” or “experiential delay task”).tw.)) or ((exp Binge-Eating Disorder/ [Medline]; Binge Eating Disorder/ [EMBASE]; Binge Eating Disorder/ [Psychinfo] or (binge eating or binge-eating).tw.) and (exp Compulsive Behavior/ [Medline]; compulsion/ [EMBASE]; or (compulsive or compulsion or “cognitive flexibility” or perseverat* or “set-shifting” or “switching” or “habit learning” or “goal-directed” or “probabilistic learning” or “rule shift card” or “Wisconsin card sorting” or “trail making” or “intra-dimensional/extradimensional” or “two-stage task” or “slips of action” or “sensory-specific satiety”).tw.))

Reference lists of recent reviews were also manually searched for additional studies.

Eligibility Criteria

Case-control studies comparing individuals over the age of 18 years with either full-threshold BED, subthreshold BED, or binge-eating symptoms to any of the following groups were considered: individuals with overweight or obesity defined as a BMI ≥ 25.0, individuals with normal weight defined as a BMI < 25.0 and ≥ 19, and individuals with an eating disorder or eating condition other than BED. Studies were required to be full-text, empirical articles published in English. Conference proceedings, position papers, and dissertations were not included. Studies must have included at least one appropriate behavioral measure and results must report on group differences between case and controls. Studies could be either single-session experiments, single-session observational studies, or prospective studies as long as case-control comparisons for baseline data were available. Studies presenting data in more than a single paper were included but noted in results as appropriate.

Results

In total, 32 studies representing 29 unique samples met inclusion criteria. See Table 1 for a summary of behavioral tasks considered; underlining indicates that at least one included study used the measure. Study descriptions and behavioral findings are summarized in Tables 2–4.

Table 1.

Behavioral Tasks Relevant to Rapid Review

Impulsivity
Rapid Response Impulsivity: Go/No-Go, Stop Signal Task, Continuous Performance Task, Five-Choice Serial Reaction Time Task,
Choice Impulsivity: Richard’s task, Monetary Choice Questionnaire, Experiential Delay Task, Delayed Discounting Task, Monetary Incentive Delay Task
Compulsivity
Set-Shifting: Rule Shift Card, Wisconsin Card Sorting Task, Trail Making Test, Intra-Dimensional/Extradimensional, DKEFS Trail Making Test, NIH dimensional Change Card Sort, Penn Conditional Exclusion Task
Cognitive Flexibility: Probabilistic Learning Task
Habit Learning: Two-Step Task, Slips of Action Task, and Sensory-Specific Satiety Task

Notes. Underlining indicates tasks that were used in studies that met inclusion criteria.

Table 2.

Summary of Included Studies Investigating Impulsivity

				Impulsivity
Author(s)	BE/BED Sample	BE/BED BMI Categorya	Control Group(s)	Construct	Task	Major Findings	Summaryb
Voon et al, 201457	BED (n = 30; n-TS)	Obesity 2	OB, NW	Response impulsivity	Five-choice serial reaction time task	BED group showed no significant differences on 5-CSRT compared to OB and NW groups.	BED ≈ NW impulsivity BED ≈ OB impulsivity
Córdova et al., 201737	BE (n = 18; u-TS)c	Obesity 2	OB	Response impulsivity	GNG	BE group showed more commission errors on the GNG task as compared with OB group.	BED > OB impulsivity
Hege et al., 201538	BED (n = 18; u-TS)	Obesity 1	OB	Response impulsivity	GNG (Food)	BED group showed no significant differences on GNG compared to OB group. BED group did show a differential pattern of activation.	BED ≈ OB impulsivity
Loeber et al., 201839	BED (n = 17; n-TS)	Obesity 2	OB, NW	Response impulsivity	GNG (Food)	BED group made significantly more errors on the neutral version of the GNG as compared to the food version. NW group made significantly more errors on the food version of the GNG as compared to the neutral version. No main effects (e.g., group differences) were significant.	BED / OB impulsivity BED / NW impulsivity
Lyu et al., 201747	BE (n = 31; n-TS)c	NW	NW	Response impulsivity	GNG (Food)	BE group showed no significant differences on GNG compared to NW group.	BED ≈ NW impulsivity
Mobbs et al., 201140	BED (n = 16; TS)	Obesity 2	OB, NW	Response impulsivity	GNG (Food, Body)	BED group show more commission errors for food and body stimuli compared to OB group. OB and BED combined group showed more errors than NW group. BED and NW group not directly compared.	BED > OB impulsivity
Oliva et al, 201945	BED (n = 21; u-TS)c	NW	NW	Response impulsivity	GNG (Food), SST (Food),	BE group showed no significant differences on GNG or SST compared to non-BE/NW group.	BE-NW ≈ nonBE-NW impulsivity
Bartholdy et al., 2017a50	BED (n = 11; p-TS)	Overweight	AN, BN, NW	Response impulsivity	SST, Cued RT task	BED group showed no significant differences for SST or Cued RT task compared to AN, BN, or NW.	BED ≈ NW impulsivity BED ≈ AN impulsivity BED ≈ BN impulsivity
Manasse et al., 201651	BED (n = 25; TS)d	Obesity 2	OB	Response impulsivity	SST (Food)	BED group showed slower stop signal reaction time compared to OB group. The effect was not stimulus dependent.	BED > OB impulsivity
Svaldi et al., 201452	BED (n = 31; n-TS)	Obesity 2	OB	Response impulsivity	SST (Food)	BED group showed slower stop signal reaction time and greater errors for food stimuli compared to OB.	BED > OB impulsivity
Wu et al., 201353	BED (n = 54; p-TS)	Obesity 1	OB	Response impulsivity	SST	BED group showed no significant differences on SST compared to OB group.	BED ≈ OB impulsivity
Preuss et al., 201954	BED (n = 24; TS)	Obesity 1	OB, NW	Response impulsivity	SST (Food),	BED group showed better performance on SST compared to OB and NW groups.	BED < OB impulsivity BED < NW impulsivity
Davis et al, 201068	BED (n = 65; n-TS)f	Obesity 2	OB, NW	Choice impulsivity	Delayed Discounting task	BED group showed steeper delay discounting compared to NW group, but the effect disappeared when education was controlled for. BED group did not show significantly steeper discounting compared to OB group.	BED ≈ OB impulsivity BED / NW impulsivity
Barthholdy et al., 2017b74,e	BED (n = 11; p-TS)	Overweight	AN, BN, NW	Choice impulsivity	Delayed Discounting task	BED group did not show steeper delay discounting compared to AN, BN, or NW groups.	BED ≈ AN impulsivity BED ≈ BN impulsivity BED ≈ NW impulsivity
Manwaring et al.,201169	BED (n = 30; n-TS)	Obesity 3	OB, NWg	Choice impulsivity	Delayed Discounting task (Food)	BED group showed steeper delay discounting for a variety of stimuli including food compared to NW and OB groups.	BED > OB impulsivity BED > NW impulsivity
Steward et al., 201772	BED (n = 24; TS)	Obesity 2	AN-R, AN-BP, NW	Choice impulsivity	Monetary choice questionnaire	BED group showed steeper discounting as compared with NW group and AN-R group. BED group was not significantly different than AN-BP subgroup.	BED > NW impulsivity BED / AN impulsivity
Yan et al, 201873	BE (n = 85; n-TS)	NW	NW	Choice impulsivity	Monetary Choice Questionnaire	BE group did not show steeper delay discounting compared to NW.	BE ≈ NW impulsivity
Balodis et al., 201370	BED (n = 19; TS)	Obesity 2	OB, NW	Choice impulsivity	Monetary incentive delay task	BED group showed no differences on task compared to OB and NW groups. BED groups showed differential pattern of activation compared to OB and NW groups.	BED / OB impulsivity BED / NW impulsivity
Mole et al, 2O1555,e	BED (n = 30; n-TS)	Obesity 2	OB, NW	Response impulsivity, Choice impulsivity	SST, Monetary choice questionnaire	BED group showed steeper discounting compared to NW group. BED group showed no differences on the SSDRT compared to NW group and better SSDRT compared to OB group. BED and OB group not compared on delay discounting measures.	BED / NW impulsivity BED < OB impulsivity

Notes. AN: individuals with anorexia; AN-BP: individuals with anorexia binge purge subtype; AN-R: individuals with anorexia restricting subtype; BE: Binge eating; BED: Binge eating disorder; BMI: body mass index; BN: individuals with bulimia; CSRT: Cued-serial reaction time; GNG: go/no-go; n-TS: non-treatment seeking; NW: individuals with normal weight; OB: individuals with obesity; p-TS: partially treatment seeking; RT: reaction time; SSDRT: stop signal delay reaction time; SST: stop signal task; TS: treatment seeking; u-TS: unknown treatment seeking status.

^a.BMI for BED group rounded to the nearest whole number and classified according to CDC guidelines of normal weight, obesity class 1, obesity class 2, and obesity class 3.

^b.> indicates greater levels of impulsivity or compulsivity, ≈ indicates similar levels of impulsivity or compulsivity, / indicates mixed findings

^c.BED cases versus controls assessed via self-report;

^d.sample included some subthreshold BED cases or BE symptom cases

^e.Sample included in two studies reporting different outcomes.

^f.NW group included individuals with BMI up to 27.

^g.Modified DSM-IV TR BED diagnosis used. Weekly binging of objectively large amounts without compensatory behaviors.

^h.Study included a delayed discounting task, but pairwise outcomes were not reported

Table 4.

Summary of Included Studies Investigating Impulsivity and Compulsivity

			Impulsivity and Compulsivity
First Author	BED Sample	BE/BED BMI Categorya	Control Group(s)	Construct	Task	Major Findings	Summaryb
Kelly et al., 201333	BE (n = 50; n-TS)c	Overweight	NW	Response impulsivity, Set-shifting	Continuous performance task, WCST	BED group showed no significant differences on commission errors or perseverative errors in the CPT or WCST respectively.	BED ≈ NW impulsivity BED ≈ NW compulsivity
Grant et al., 201956	BED (n = 17; n-TS)	Obesity 1	OB	Response impulsivity, Set-shifting	SST, IED	BED group showed significantly slower SSDRT and no difference in perseverative errors on the IED compared to the OB group.	BED > OB impulsivity BED ≈ OB compulsivity
Kollei et al., 201841	BED (n = 48; p-TS)	Obesity 3	OB, NW	Response impulsivity, Set-shifting	GNG (Food), Choice RT Task, IED	BED group showed no significant difference on the GNG as compared with the OB and NW groups. Remaining test used a composite dependent variable and discriminant function analysis. Reversal learning discriminated well between BED and OB groups, while commission errors from choice serial reaction time test did not. Group by behavioral construct for other differences not reported.	BED ≈ OB impulsivity BED ≈ NW impulsivity BED / OB compulsivity BED ≈ NW compulsivity
Blume et al, 201942,h	BED (n = 23; TS) BED+FA (n = 19; TS)	BED: Obesity 3 BED+FA : Obesity 3	OB, FA	Response impulsivity Set-shifting	GNG WCST	BED and BED+FA groups did not show significant differences on the GNG or WSCT compared with OB and FA groups.	BED ≈ OB impulsivity BED ≈ FA impulsivity BED ≈ BED+FA impulsivity BED+FA ≈ OB impulsivity BED + FA ≈ FA impulsivity BED ≈ OB compulsivity BED ≈ FA compulsivity BED ≈ BED+FA compulsivity BED+FA ≈ OB compulsivity BED + FA ≈ FA compulsivity
Wollenhaupt et al., 201946	BED (n = 10; TS)	Obesity 3	AN, BN, NW	Response impulsivity, Set-shifting	GNG (food), WCST, TMT-B	BED group did not show significant differences on the GNG, WSCT, or TMT-B as compared with AN, BN, and NW groups.	BED ≈ AN impulsivity BED ≈ BN impulsivity BED ≈ NW impulsivity BED ≈ AN compulsivity BED ≈ BN compulsivity BED ≈ NW compulsivity
Manasse et al., 201571	BED (n = 31; TS)d	Obesity 2	OB	Choice impulsivity, Set-shifting	Delayed Discounting task, Penn Conditional Exclusion Task	BED group showed steeper discounting delays as compared with OB group. BED group showed no significant differences in the Penn Conditional Exclusion task compared to OB group.	BED > OB impulsivity BED ≈ OB compulsivity

Notes. AN: individuals with anorexia; BE: Binge eating; BED: Binge eating disorder; BMI: body mass index; BN: individuals with bulimia; FA: Food addiction; GNG: go/no-go; IED: intra-dimensional/extra-dimensional shift task; n-TS: non-treatment seeking; NW: individuals with normal weight; OB: individuals with obesity; p-TS: partially treatment seeking; SSDRT: stop signal delay reaction time; SST: stop signal task; TMT-B: trail making test B; TS: treatment seeking; WCST: Wisconsin card sorting test.

^a.BMI for BED group rounded to the nearest whole number and classified according to CDC guidelines of normal weight, obesity class 1, obesity class 2, and obesity class 3.

^b.> indicates greater levels of impulsivity or compulsivity, ≈ indicates similar levels of impulsivity or compulsivity, / indicates mixed findings

^c.BED cases versus controls assessed via self-report;

^d.sample included some subthreshold BED cases or BE symptom cases

^e.Sample included in two studies reporting different outcomes.

^f.NW group included individuals with BMI up to 27.

^g.Modified DSM-IV TR BED diagnosis used. Weekly binging of objectively large amounts without compensatory behaviors

^h.Study included a delayed discounting task, but pairwise outcomes were not reported

Impulsivity

Rapid Response Impulsivity

Rapid response impulsivity describes a tendency to act immediately, and such behavior occurs with minimal forethought and does not meet the demands of the current context.²⁴ When behaviorally operationalized, it may be understood as reduced tendencies to inhibit prepotent responses.⁴ In humans, rapid-response impulsivity is primarily assessed via two tasks: the go/no-go task (GNG) and stop-signal task (SST). For the following review, the procedures and outcomes of interest for each test will be introduced. Next, behavioral results for each task will be summarized. Finally, an overall summary of the available evidence for rapid response impulsivity will be presented. A similar approach will be applied in other sections.

Go-No-Go: Task Description.

Ten studies investigated rapid response impulsivity using the GNG (Table 2 and 3). GNGs were originally adapted from a preclinical procedure, though there are important differences²⁴. For included studies, the GNGs used were computerized and instructed individuals to provide a motor response (usually pressing a space bar) in response to specific stimuli. These trials are known as go-trials. Go-trials are more frequent to establish this motor output as the dominant response. In contrast, stop-trials require a participant to withhold the previously established, prepotent motor response. Errors in response to stop-trials, or commission errors, are the primary measure of rapid response impulsivity²⁴. The Conner’s Continuous Performance Task is a proprietary version of the GNG, and it was used in single study³³. Seven studies adapted the GNG to include food stimuli for go- and stop-trials. Trials may also vary according to the caloric content of the food item presented, where most often, the outcome of interest is the relative difference in an individual’s performance of inhibiting prepotent responses following high-calorie food trials as compared with low-calorie food trials or neutral trials. The GNG has demonstrated external validity²⁴ and reliability over time, i.e., r = .65, p < .001 for commission errors.³⁴ However, like most behavioral tasks, the ability of the GNG to discriminate functions is limited. The GNG is highly dependent upon complex attention and working memory, in addition to response inhibition^35,36.

Table 3.

Summary of Included Studies Investigating Compulsivity

				Compulsivity
First Author	BED Sample	BE/BED BMI Categorya	Control Group(s)	Construct	Task	Major Findings	Summaryb
Voon et al., 2015102,ef	BED (n = 31; n-TS)	Obesity 2	OB, NW	Habit Learning	Two-Step task	BED group used model-based, goal directed behavior less frequent than OB and NW groups.	BED > OB compulsivity
Duchesne et al., 201088	BED (n = 38; TS)	Obesity 2	OB	Set-shifting	Rule shift card task, WCST, TMT-B	BED group showed no significant differences in performance on the Rule shift card task, or TMT-B as compared with OB group. BED group showed significantly more perseverative errors on the WCST as compared with the OB group.	BED / OB compulsivity
Aloi et al., 201895	BED (n = 35, TS)	Obesity 2	OB, NW	Set-shifting	TMT-B	BED group showed significantly worse performance on the TMT-B as compared with the NW group. BED group did not show worse performance on TMT-B compared to OB group.	BED > NW compulsivity BED ≈ OB compulsivity
Galioto et al., 201296	LT or Cur. BED (n = 41; TS)	Obesity 3	OB	Set-shifting	TMT-B	BED group showed no significant differences on TMT-B compared to OB group.	BED ≈ OB compulsivity
Eneva et al., 201789	BED (n = 55; u-TS)	NW Obesity 1	OB, NW	Set-shifting	DKEFS trail making, NIH dimensional change card sort	There were no significant group differences on DKEFS trail making test and NIH dimensional card sort.	NW-BED ≈ OB-BED compulsivity NW-BED ≈ OB compulsivity NW-BED ≈ NW compulsivity OB-BED ≈ OB compulsivity OB-BED ≈ NW compulsivity
Aloi et al, 201592	BED (n = 45; TS)	Obesity 2	AN, NW	Set-shifting	TMT-B, WCST	BED group showed worse performance on TMT-B compared AN or NW groups. There were no significant differences on the WCST perseverative errors between the BED group and AN or NW groups.	BED / NW compulsivity BED / AN compulsivity
Banca et al., 201690	BED (n = 32; n-TS)	Obesity 2	OB, NW	Cognitive flexibility, Set-shifting	Probabilities learning task, IED	BED group showed significantly greater difficulties with reversal learning in “win” trials and significantly better performance in “loss” trials compared to OB group. BED group showed no differences on probabilistic learning task as compared with NW group. BED group not compared to the OB group on IED. BED group showed greater set-shifting difficulties compared to NW group.	BED / OB compulsivity BED > NW compulsivity

Notes. AN: individuals with anorexia; BE: Binge eating; BED: Binge eating disorder; BMI: body mass index; IED: intra-dimensional/extra-dimensional shift task; n-TS: non-treatment seeking; NW: individuals with normal weight; OB: individuals with obesity; TMT-B: trail making test B; TS: treatment seeking; WCST: Wisconsin card sorting test.

^a.BMI for BED group rounded to the nearest whole number and classified according to CDC guidelines of normal weight, obesity class 1, obesity class 2, and obesity class 3.

^b.> indicates greater levels of impulsivity or compulsivity, ≈ indicates similar levels of impulsivity or compulsivity, / indicates mixed findings

^c.BED cases versus controls assessed via self-report;

^d.sample included some subthreshold BED cases or BE symptom cases

^e.Sample included in two studies reporting different outcomes.

^f.NW group included individuals with BMI up to 27.

^g.Modified DSM-IV TR BED diagnosis used. Weekly binging of objectively large amounts without compensatory behaviorsh. Study included a delayed discounting task, but pairwise outcomes were not reported

Go/No-go: Behavioral Data.

Six studies investigated potential differences between individuals with BED and individuals with obesity and without BED (obesity-only) using GNGs (Tables 2 and 3)^37–42. In most studies, individuals with BED also had overweight or obesity. Two studies reported significant differences between individuals with BED and individuals with obesity-only^37,40. In one of the studies that found a significant difference³⁷, binge eating was assessed using the Binge Eating Scale⁴³. The Binge Eating Scale is a widely used self-report measure that demonstrates greater sensitivity than specificity⁴⁴. Four studies also using the GNG failed to find significant differences in commission errors when comparing individuals with BED and obesity-only^38,39,41,42. Overall, more studies are needed to determine if individuals with BED have higher or similar levels of rapid response impulsivity compared to individuals with obesity-only as measured by the GNG.

There were few differences between people with NW and those with BED on the GNG. Seven studies compared NW and BED groups (Tables 2 and 3)^{33,39–41,45–47}; five described no between-group differences^{33,41,45–47}, and the remaining two reported partial differences (i.e., no main effects)^39,40. One study found that individuals with BED made significantly more errors on the neutral version of the GNG as compared to the food version³⁹. In contrast, individuals with NW made significantly more errors on the food version of the task as compared to the neutral version.³⁹. Finally, one study reported that a combined group (BED and obesity-only) showed more GNG-related errors as compared with the group with NW^39,40. Pairwise comparisons were not reported when comparing individuals with BED exclusively and NW.

Two studies compared individuals with BED to other eating disorders⁴⁶ or food addiction (Table 4)⁴². Individuals with BED did not show significant differences on the GNG as compared with individuals with either anorexia nervosa or bulimia nervosa, though the study sample was quite small (BED, n=11)⁴⁶. Regarding food addiction, this is a controversial, albeit growing, area of research within the eating-disorder field. Some argue that food addiction represents a condition specifically marked by impulsivity and compulsivity related to food. Only one included study investigated food addiction⁴², and no significant differences in commission errors distinguished the four groups: obesity-only, food addiction only, BED-only, and BED plus food addiction.

Go/No-Go: Neuroimaging Data.

Two studies included neuroimaging procedures and the GNG^38,45. As inhibition is a primary variable of interest, data related to no-go trials will be summarized. In both studies, group differences in task performance were not observed, despite differential patterns of activation. Magnetoencephalography compared individuals with BED and obesity-only, and a significant three-way interaction revealed differential activity between groups that was influenced by the type of trial (go or no-go) and stimulus category (e.g., neutral versus food-based cues). Among individuals with BED, there was hypoactivation in the dorsolateral prefrontal cortex, suggesting under-recruitment of the areas necessary for successful inhibition³⁸. Hyperactivation in the left middle occipital gyrus for food stimuli, as compared with toy stimuli, among individuals with BED was also found. As the middle occipital gyrus is involved in object recognition, the finding may speculatively reflect greater visual processing of food versus toy stimuli in BED. The other study used functional magnetic imaging and the GNG to compare individuals with NW to individuals with NW plus subthreshold-BED⁴⁵. For no-go trials, individuals with NW plus subthreshold-BED showed lower activation in the right middle frontal gyrus, right precuneus, right caudate, right putamen, and left cerebellum as compared with those without binge symptoms. In contrast to this pattern of hypoactivation, individuals with BE symptoms showed increased activation in occipital regions. Most group differences were observed in both food and neutral no-go trials, though hypoactivation of the right caudate no longer showed significant differences in either type of stimulus blocks, and differences in the right middle frontal gyrus appeared specific to neutral stimuli, while the right putamen hypoactivation may be specific to food stimuli⁴⁵. Findings suggest that the subthreshold-BED group shows lower activation of executive control regions during no-go trials given the involvement of the middle frontal gyrus, a region that includes dorsolateral prefrontal cortex.

Stop-Signal Task: Task Description.

The SST is a computerized task during which an individual presses a key based on either the location (e.g., left or right) or type (e.g., square or circle) of stimuli for go-trials. In a minority of trials (sometimes explicitly operationalized as 25%), a participant must inhibit responses. The cue to inhibit the response is usually a tone, but sometimes an image is used as a cue. The delay between the go-stimuli and the indication to stop varies with varying adjusting procedures used across studies to achieve 50% correct inhibition. The primary outcome of interest is the average stop-signal-delay reaction time (SSDRT). The SSDRT is calculated by subtracting a participant’s average delay when achieving 50% accuracy from the average reaction time for go-trials. Like the GNG, the SST has been adapted to include food stimuli. While similarities exist regarding GNGs and SSTs, important differences exist with respect to performance and neural correlates²⁴. The GNG is best characterized as action restraint, while the SST is best understood as representing action cancellation. The reliability of the SST depends on methods used to calculate SSDRT with the most reliable using an average of all runs without excluding outliers⁴⁸. The SST has external validity perhaps greater superior discriminant validity than the GNG as scores are not conflated with other factors beyond attention/alertness⁴⁹.

Stop-Signal Task: Behavioral Data.

Eight studies investigated rapid response impulsivity using the SST (Tables 2 and 4)^45,50–56, and most investigated differences between individuals with BED and those with obesity-only, with mixed findings. Three studies found slower SSDRTs for individuals with BED as compared with obesity-only^51,52,56. One⁵¹ included both subthreshold/threshold patients with BED, and the findings did not support an effect for stimulus (food versus neutral stimuli), while group differences between individuals with BED and obesity-only were significant. A second study similarly found slower SSDRTs, and the effect for stimulus was significant, with post-hoc tests demonstrating that differences between individuals with BED and obesity-only were more pronounced for trials using food stimuli. One study found no differences⁵³ and two studies found that individuals with BED performed better on the SST than those with obesity-only^54,55.

Four studies compared individuals with BED to individuals with NW on the SST (Tables 2 and 4)^45,50,54,55. Superior performance by BED versus NW individuals was observed in some⁵⁴ but not other⁵⁵ studies. No significant differences on SSDRTs were observed in two additional studies^45,50. Only one small study (BED, n=11; Table 3) investigated potential differences in SST performance across multiple types of eating disorders. The study found no significant differences between individuals with BED, anorexia, or bulimia⁵⁰.

Stop Signal Task: Neuroimaging Data.

One study used the SST during functional magnetic resonance imaging to compare individuals with NW to individuals with NW and subthreshold-BED⁴⁵. As described above, the groups differed on brain activations but not task performance. During stop-trials, the binge-eating group showed greater activation of the left middle frontal gyrus and cerebellum, while the non-binge-eating group showed greater activation in the right postcentral gyrus during the stop trials. The relatively greater activation of the middle frontal gyrus suggests a possible need to exert more executive control during SST performance in the binge-eating group.

Other Rapid Response Tasks: Task Description.

Several other tasks were less widely utilized in BED studies to assess rapid response impulsivity, including the five-choice serial reaction time task (n =1; see Table 2)⁵⁷ and cued reaction time task (n =1; See Table 2)⁵⁰.

Other Rapid Response Tasks: Behavioral Data.

Findings from studies using the five-choice serial reaction time task and cued reaction time task found no significant differences in performance when comparing individuals with BED and obesity-only⁵⁷ and NW^50,57.

Other Rapid Response Tasks: Neuroimaging Data.

No neuroimaging studies met inclusion criteria.

Rapid Response Impulsivity Summary

Overall, data are equivocal as to whether individuals with BED have different levels of rapid response impulsivity as compared to the three groups of interest: obesity-only, NW, and other eating disorders. When comparing obesity-only and BED groups across tasks, five studies showed impaired performance^{37,40,51,52,56}, six showed nonsignificant differences^{38,39,41,42,53}, and two showed superior performance^54,55. When comparing individuals with BED to individuals with either NW or other eating disorders, most studies did not support higher levels of impulsivity^{33,41,45–47,50,55}, but one study did find significant differences³⁹.

Choice Impulsivity

Choice impulsivity is characterized by diminished tendencies to tolerate temporal delays. Choice impulsivity is higher in childhood and adolescence and diminishes in adulthood. Choice impulsivity is weakly correlated or uncorrelated with rapid-response impulsivity. Cross-species translational work indicates that rapid-response and choice impulsivity are underpinned by different neural substrates^58,59. The two types of impulsivity also differ in relation to psychopathology, with rapid-response impulsivity particularly associated with multiple psychiatric disorders⁶⁰. However, choice impulsivity is associated with psychopathologies, including bipolar disorder, borderline personality disorder, schizophrenia and, to a lesser extent, major depressive disorder⁶¹.

Choice Impulsivity: Task Descriptions.

Choice impulsivity is assessed via delay-discounting procedures, including the Monetary Choice Questionnaire (MCQ)⁶². In other studies, the stimulus is also money, but procedures vary in amounts of rewards or time delays. One study varied the stimuli to include food, a preferred sedentary activity, and massage in addition to money. Across tasks, the outcome of interest is the discounting rate, though this is expressed variably. For example, the MCQ sometimes expresses an overall discounting rate or the discounting rate can be expressed at various magnitudes of reward. The outcomes of all delay discounting tasks will be considered collectively with attention paid to stimulus type and comparison group. Most studies demonstrate good test-retest reliability for measures of choice impulsivity^63,64. Considerations of validity depend on types of validity under investigation. Considerable evidence supports the external validity of choice impulsivity measures based on abilities to distinguish psychiatric patients from controls^61,65 and moderate evidence supports predictive validity^71,66. Other data, however, indicate limitations, including confounding associations with mood³⁴ and stress⁶⁷. Limitations in the measurements should be considered when interpreting findings.

Choice Impulsivity: Behavioral Data.

Four studies investigated differences between individuals with BED and obesity-only on discounting tasks (Tables 2 and 4)^68–71, and three showed nonsignificant differences in discounting rates for money^68–70. One study showed a steeper discounting for individuals with BED for money⁷¹. Additionally, one of the studies showing nonsignificant differences in monetary discounting also tested food, a preferred sedentary activity and massage rewards and found that individuals with BED showed steeper discounting rates compared to individuals with obesity-only⁶⁹.

Seven studies have compared individuals with BED and NW (Tables 2 and 4)^{55,68–70,72–74}, with most finding that individuals with BED discount monetary^55,68,69,72 and non-monetary rewards⁶⁹ more steeply. For non-monetary rewards, differences between groups were largest when the stimulus was food, potentially providing evidence of stimulus-specific effects⁶⁹. Three studies did not find significant differences^70,73,74. However, two finding nonsignificant differences either used self-report measures to create BE groups⁷³ or individuals with BED and overweight (not obesity)⁷⁴. One additional, potential confounding variable was identified by Davis and colleagues⁶⁸, who showed that individuals with BED and NW no longer showed different discounting rates when controlling for education. Two other studies investigated effects of education and either found it unrelated to the primary delay-discounting outcome⁷⁴ or observed no between-group differences⁶⁹.

Two studies compared individuals with BED to other eating-disorder groups (Table 2)^72,74. A small study comparing individuals with BED to those with anorexia nervosa and bulimia nervosa on a delay-discounting task involving money found no differences⁷⁴. One study found that individuals with BED showed steeper monetary delay discounting compared to individuals with anorexia nervosa-restricting subtype, and comparable discounting compared to anorexia nervosa-binge purge subtype⁷². More research is needed to better understand how BED may or may not differ from other eating disorder groups with respect to choice impulsivity.

Choice Impulsivity: Neuroimaging Data.

One study compared patterns of activation among individuals with BED, obesity-only, and NW⁷⁰. The authors found significant differences between all groups. Most notably, individuals with BED showed hypoactivation in the ventral striatum regions during anticipatory phases of the task, but only in comparison of individuals with obesity-only. Most contrasts were not significant when compared to individuals with NW. During the outcome phases of the task, individuals with BED showed hypoactivation in prefrontal and insular regions, relative to both the obesity-only and NW groups⁷⁰. The findings of relatively diminished activation of the ventral striatum and insula during anticipation and outcome task phases resonate with BED-related findings on the monetary incentive delay task⁷⁰, with some of these patterns linked to poorer treatment outcomes⁷⁵.

Choice Impulsivity Summary

Overall, the data support some differences in choice impulsivity among individuals with BED, though the evidence varies. When comparing individuals with BED to individuals with obesity-only, only four studies met inclusion criteria^68–71. Thus, conclusions should be considered cautiously until further replication is possible. Three of four studies did not find increased choice impulsivity with respect to hypothetical monetary rewards^59,67,68. A single study tested other types of rewards and found evidence for increased impulsivity, with the largest effect sizes observed for food rewards⁶⁹.

Regarding comparisons to individuals with NW, the literature is comparatively larger with seven studies meeting inclusion criteria^{55,69,70,72–74,76}. Much like the literature on rapid-response impulsivity, the findings are mixed, though some evidence suggests that individuals with BED have higher choice impulsivity as compared to NW individuals. Specifically, four of seven studies found increased choice impulsivity among individuals with BED compared to individuals with NW^55,68,69,72. This finding is echoed in a recent meta-analysis⁶¹. Although the study had different inclusion criteria (i.e., required validated diagnostic instruments such as the SCID), it reported a moderate effect size (Hedge’s g =.34) for increased choice impulsivity among individuals with BED. There are not yet enough data specifically comparing individuals with BED to individuals with other eating disorders to draw definitive conclusions.

All studies meeting inclusion criteria used hypothetical rewards. However, research in other clinical groups has demonstrated that hypothetical rewards may be inadequate in evoking discounting, particularly moment-to-moment variability in discounting⁷⁷. Experiential approaches that use actual rewards, such as the Experiential Delay Task⁷⁷, may be more effective in evoking delay discounting. To our knowledge, no studies have investigated differences in delay discounting among individuals with BED using experiential procedures. Other aspects of task methodology are important in designing studies of choice impulsivity among people with BED. One study demonstrated group differences when rewards other than money were used⁶⁹. This is consistent with a broader literature that shows differences in discounting when consumable rewards (e.g., food or drugs of abuse) versus rewards that cannot be consumed (e.g., money) are used⁷⁸. Future research would benefit from combining experiential procedures with rewards that are both consumable and non-consumable in order to better understand delay discounting among individuals with BED.

Compulsivity

Set-shifting Compulsivity

Set-shifting refers to the degree of use of flexible versus concrete or rigid strategies to respond to changing rules⁷⁹. While flexible strategies typically result in behaviors that are better able to meet the demands of the goal or environment, rigid or concrete strategies often lead to perseverative behaviors⁸⁰. Set-shifting is typically assessed via two tasks: dimensional card-sorting tasks and Trail Making Test Part B (TMT-B).

Dimensional Card Sorting Tasks: Task Descriptions.

The Wisconsin Card Sorting Test (WCST) was one of the first to assess set-shifting and other tendencies using card sorting and rule changes. A primary outcome of interest for the WCST, as it relates to compulsivity, is perseverative errors. While the WCST is the most widely used card-sorting task in studies of BED, limitations in its psychometric properties should be considered. Some studies demonstrate poor reliability for perseverative errors^81,82, though evidence of good test-retest reliability also exists⁸³. Questions regarding construct validity of the WCST have also been raised. The WCST generally, and perseverative errors more specifically, have limited specificity with respect to particular cognitive functions and underlying neural substrates⁸⁴. In addition to the WCST, multiple related variants have been developed, including the intradimensional/extradimensional card sort, NIH’s dimensional card sort, and the Penn Conditional Exclusion Task. The intradimensional/extradimensional card sort distinguishes between problems with set-shifting due to distraction (intradimensional set-shift) versus difficulties due to problems shifting from a previously learned rule⁸⁵ (extradimensional set-shift). Compulsivity relates to shifts from a previously learned rule (i.e., extradimensional shifts). The Penn Conditional Exclusion task is similar to the WCST, but it includes more sorting dimensions and alternative forms, which was motivated to reduce potential practice effects⁸⁶. The NIH dimensional card-sorting task is based on a widely utilized task for children⁸⁷. This task differs in that sorting rules are explicit. The rule shift occurs both between blocks and within blocks. Unlike the previous set-shifting tasks, reaction time is considered within the scoring scheme. For individuals who get at least 80% of the trials right, both accuracy and reaction times are included in a composite measure that is scaled based on age group⁸⁷. A final dimensional card sorting task is the Rule Shift Cards task. In the task, individuals learn a single, simple rule: to say “Yes” in response to a red card and “No” in response to a black card. In the subsequent phase, a new rule is introduced, and the prior rule no longer applies. These rules are explicit, written, and available to the individuals throughout the task. A primary outcome related to compulsivity is the number of errors made. Scant investigation of the psychometric properties of card sorting tasks beyond the WCST exist except for a single study supporting convergent and divergent validity for the Penn Conditional Exclusion Test⁸⁶.

Dimensional Card Sorting Tasks: Behavioral Data.

Six studies explored differences between individuals with BED and obesity-only using dimensional card-sorting tasks (Tables 3 and 4)^{42,56,71,88,89}. Seven studies had samples that met inclusion criteria but one study⁹⁰ did not report comparisons between individuals with BED and obesity-only on the set-shifting task. Four studies did not find significant differences in perseverative errors on the tasks^56,71,89,91. Only one study reported a direct, significant mean difference in perseverative errors, with individuals with BED showing greater errors⁸⁸. Kollei and colleagues also found support for differences, but the analytic plan and outcomes were less direct. In particular, the authors used a multivariate dependent-variable approach and discriminant analyses. The authors found two discriminating factors, one of which they characterized as representing attention, reversal learning, and risk-taking in decision-making. This discriminant function distinguished individuals with BED, obesity-only, and NW. Reversal learning is relevant to set-shifting and provides some support for potential differences. However, the absence of pairwise comparisons in this study prohibit characterizing the evidence specific to set-shifting further.

Six studies examined differences between individuals with BED and NW using dimensional card-sorting tasks (Tables 3 and 4)^{33,41,46,89,90,92}. Four of six did not support differences in perseverative errors^33,46,89,92. Two studies reported greater set-shifting difficulties for individuals with BED as compared with NW^41,90. Some studies reported overall poorer performance on the WCST, although not on perseverative errors⁹²

Three studies described above also tested for differences between individuals with BED and other types of eating disorders or food addiction (Tables 3 and 4)^46,91,92. None found differences in perseverative errors when comparing individuals with BED to individuals with anorexia nervosa^46,92, bulimia nervosa⁴⁶ or food addiction⁹¹.

Dimensional Card-Sorting Tasks: Neuroimaging Data.

No studies met inclusion criteria.

Trail Making Test-Part B: Task Description.

The TMT-B is a longstanding behavioral task that requires an individual to set-shift between sequencing letters and numbers. Outcomes of interest including errors and completion time after controlling for processing and motor speed, which is accomplished by taking into the account the time need to complete trails that do not require set-shifting (TMT-A) Controlling for the effect of TMT-A has been shown to increase the construct validity of TMT-B and its association with other set-shifting measures⁹³. Assessments using alternate forms of the TMT-B to avoid practice effects indicate good test-retest reliability⁹⁴.

Trail Making Test-Part B: Behavioral Data.

Four studies investigating performance on the TMT-B in individuals with BED and individuals with obesity-only observed no between-group differences (Tables 3 and 4)^88,89,95,96. Four studies examined differences between individuals with BED and individuals with NW on the TMT-B (See Tables 3 and 4)^46,89,92,95. Two studies found that compared to individuals with NW, those with BED took more time to complete TMT-B, made more errors, and demonstrated larger differences in completion times of TMT-A and TMT-B^92,95. In contrast, two studies did not find differences in errors made or completion time^46,89. One study showing null findings included a small sample (BED n = 10)⁴⁷, which may be underpowered. Most studies included in this review (N = 32) did not report a priori power analyses, with the exception of two studies^39,41.

Two studies investigated potential differences among individuals with BED, anorexia nervosa or bulimia nervosa, with differing findings (Tables 3 and 4)^46,92. The earlier study⁹² reported that compared to individuals with anorexia nervosa, individuals with BED showed longer TMT-B completion times and larger completion-time differences between TMT-A and TMT-B. Individuals with BED did not commit more errors than individuals with anorexia nervosa⁹². The subsequent small study⁴⁶ did not observe significant differences between individuals with BED, anorexia nervosa, or bulimia nervosa.

Trail Making Test-Part B: Neuroimaging Data.

No studies met inclusion criteria.

Set-Shifting Compulsivity Summary

The data did not broadly support significant deficits in set-shifting for individuals with BED compared to individuals with obesity-only, NW, or other eating disorders, though more research is needed. When focusing on comparisons with individuals with obesity-only, ten results from seven studies were included^{41,42,56,71,88,89,95,96}, and eight results did not find evidence of significant differences^{42,56,71,89,95,96}, though there appears to be relevant differences across tasks. The WCST data included mixed findings for BED and obesity-only comparisons, with two studies^41,88 finding some evidence of poorer performance for individuals with BED and four studies^56,71,89,91 not finding evidence of group differences. In contrast, the TMT-B results were unambiguous, with four studies failing to find a difference between BED and obesity-only groups.

Data related to differences in set-shifting between individuals with BED and NW were also mixed, with most supporting no differences, particularly on the WCST. The more consistent WCST findings in the NW group differs from the more consistent TMT-B findings in the obesity-only group. Sample characteristics are likely relevant when interpreting TMT-B findings comparing individuals with BED to those with NW. Specifically, two studies that did not find differences either used small samples⁴⁶ or samples that were atypical with respect to weight status⁸⁹. Further study is warranted.

Studies comparing individuals with BED to those with other types of eating disorders are also needed. Across tasks, three studies compared individuals with BED to other eating-disorder groups^42,46,92, and only the TMT-B performance in a single study was significantly different between groups⁹².

Future directions related to the role of set-shifting compulsivity in BED should consider task demands. Earlier reviews of eating disorders (BED was not included) argued that the potency of set-shifting requirements might differ across tasks, and significantly fewer differences were observed between individuals with anorexia and bulimia when TMT-B was used as compared with other set-shifting tasks⁹⁷. This was also partially supported by meta-analytic work of other “impulsive-compulsive-spectrum” disorders, though conflicting findings exist here too. Specifically, meta-analysis of studies of gambling disorder and healthy controls found significantly smaller differences when comparing TMT-B results to other related behavioral tasks such as the WCST⁹⁸. In contrast, a meta-analysis of individuals with obsessive-compulsive disorder revealed large deficits on the TMT-B as compared with healthy controls⁹⁹. As such, more research is needed to disambiguate findings. Although large sample sizes are needed, factor analytic approaches to behavioral outcomes, such as perseverative errors on the WSCT and completion time on the TMT-B, could help to identify if task differences between groups are meaningful.

Cognitive Flexibility Compulsivity

Cognitive flexibility and set-shifting are highly similar. In line with definitions presented by Fineberg and colleagues and Kakoschke and colleagues, cognitive flexibility was differentiated from set-shifting by considering the role of negative contingencies (e.g., punishments). Specifically, tasks that included negative contingencies and a measure of response to that contingency were considered measures of cognitive flexibility. Cognitive flexibility tasks are similar to tests of risky decision-making. Cognitive flexibility tasks assess rigid adherence to rules or strategies (perseveration) even after one experiences loss associated with use of that strategy. Tests of risky decision-making instead aim to identify choices under risk and reward manipulations, usually in the form of gambling or betting. Only one study met inclusion criteria for studying cognitive flexibility (Table 3)⁸⁸. Given the paucity of available evidence related to cognitive flexibility and habit learning in BED, a summary of included studies will consider cognitive flexibility and habit learning in aggregate.

Probabilistic Reversal Learning task: Task Description.

The probabilistic reversal learning task involves rule acquisition, rule reversal, and changes in magnitude⁹⁰. In short, there are two phases, one in which the rule is learned and one in which the rule is reversed. The individual chooses between a stimulus pair, with varying probabilities of winning and losing. There are three magnitude conditions, with 30 trials each, across the two phases, for a total of 180 trials. Individuals are told that one choice within each pair was more likely to be associated with winning money or not losing money. They are also told that at some point that the rule might change, and accordingly, the symbol associated with winning or not losing could change. Psychometric analyses of probabilistic reversal learning tasks among people with or without eating disorders are sparse. However, a small event-related fMRI study (n = 10) assessing test-retest reliability revealed a high concordance for activation patterns during task completion¹⁰⁰.

Probabilistic Reversal Learning task: Behavioral Data.

There were no main effect group differences between individuals with BED and obesity-only in cognitive flexibility⁹⁰. However, there was a significant interaction effect; individuals with BED took more trials to learn in the reversal phase specifically for reward trials, which suggests that despite punishment (in the form of monetary losses), individuals with BED were likely to continue trying a previously successful strategy in pursuit of a reward. When attempting to avoid losses, individuals with BED required significantly fewer trials as compared with individuals with obesity-only. Individuals with BED showed no significant differences on the probabilistic learning task as compared with individuals with NW.

Probabilistic Reversal Learning task: Neuroimaging Data.

No studies met inclusion criteria.

Habit-Learning Compulsivity

Habit-learning refers to a lack of sensitivity to goals or outcomes of action⁷. Habit-learning has been proposed as a parallel system to goal-directed behavioral systems, and compulsivity may in part arise from an imbalance in these systems. To date, most task-based research has focused on testing goal-directed behavior (e.g., set-shifting and cognitive flexibility), and accordingly habit-learning data both in BED and other psychiatric disorders are limited¹⁰¹. One study of habit-learning met inclusion criteria (Table 3)¹⁰².

Two-Step Task: Task Description.

The Two-Step Task involves attempts to differentiate between model-based and model-free learning processes. Model-based approaches are goal-directed and involve an individual dynamically updating what is known about the probability of winning and losing. In contrast, model-free learning is the result of more habitual or compulsive behavior and information is not dynamically updated to direct behavior^102,103. The primary outcome is termed the ‘w parameter’, which conceptually demonstrates reliance on model-free (habit strategies).

Two-Step Task: Behavioral Data.

Individuals with BED showed significantly greater reliance on model-free approaches as compared with both individuals with NW and obesity-only¹⁰².

Two-Step Task: Neuroimaging Data.

The same study investigated anatomical correlates of compulsivity using MRI¹⁰². However, the neuroimaging did not include the sample with BED and as such the findings are not reviewed here.

Cognitive Flexibility and Habit Learning Compulsivity Summary

There are insufficient data to comment on roles for cognitive flexibility and habit learning in BED.

Investigation of Confounds

We aimed to investigate the role of five potential confounds: subthreshold/threshold BED status, treatment-seeking status of the BED group, BMI of the BED group, use of food stimuli, and biological sex/gender. When considering the direction and significance of group differences, the data did not reveal any clear systematic relationships for any of these five factors. To aid in future research, Supplement 1 includes detailed information about the representation of confounds in the included studies. Supplement 1 also includes any studies that statistically tested for the effects of these confounds, though this was very rare.

Conclusions

Based on the outlined inclusion criteria, 32 studies investigating impulsive and compulsive aspects of BED using tasks with or without neuroimaging techniques have been published to date. The evidence base for impulsivity was comparatively larger than that for compulsivity, but complexities in interpretation exist across domains. Arguably the most consistent data suggest that choice impulsivity is increased in individuals with BED as compared with individuals with NW, but the increased delay discounting may not be as pronounced or may be altogether absent when comparing individuals with BED to individuals with obesity-only or other eating disorders. Evidence related to rapid response impulsivity was mixed, with most studies suggesting similar rapid-response impulsivity among individuals with and without BED. Regarding compulsivity, only studies of set-shifting allowed for systemic review. Heterogeneity in tasks prohibits definitive conclusions, but most data did not indicate significant differences in BED samples with regard to set-shifting. For the remaining domains of compulsivity—cognitive flexibility and habit learning—only two studies met inclusion criteria, and as such, specific conclusions cannot be drawn currently.

When considering discrepant findings for rapid-response impulsivity, choice impulsivity, and set-shifting, several points are worth noting. The current qualitative review of potential confounds did not reveal obvious relationships. Nonetheless, they could be contributing, and quantitative methods may be more appropriate for identifying this effect. Additionally, other confounds not considered here could also be contributing. Education, depression, and meal timing have all been implicated in the literature as relevant to BED presentation. Depression may be a particularly salient consideration. Approximately 66% of people with BED have a comorbid depressive disorder¹⁴, and people with depressive conditions have deficits in inhibitory control¹⁰⁴. Future research should disentangle possible effects due to depressive disorders and to BED.

Multiple studies failed to support hypotheses that individuals with BED have increased impulsivity and compulsivity. However, the null hypothesis that individuals with BED have the same levels of impulsivity and compulsivity as the comparison groups of interest was not directly examined. This conclusion would only be supported by alternative analytic approaches such as equivalency analyses, which are notably absent from the behavioral literature on BED (and much neuropsychiatric research generally). For traditional hypothesis testing, adequate power is critical. Of the 32 studies identified, only two studies included data on a priori power analyses^39,41. Demonstration of adequately powered studies is important for future studies.

In addition to these analytic considerations, several cross-cutting design issues and future directions are apparent from the current review. First, future research should consider using neuroimaging methods to add important information to studies of impulsivity and compulsivity among people with BED. Relatively few studies used neuroimaging, and when used, data generally supported group differences that were not evident in behavioral task performance^38,45,70. Although interpreting different patterns of activation in the absence of behavioral differences is complex, more studies would help.

Second, more research is needed to understand how specific sample characteristics and methods influence outcomes. The current review attempted to consider how subthreshold/threshold BED status, treatment-seeking status of the BED group, BMI of the BED group, use of food stimuli, and biological sex/gender were related to outcomes. In large part, the data did not reveal clear, unequivocal patterns. One potential exception included studies of individuals with subthreshold BED. Outcomes with subthreshold patients rarely diverged notably from studies with full threshold cases. Given the emphasis on adequately powering studies, consideration of this type of approach could increase the feasibility of research related to impulsivity and compulsivity in BED. The use of food stimuli warrants further attention. Several studies found significant interactions between type of case (e.g., BED, obesity-only, NW) and stimulus type, where people with BED had faster reaction times or greater inhibitory difficulties when exposed to food stimuli as compared with neutral stimuli^38,40,47,52. However, multiple studies failed to find effects for stimulus type among BED cases^45,46,54,105. Research with other types of psychiatric conditions have demonstrated greater specific effects for stimulus type. For example, studies of alcohol use disorder demonstrate a stronger effect when alcohol stimuli are used in rapid-response tasks as compared with neutral cues¹⁰⁶. Future studies should consider use of food stimuli in adequately powered designs to test interaction effects. Finally, the role of sex/gender requires further consideration. Beginning in January 2016, the National Institutes of Health required consideration of sex/gender in matters of research design, analysis, and reporting. Of the unique samples, about half (n = 14) of the studies included female-only participants. Only one study investigated and reported on the effects of sex/gender specifically, and no interactions between sex/gender and binge-eating status were found on primary outcomes⁷³. Within mixed-sex/gender samples, the samples remained largely female (average percent females = 75%). As such, the results presented are likely most consistent with impulsivity and compulsivity among female BED patients. Future studies are needed to examine impulsivity and compulsivity in large samples with more male participants with BED.

Study conclusions should be considered within the context of several limitations. First, while theoretical work was used to operationalize the literature search and reporting of findings, precise boundaries between tasks and behavioral domains may be debated, which may obscure potentially relevant data. Second, the current review is qualitative in nature. The qualitative approach did not reveal clear explanatory variables associated with the heterogeneity in findings. A quantitative approach (e.g. meta-analysis, including assessment of the quality of evidence and potential for bias) and/or consideration of additional relevant covariates represent important future directions. Additionally, while preclinical literature was outside of the scope of the current review, examination of preclinical data would be helpful in elucidating conclusions. A recent narrative review of rodent models of binge eating and emotional eating found evidence of a transition from impulsive food choice to compulsive food choice¹⁰⁷. See [98] for a discussion on the pre-clinical evidence of impulsivity and compulsivity in food choice. See [99] for a discussion of pre-clinical and clinical studies that have furthered our understanding of the neural circuits of binge-eating and related eating pathology. Notwithstanding these important limitations, several strengths are worth noting. This review represents an update to several prior reviews both systematic and nonsystematic in nature. A prior nonsystematic review²² covered the widest range of topics, but substantial research has been conducted in the time since this review was published; approximately 12 studies met inclusion criteria for the current study that were not published at the time of the earlier review²². Finally, the systematic nature of the search with associated screening increases confidence that all available studies published on impulsivity and compulsivity in BED were considered here.

Highlights.

• Binge-eating disorder (BED) is associated with impulsivity and compulsivity

• Increased choice impulsivity may be particularly relevant to BED

• Behavioral data related to impulsivity/compulsivity are mixed

• Neuroimaging data related to impulsivity/compulsivity are largely lacking