Subtyping Patients with Binge-Eating Disorder by Dietary Restraint and Negative Affect: Characteristics and Treatment Outcome

Abstract

Objective:

This study examined patients with binge-eating disorder (BED) subtyped by dietary restraint (DR) and the negative affect (NA) dimension of depression, anxiety, hostility, and self-esteem, comparing clinical features and outcome of evidence-based psychological treatments. It was hypothesized that individuals with DR and high NA (DR-HNA) would have lower functioning and poorer immediate and long-term BED treatment outcomes compared to those with DR and low NA (DR-LNA).

Method:

Cluster analysis was conducted (n = 159) as a secondary analysis of data from a randomized group cognitive-behavioral therapy and interpersonal psychotherapy trial of women and men with BED.

Results:

At pretreatment, participants with DR-HNA were higher in eating disorder and general psychopathology, emotional eating, and poor social adjustment. At post-treatment, differences in binge days between cluster groups were small and statistically nonsignificant, and differences in rates of binge-eating abstinence were only marginally significant. However, by 1-year follow-up, patients with DR-HNA had a greater increase in binge days, less abstinence from binge eating, a higher risk for relapse, and were significantly less likely to be in BED remission than those with DR-LNA.

Conclusions:

DR-HNA served as a predictor of poorer maintenance of BED therapeutic improvement. Future directions to better sustain outcomes among patients with DR-HNA include developing a brief measure to assess for the full construct of NA prior to and throughout treatment, enhancing BED interventions by focusing more on NA, and augmenting treatment dose.

Binge-eating disorder (BED) is characterized by recurrent and persistent binge episodes in the absence of regular use of compensatory behaviors (American Psychiatric Association [APA], 2013). Subtyping can facilitate explanations of differences in treatment outcome that would be masked when all cases are considered together. For example, Loeb and colleagues (2000) examined psychopathology subtypes within a trial of guided and unguided cognitive-behavioral therapy (CBT) self-help interventions among women who met subthreshold/threshold diagnostic criteria for BED or bulimia nervosa, yielding a two cluster solution. Participants with higher psychopathology had poorer post-treatment binge-eating outcomes. Wilson et al. (2010) examined interpersonal psychotherapy (IPT), guided self-help based on CBT, and behavioral weight loss in the treatment of BED. Patients with high negative affect (NA), as measured by the Beck Depression Inventory, showed less improvement in binge-eating outcomes. Moderator findings revealed that individuals with low self-esteem and a high level of eating disorder psychopathology fared best in IPT.

Subtyping along the construct of NA, defined as a general trait of negative emotionality that includes the aversive mood states of anxiety, sadness, hostility, and low self-concept (Watson & Clark, 1984), appears to have important clinical predictive value for patients with BED. Stice and colleagues (2001) examined pure dietary restraint (DR) and mixed DR-NA subtypes among women with BED. Those in the mixed subtype had a higher frequency of objective binge episodes; significantly greater eating, weight, and shape concerns; as well as higher levels of psychiatric and social maladjustment. Grilo, Masheb, and Wilson (2001) replicated Stice et al.’s (2001) results and found that those in the mixed subtype had significantly higher eating, weight, and shape concerns, as well as greater body dissatisfaction and higher levels of impulsivity. Additional secondary analyses have found NA combined with DR to be a predictor but not a moderator of BED treatment outcome among individuals treated with CBT or behavioral weight-loss guided self-help (Masheb & Grilo, 2008). Grilo, Masheb, & Crosby (2012) studied patients with BED treated with CBT or fluoxetine, finding subtyping by overvaluation of shape/weight the most salient predictor and moderator of outcomes.

The current study sought to examine immediate and long-term treatment outcome within DR-NA subtypes, as well as how DR-NA subtypes differ on a broader set of clinical features and more comprehensive construct of NA than previously studied, among a large and diverse sample of patients with BED receiving CBT or IPT. It was hypothesized that individuals in the DR - high NA (DR-HNA) subtype would have higher levels of dysfunctional clinical characteristics at pretreatment than would those in the DR - low NA (DR-LNA) subtype. Across treatments, it was hypothesized that the DR-HNA group would have poorer immediate and long-term outcomes compared to the DR-LNA group. Because IPT targets the management of affect within an interpersonal context, a moderation effect was hypothesized such that the DR-HNA group would do better in IPT, whereas the DR-LNA group would do better in CBT across time periods.

Method

This study used data from a comparative psychotherapy trial for BED (n = 162; Wilfley et al., 2002); 3 participants were missing relevant pretreatment data and were therefore excluded from the present study.

Participants

Participants were women and men who met Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5 [APA, 2013]) criteria for BED, as well as the more stringent DSM-IV (APA, 1994) criterion of at least two binge-eating episodes a week for a period of at least 6 months. Inclusion criteria were being 18–65 years old; not pregnant, in psychotherapy or a weight-loss program, or taking weight-affecting or psychotropic medications; having a body mass index (BMI) of 27–48 kg/m²; and no psychosis, active substance dependence or suicidality. The study received Institutional Review Board approval, and participants gave written consent to participate.

Treatment

Participants were stratified by sex and randomized to 20 weekly 90-minute sessions of group CBT or IPT. Blind ratings confirmed that in this trial they were distinct and delivered with high treatment integrity (Wilfley et al., 2002).

Measures

At pretreatment, BMI and demographics were assessed, as well as a weight cycling item adapted from the Brownell Weight-Cycling Questionnaire (Foreyt et al., 1995).

Trained assessors administered the interviewer-based Eating Disorder Examination (EDE), 12^th Edition (Fairburn & Cooper, 1993), at pretreatment, post-treatment, and 1-year follow-up. The EDE assessed binge-eating frequency (i.e., days within the past 28 days in which at least one objective binge episode occurred) and eating disorder pathology (i.e., four subscales: Restraint, and Eating, Shape, and Weight Concern). The EDE’s Restraint subscale was aggregated with the NA variables listed below for clustering purposes.

The Emotional Eating Scale (EES; Arnow et al., 1995) measured an individual’s urge to cope with NA by eating. The Weight Efficacy Lifestyle Questionnaire (WEL; Clark et al., 1991) assessed confidence in being able to successfully resist the desire to eat across situations.

In line with Watson and Clark’s (1984) definition of NA, depression, anxiety, hostility, and self-esteem were aggregated for clustering purposes. The Rosenberg Self-Esteem Scale (RSES; Rosenberg, 1979) and the Depression, Anxiety, and Hostility subscales of the Symptom Checklist-90-Revised (SCL-90-R; Derogatis, 1977), assessed NA constructs.

Current social functioning/adjustment was measured using the Social Adjustment Scale (SAS; Weissman & Bothwell, 1976). Trained interviewers used the Structured Clinical Interview for the DSM (SCID; Spitzer et al., 1990a, 1990b) to assess comorbid current and lifetime psychopathology.

Data Analytic Plan

For all analyses, family-wise alpha rate was set at .05. First, an SPSS k-means cluster analysis (Jain, 2010) was performed on the standardized [per cluster analysis requirements (Afifi & Clark, 1996)] scores of DR (i.e., the Restraint subscale of the EDE) and NA (i.e., the RSES and the Depression, Anxiety, and Hostility subscales of the SCL-90-R). The k-means cluster analysis is an iterative method that specifies clusters of homogenous groups based on Euclidean distance and assigns each participant to one of k clusters in order to minimize the distances of cluster members from their corresponding cluster centers (Aldenderfer & Blashfield, 1984). We chose cluster analysis for the purpose of replication.

Second, to examine potential demographic differences in the subtypes, t-tests and chi-square analyses were conducted. To test whether the subtypes differed in pretreatment binge-eating frequency and various related clinical features, analyses of variance (ANOVAs) and multivariate ANOVAs (MANOVAs) were used. Chi-square analyses examined differences by subtype in general psychopathology and history of weight cycling. In cases where a cell had an n < 5, Fisher’s exact tests were conducted.

Third, to assess whether subtype predicted and/or moderated treatment outcome (binge days), an intent-to-treat 3 (time: pretreatment, post-treatment, or 1-year follow-up) × 2 (subtype) × 2 (treatment) repeated-measures ANOVA and follow-up contrasts were conducted, with missing data points (3 at post-treatment, and 17 at 1-year follow-up) replaced with pretreatment values.

Finally, chi-square analyses examined outcome, by subtype, in terms of remission from full-syndrome BED (i.e., binge-eating frequency below a DSM-5 diagnostic level of ≥4 days during the past month) and abstinence (i.e., no binge episodes during the past month). Similar analyses were used to examine relapse – i.e., an increase in binge eating, for those in remission at post-treatment, to above remission threshold at 1-year follow-up.

The pattern of results for continuous and categorical binge-eating outcomes using completer outcome analyses (excluding participants with missing data) was similar to those for intent-to-treat analyses, so only intent-to-treat results are reported.

Results

Cluster Analysis

From the final k-means cluster analysis with the DR-NA variables and k = 2 clusters specified, 39 (24.5%) participants were categorized within the DR-HNA subtype and 120 (75.5%) were categorized within the DR-LNA subtype. Distribution of the subtypes was equal across treatments, χ²(1, 159) = 0.36, p = .550, ϕ = .047. There were no significant differences by subtype in sex, race/ethnicity, marital status, age, or BMI (all ps > .199), but those within the DR-HNA subtype had about 1.5 fewer years of education (mean 14.5 versus 16.0; p = .001) compared to those within the DR-LNA subtype.

Table 1 shows a manipulation check of the cluster analysis; participants in the DR-LNA subtype had lower mood and higher self-esteem compared to those in the DR-HNA subtype (all ps < .001), but did not significantly differ in DR (p = .090). As indicated by their unstandardized mean NA scores (see Table 1), participants within the DR-HNA subtype experienced a moderate amount of distress due to symptoms of depression, and a small amount due to anxiety or hostility. In contrast, participants within the DR-LNA subtype experienced minimal distress related to depressive symptoms, and almost no distress due to anxiety or hostility. In both subtypes, participants had a mild amount of DR. Notably, analyses were also run with NA alone (that is, excluding DR from the cluster subtypes), revealing a similar pattern of findings to results presented below with both DR and NA included in the cluster.

Table 1.

Unstandardized Mean Negative Affect and Dietary Restraint Scores by Subtype

Negative Affect	DR-HNA (n = 39)	DR-LNA (n = 120)
Depressiona	2.3	0.9
Anxietya	1.3	0.4
Hostilitya	1.3	0.4
Self-Esteemb	22.4	28.6
Dietary Restraintc	2.3	1.9

Note. DR-HNA = Dietary Restraint - High Negative Affect, DR-LNA = Dietary Restraint - Low Negative Affect.

^aAssessed by the Symptom Checklist-90-Revised (SCL-90-R), on a 5-point rating scale from 0 to 4, where higher scores indicate more symptom distress

^bAssessed by the Rosenberg Self-Esteem Scale (RSE), on a 4-point rating scale from 1 to 4 (range for total score is 10 to 40), where higher scores indicate higher levels of self-esteem

^cAssessed by the Eating Disorder Examination (EDE), on a 7-point rating scale from 0 to 6, where higher scores indicate higher levels of dietary restraint

Pretreatment Clinical Features by Subtype

When examining clinical features by subtype (see Table 2), the relation between DR-NA subtype and pretreatment binge-eating frequency was nonsignificant, F(1, 157) = 3.01, p = .085, η² = .019. Multivariate analyses did indicate that there was a significant relation between DR-NA subtype and pretreatment eating disorder pathology, p < .001 (with significant differences on all subscales). In terms of associated features of BED (see Table 2), participants with DR-HNA had greater emotional eating, poorer social adjustment, and were more likely to have engaged in weight cycling, but subtype was not associated with confidence in controlling eating.

Table 2.

Pretreatment Eating Disorder Pathology and Associated Features of BED, by Subtype

	DR-HNA (n = 39)	DR-LNA (n = 120)
Variable	M (SD)	M (SD)	F (p)	η2
ED Pathology (Multivariate effect)a			6.51 (< .001)	.112
Eating Concern	3.0 (1.3)	2.1 (1.3)	13.30 (< .001)	.078
Weight Concern	3.8 (0.9)	3.1 (1.1)	12.11 (.001)	.072
Shape Concern	4.3 (0.7)	3.6 (1.0)	15.80 (< .001)	.091
Emotional Eating (Multivariate effect)b			3.95 (.010)	.075
Depression	3.1 (0.6)	2.7 (0.8)	5.70 (.018)	.037
Anxiety	2.4 (0.8)	2.1 (0.9)	3.05 (.083)	.020
Anger/Frustration	2.9 (0.7)	2.4 (0.9)	11.36 (.001)	−.071
Confidence in Controlling Eatingc	72.7 (31.2)	76.5 (29.7)	0.43 (.512)	.003
Social Adjustmentd	2.3 (0.4)	2.0 (0.5)	9.51 (.002)	.058
	n (%)	n (%)	χ2(p)	φ
History of Weight Cyclinge	36 (92.3)	82 (70.1)	9.29 (.002)	−.224

Note. DR-HNA = Dietary Restraint - High Negative Affect, DR-LNA = Dietary Restraint - Low Negative Affect, ED = Eating Disorder.

^aAssessed by the Eating Disorder Examination (EDE); subscales are on a 7-point scale from 0 to 6, with higher numbers indicating greater pathology

^bAssessed by the Emotional Eating Scale (EES); subscales are on a 5-point scale from 0 to 4, with higher scores indicating greater emotional eating

^cAssessed by the Weight Efficacy Lifestyle Questionnaire (WEL), on a 10-point scale from 0 to 9 (range for total score is 0 to 180), with higher scores indicating greater confidence

^dAssessed by the Social Adjustment Scale (SAS), on a 5-point scale from 1 to 5, with higher scores indicating greater maladjustment

^eAssessed by a dichotomous item adapted from the Brownell Weight-Cycling Questionnaire

In terms of overall current general psychopathology, including personality disorders, and lifetime general psychopathology, all were significantly more prevalent among participants with DR-HNA compared to those with DR-LNA (see Table 3). These findings held especially for current mood disorders (specifically major depression) and personality disorders (specifically cluster B). Finally, participants in the DR-HNA subtype were 1.2 times more likely to have at least one comorbid SCID diagnosis than those in the DR-LNA subtype, χ²(1, 159) = 5.81, p = .016, ϕ = −.177.

Table 3.

General Psychopathology by Subtype

	DR-HNA (n = 39)		DR-LNA (n = 120)
Diagnosis	%	n	%	n	χ 2	p	φ	RR
Current Diagnoses
Any Current Diagnosis other than a Personality Disorder	53.8	21	25.8	31	10.05	.002	−.257	2.1
Mood Disorders	41.0	16	15.8	19	9.95	.002	−.262	2.6
Major Depression	33.3	13	10.0	12	10.68	.001	−.276	3.3
Anxiety Disorders	20.5	8	10.0	12	2.70	.099	−.136	2.1
Substance Use Disorders	5.1	2	3.3	4	0.24	.636	−.041	1.5
Any Personality Disorder	56.4	22	30.8	37	8.04	.005	−.228	1.8
Cluster A	10.3	4	4.2	5	1.81	.224	−.113	2.5
Cluster B	25.6	10	8.3	10	7.06	.010	−.225	3.1
Borderline	17.9	7	6.7	8	3.87	.055	−.166	2.7
Cluster C	35.9	14	23.3	28	2.29	.130	−.123	1.5
Lifetime Diagnoses
Any Lifetime Diagnosis other than a Personality Disorder	89.7	35	72.5	87	5.57	.018	−.176	1.2
Mood Disorders	74.4	29	56.7	68	4.03	.045	−.156	1.3
Major Depression	71.8	28	53.3	64	4.25	.039	−.161	1.3
Anxiety Disorders	41.0	16	25.8	31	3.14	.076	−.143	1.6
Substance Use Disorders	30.8	12	34.2	41	0.15	.694	.031	0.9

Note. DR-HNA = Dietary Restraint - High Negative Affect, DR-LNA = Dietary Restraint - Low Negative Affect, RR = Relative Risk, calculated as the percent of the DR-HNA group with that diagnosis divided by the percent of the DR-LNA group with that diagnosis. All diagnoses were made using the Structured Clinical Interview for the DSM (SCID).

Outcome Analyses

For the ANOVA testing the impact of subtype on binge-eating outcome (see Table 4), the 3-way interaction among time, subtype, and treatment was nonsignificant, F(2, 154) = 0.84, p = .436, η² = .011, indicating no moderation effects of NA cluster. However, consistent with the general predictor hypothesis, there was a significant 2-way time by subtype interaction, whereby across treatments, improvement in binge days was lower in DR-HNA than DR-LNA participants (p = .011). Follow-up contrasts on this interaction indicated no significant subtype differences from pretreatment to post-treatment, but from post-treatment to 1-year follow-up there was a greater increase in binge eating for the DR-HNA than the DR-LNA subtype (p = .003; see Figure 1).

Table 4.

Mean Binge Days by Subtype Within and Across Treatment

	Overall (N = 159)	DR-HNA (n = 39)	DR-LNA (n = 120)
Treatment Type	M (SD)	M (SD)	M (SD)
Pretreatment
Overall	16.8 (7.1)	18.5 (6.7)	16.2 (7.1)
CBT	17.2 (7.0)	20.3 (6.8)	16.0 (6.7)
IPT	16.4 (7.2)	16.3 (6.1)	16.4 (7.5)
Post-Treatment
Overall	1.2 (3.7)	1.8 (4.5)a	1.0 (3.5)a
CBT	1.1 (3.9)	2.1 (5.7)	0.8 (3.0)
IPT	1.2 (3.6)	1.4 (2.7)	1.2 (3.8)
1-Year Follow-up
Overall	3.2 (6.3)	6.2 (8.5)a	2.3 (5.1)a
CBT	3.9 (7.1)	6.9 (9.6)	2.8 (5.6)
IPT	2.5 (5.4)	5.3 (7.3)	1.7 (4.5)

Note. Data are based on intent-to-treat analysis. DR-HNA = Dietary Restraint - High Negative Affect, DR-LNA = Dietary Restraint - Low Negative Affect, CBT = Cognitive-behavioral therapy; IPT = Interpersonal psychotherapy. Binge days are the number of days, out of the last 28 days, on which the participant had at least one objective binge episode, as assessed by the Eating Disorder Examination (EDE).

^aPost-hoc analyses revealed significant time by subtype effect in change from post-treatment to 1-year follow-up, across treatments, with a greater increase in binge eating for the DR-HNA than the DR-LNA subtype, F(1, 155) = 9.22, p = .003, η² = .056.

Figure 1. Days Binged in Past Month by Subtype, as Assessed by the Eating Disorder Examination (EDE).

Note. Data are from intent-to-treat analysis. There was a significant time by subtype effect in change from post-treatment to 1-year follow-up, across treatments, with a greater increase in binge eating for the DR-HNA than the DR-LNA subtype (p = .003).

Similarly, chi-square analyses (see Table 5) indicated that at post-treatment, rates of BED remission were not significantly different between participants within the DR-HNA and DR-LNA subtypes (84.6% and 91.7%, respectively; p = .225), but by 1-year follow-up, significantly fewer participants with DR-HNA than DR-LNA were in remission (56.4% and 82.5%, respectively; p = .001). Further, fewer participants within the DR-HNA than DR-LNA subtypes were abstinent at both post-treatment (at a trend level: 64.1% versus 80.0%; p = .050) and 1-year follow-up (significant: 48.7% versus 66.7%; p = .047). Finally, a significant difference by subtype was found for relapse at 1-year follow-up (p = .005), with the relative risk of relapse for participants within the DR-HNA subtype 2.5 times that of those within the DR-LNA subtype (i.e., 39.4% and 15.5% relapse, respectively).

Table 5.

Remission, Abstinence, and Relapse by Subtype

	DR-HNA		DR-LNA
Outcome	n	%	n	%	χ2	p	φ	RR
Post-Treatment	(n = 39)		(n = 120)
Remission	33	84.6	110	91.7	1.49	.225	.101	0.9
Abstinence	25	64.1	96	80.0	3.86	.050	.160	0.8
1-Year Follow-up	(n = 39)		(n = 120)
Remission	22	56.4	99	82.5	10.16	.001	.263	0.7
Abstinence	19	48.7	80	66.7	3.95	.047	.159	0.7
From Post-Treatment to 1-Year Follow-upa	(n = 33)		(n = 110)
Relapse	13	39.4	17	15.5	7.95	.005	−.248	2.5

Note. DR-HNA = Dietary Restraint - High Negative Affect, DR-LNA = Dietary Restraint - Low Negative Affect, RR = Relative Risk, calculated as the percent of the DR-HNA group at that time point with that outcome divided by the percent of the DR-LNA group with that outcome.

Data are from intent-to-treat analyses. Remission is defined as binge eating at a frequency below a DSM-5 diagnostic level of ≥4 days during the past month. Abstinence is defined as no binge-eating episodes during the past month. Relapse (only out of those who were in remission at post-treatment) is defined as an increase in binge eating to above the remission threshold from post-treatment to 1-year follow-up. All outcomes were assessed by the Eating Disorder Examination (EDE).

Note that the number of participants who relapse is not simply the difference between the numbers of participants in remission at post-treatment versus 1-year follow-up. This is because some participants who had not remitted at post-treatment had remitted at 1-year follow-up.

^aSample size represents those who were remitted at post-treatment.

Discussion

The current study utilized a cluster analysis by DR-NA to investigate pretreatment clinical characteristics and short- and long-term treatment outcome in patients with BED. As predicted, classification into a DR-HNA subtype predicted higher levels of nearly all forms of dysfunctional clinical characteristics, including more severe eating disorder pathology, higher rates of comorbid psychopathology, and more emotional eating, prior to treatment. In addition, participants within the DR-HNA subtype had a marginally lower abstinence rate at post-treatment and a significantly less robust long-term treatment outcome, regardless of treatment received. History of weight cycling was more common in the DR-HNA subtype than in the DR-LNA subtype. Participants across subtypes had similar levels of confidence in controlling eating across situations, though post-hoc analyses did reveal a trend-level difference by subtype for the Negative Emotions subscale (i.e., those with DR-HNA had lower confidence about being able to successfully resist the desire to eat when experiencing negative emotions). Finally, and not surprising given the relation between self-esteem and social functioning, those with DR-HNA experienced more disturbance in social adjustment.

Participants within the DR-HNA subtype had higher rates of current mood disorders (specifically major depression) and cluster B personality disorders. Although significantly more participants within the DR-HNA than the DR-LNA subtype had some form of comorbid psychopathology, a post-hoc analysis of binge days showed that the DR-NA subtype was associated with BED treatment outcome independent of presence or absence of comorbid general psychopathology. Although characteristics are shared between the diagnostic features of these disorders and the DR-NA subtyping dimension, NA is a broader, trait-level, distinct phenomenon not driven by presence of psychopathology alone.

A primary hypothesis of the current study was that participants within the DR-HNA subtype would have poorer short- and long-term outcomes than would those within the DR-LNA subtype. Post-treatment differences in binge days were relatively small and were nonsignificant. However, differences in rates of abstinence by DR-NA were evident at a trend level at post-treatment and significantly at 1-year follow-up. DR-HNA predicted poorer maintenance of treatment gains in the long term (i.e., from post-treatment to 1-year follow-up). These treatment results were robust, as findings of poorer outcomes among participants within the DR-HNA subtype held when examining remission to below a DSM-5 diagnostic level of binge eating; in particular, given the potency of both interventions, rates of BED remission were not significantly different between DR-HNA and DR-LNA at post-treatment. However, by 1-year follow-up, significantly fewer participants with DR-HNA, compared to DR-LNA, were in remission. Results also held when either intent-to-treat or completer analyses were used. Thus, DR-HNA emerged as a significant barrier to complete abstinence and maintenance of treatment gains. We had also predicted a moderator effect which was not found, as participants with DR-HNA did not do better in IPT, nor DR-LNA in CBT. Yet, this was not entirely surprising, as both therapies were adapted from treatments for depression.

The present study adds to the literature by revealing that DR-NA may serve as a predictor of maintenance of BED therapeutic improvements in both CBT and IPT. Although previous research has studied subtyping and treatment outcome in patients with BED, this is the first to examine subtyping using the more complete definition of the NA mood state, as well as long-term outcome one year after treatment cessation. The pattern of findings was similar when DR was excluded from the clustering. Given both subtypes had similar levels of DR, the future directions of these findings apply more specifically to targeting NA. The significantly smaller rate of abstinence at 1-year follow-up and higher rate of relapse among participants with DR-HNA calls for a brief, easily scored NA assessment tool to determine into which subtype a patient falls and to track NA throughout treatment. Findings also suggest the utility of future research on the specific needs of patients with HNA, such as a preliminary treatment phase solely targeting NA, and booster sessions to augment BED treatments.

Public Health Significance Statement:

Binge-eating disorder (BED) is often associated with significant impairment in physical and mental health as well as quality of life, and following evidence-based psychological treatment some individuals with BED may fail to achieve abstinence or maintain their treatment gains in the long term. This research extends the subtyping literature by suggesting that individuals with BED and dietary restraint who have high levels of negative affect at pretreatment have less abstinence and poorer maintenance of treatment gains in the long term, highlighting the need for a screening tool and interventions tailored to reducing negative affect within the context of dietary restraint.