Subtypes of Exercise are Differentially Associated with Baseline Eating Disorder Pathology & Treatment Outcome Among Individuals with Bulimia Nervosa

Abstract

Objective.

Individuals with bulimia nervosa (BN) engage in both maladaptive (i.e., compulsive and/or compensatory) and adaptive exercise (e.g., for enjoyment). No research has examined whether those who engage in adaptive, compulsive, and/or compensatory exercise exhibit differences in BN pathology or treatment outcome compared to those not engaging in exercise, limiting intervention efficacy.

Method.

We examined associations of baseline exercise engagement with baseline and post-treatment BN pathology among 106 treatment-seeking adults (M_age=37.4, SD_age = 12.95, 87.74% female, 68.87% White) enrolled across four clinical trials of outpatient Enhanced Cognitive Behavioral Therapy (CBT-E) for BN (range: 12–16 sessions). ANCOVAs examined associations between baseline exercise type and baseline/post-treatment global eating pathology, dietary restraint, loss of control (LOC) eating, and purging frequency.

Results.

Those engaging in only adaptive exercise reported lower global eating pathology compared to those engaging in compulsive-only exercise (Est = −1.493, p = .014, M_diff = −0.97) while those engaging in baseline compulsive exercise reported less LOC eating compared to those not engaging in exercise (Est = −22.42, p = .012, M_diff = −12.50). Baseline engagement in compulsive-only exercise was associated with lower post-treatment global eating pathology compared to baseline engagement in no exercise (Est = −0.856, p = .023, M_diff = −0.64) and both compulsive and compensatory exercise (Est = 0.895, p = .026, M_diff = −1.08).

Discussion.

Those engaging in compulsive, compensatory, adaptive, and no exercise exhibit different patterns and severity of BN pathology. Future research is needed to position treatments to intervene on maladaptive, while still promoting adaptive, exercise.

Introduction

Exercise is typically an adaptive behavior across populations, with an exception being individuals with eating disorders, including bulimia nervosa (BN). Over 50% of treatment-seeking patients with BN reported engaging in maladaptive exercise (Monell, Levallius, Mantilla, & Birgegård, 2018) which can be compulsive and/or compensatory in nature. Compulsive exercise is performed according to rigid rules to reduce distress (Fairburn, Cooper, & O’Connor, 2014; Meyer, Taranis, Goodwin, & Haycraft, 2011), whereas compensatory exercise is performed in response to food intake with or without a sense of loss-of-control (Davis, Holland, & Keel, 2014; Fairburn et al., 2014; LePage, Crowther, Harrington, & Engler, 2008).

Maladaptive exercise engagement has been cross-sectionally associated with a range of eating disorder (ED) symptoms within samples comprising male and female participants between the average ages of 18 to 34. Compulsive exercise engagement is positively associated with dietary restraint among inpatient (Dalle Grave et al., 2008) and outpatient samples (Monell et al., 2018), and within non-clinical samples (Holland et al., 2014; Taranis & Meyer, 2011). Compensatory exercise engagement is associated with increased dietary restraint and a greater frequency of both subjective and objective binge eating episodes across several non-clinical samples (Garner, Davis-Becker, & Fischer, 2014; Kerrigan, Lydecker, & Grilo, 2019; Martin, Schell, Srivastav, & Racine, 2020). However, no research to date has directly compared associations between compulsive and/or compensatory exercise engagement and BN symptoms. Consequently, little is known about how individuals who engage in these types of exercise compare to those not engaging in any exercise at treatment initiation.

Most longitudinal research within primarily female samples between the average ages of 21 to 27 indicates that baseline compulsive exercise engagement is associated with worse treatment outcomes, but has not examined outcomes specific to individuals with BN and the impact of compulsive exercise on ED treatment outcomes is not consistent across studies. For example, in both inpatient and outpatient transdiagnostic ED samples, higher compulsive exercise engagement during treatment and cognitive features of compulsive exercise (i.e., rigidity, persistence) post-treatment have been associated with worse treatment outcome (Danielsen, Rø, Romild, & Bjørnelv, 2016; Monell et al., 2018). Conversely, another study of female inpatients with AN, BN, and OSFED found that the presence of compulsive exercise at baseline did not have any specific effect on treatment outcome (Dalle Grave, Calugi, & Marchesini, 2008). No research has isolated the impact of compensatory exercise engagement at baseline on BN treatment outcomes. Further, given the frequent endorsement of both compulsive and compensatory exercise among individuals with BN (either within the same episode or within an individual across episodes), we must understand whether there is an additive effect of engagement in both compulsive and compensatory exercise on BN pathology. Thus, there is a need to understand the specific contributions of baseline compulsive and compensatory exercise engagement, and their combination, on ED treatment outcomes among individuals with BN.

While many individuals with BN report engagement in maladaptive exercise, some individuals with BN report engagement in adaptive exercise (i.e., exercise that is neither compulsive nor compensatory; Lampe, Forman, Juarascio, & Manasse, 2021; Lampe, Trainor, et al., 2021; Lampe, Wons, Taylor, Juarascio, & Manasse, 2022; Srivastava, Lampe, Wons, Juarascio, & Manasse, 2022; Wons, Lampe, Patarinski, Schaumberg, & Juarascio, 2023). An individual who engages in adaptive exercise may do so to spend time with friends, for enjoyment, or for general health benefits (Cash, Now, & Grant, 1994; Lampe, Trainor, et al., 2021). Increased engagement in adaptive exercise episodes over the course of Cognitive Behavior Therapy (CBT) shows promise in reducing binge eating and purging episodes within primarily female outpatient samples (Lampe, Forman, et al., 2021; Wons et al., 2023). It is important to better understand whether baseline adaptive exercise is associated with reductions in specific ED symptoms across treatment.

Exercise engagement patterns in BN are highly heterogeneous. Thus, more research is needed to better understand how engagement in each type of exercise at baseline (i.e., compulsive-only, compensatory-only, both compulsive and compensatory, adaptive-only exercise, and no exercise) is associated with baseline BN pathology and treatment outcomes. Thus, the current study examined whether individuals’ baseline engagement in adaptive-only, compulsive-only, compensatory-only, both compulsive and compensatory, or no exercise was 1) related to baseline ED pathology (i.e., dietary restraint, frequency of loss-of-control [LOC] eating, and global eating disorder pathology), and 2) related to eating disorder treatment outcome (i.e., post-treatment levels of dietary restraint, loss-of-control eating, and global eating disorder pathology) among 106 adults with BN.

Associations with baseline BN pathology.

We hypothesized that those engaging in adaptive-only exercise at baseline would have the lowest levels of baseline dietary restraint and global eating pathology, followed by those not engaging in exercise, those engaging in compensatory-only exercise, those engaging in compulsive-only exercise, and those engaging in both compulsive and compensatory exercise. We also hypothesized that those engaging in adaptive-only exercise at baseline would have the lowest frequency of LOC eating, followed by those not engaging in exercise, those engaging in compulsive-only exercise, those engaging in compensatory-only exercise, and those engaging in both compulsive and compensatory exercise. Given the lack of research examining associations between exercise and purging in EDs, an exploratory analysis examined whether baseline exercise engagement was associated with baseline purging frequency.

Associations with treatment outcome.

We hypothesized that those engaging in adaptive-only exercise at baseline would have the greatest reductions in dietary restraint and global eating pathology at post-treatment followed by those not engaging in exercise, those engaging in compensatory-only exercise, those engaging in compulsive-only exercise, and those engaging in both compulsive and compensatory exercise. Contrary to hypothesized baseline associations and given the theoretical link between LOC eating and compensatory exercise, we hypothesized that those engaging in adaptive-only exercise at baseline would have the greatest reductions in LOC eating at post-treatment followed by those not engaging in exercise, those engaging in compensatory-only exercise, those engaging in both compensatory and compulsive exercise, and those engaging in compulsive-only exercise. An exploratory analysis examined whether baseline exercise engagement was associated with post-treatment purging frequency.

Methods

Participants & Procedures

Data used in the current analyses come from the baseline and post-treatment assessments of 106 subjects diagnosed with BN enrolled across 4 clinical trials of enhanced Cognitive Behavioral Therapy for binge eating (CBT-E; Juarascio et al., 2021; Juarascio, Presseller, Srivastava, Manasse, & Forman, 2022; Juarascio, Srivastava, et al., 2022; Manasse et al., 2020). Participants enrolled in the parent studies were recruited from the Philadelphia metro area for in-person treatment studies, and nationally for remote treatment studies. In all studies, participants were included if they reported an average of at least one objective binge episode per week over the past 3 months, were > 18 years old, and had a BMI > 17.5. Participants were excluded across all studies if they had severe psychopathology (e.g., severe suicidality, significant substance dependence, assessed at baseline via self-report and Beck Depression Inventory) or a severe medical condition preventing safe participation, were diagnosed with anorexia nervosa, had acute suicide risk, were pursuing concurrent ED or weight loss treatment, had previously received CBT-E, were unable to speak, read, or write English, or reported current/planned pregnancy or nursing. Participants were included in the current analyses if they reported ≥ 12 objective binge episodes and ≥ 12 compensatory behaviors (e.g., purging, laxative use, etc.) over the previous 3 months at baseline (i.e., at least 1 binge episode and compensatory behavior per week over the past 3 months). The 50-minute Enhanced Cognitive Behavioral Therapy (CBT-E) for eating disorders sessions included content on self-monitoring, regular eating, reducing dietary restraint, alternative activities, and reducing overvaluation of shape and weight. More information about each parent study and specific inclusion/exclusion criteria can be found in Table 1.

Table 1.

Study information and study specific inclusion/exclusion criteria across parent clinical trials of CBT-E.

Clinical Trial	Study Description	Number of Sessions	Study Specific Inclusion Criteria	Study Specific Exclusion Criteria
Project COMPASS : Juarascio, Felonis et al., 2022	Project COMPASS recruited 256 individuals who were assigned to one of 16 treatment combinations to identify the optimal combination of four acceptance-based treatment components adjunctive to CBT-E (mindful awareness, distress tolerance, emotion modulation, values-based decision making). The current analyses included 5 individuals randomized to receive base CBT-E only with no acceptance-based components.	16	None	None
The Acquire Project: Juarascio, Presseller et al., 2022	The Acquire Project recruited 56 participants (n = 53 included in the current analyses) who completed electronic self-monitoring in a smartphone app and received just-in-time-adaptive interventions (JITAIs) or no JITAIs alongside 16 sessions of CBT-E.	16	≥ 12 compensatory behaviors in the past 3 months Willingness to use smartphone application to track eating episodes during treatment	History of bariatric surgery
SenseSupport Study: Juarascio, Srivastava et al., 2022	The SenseSupport Study recruited 30 adults (n = 21 included in the current analyses) who received CBT-E augmented with JITAIs delivered via smartphone app addressing episodes of fasting > 5 hours detected by a continuous glucose monitor.	12	≥ 12 episodes of fasting for 5+ waking hours in the last 4 weeks	Not on stable psychiatric medication for at least 1 month Diagnosis of diabetes Taking medication known to impact insulin or glucose levels History of bariatric surgery
Project REBOOT: Manasse, Lampe et al., 2020	Project REBOOT recruited 63 participants (n = 27 included in the current analyses) who received a 12-week course of CBT-E for binge eating, while completing active inhibitory control trainings (ICT) or control “sham” trainings daily for the first 4 weeks and weekly in weeks 5–12.	12	None	High baseline inhibitory control Binge episodes consists primarily of uncooked fruits and vegetables, as per a modified Food Frequency Questionnaire (FFQ) Prescribed stimulant medication

Measures

All participants completed assessments at baseline and post-treatment with a trained interviewer which assessed exercise behavior and eating disorder symptoms via the Eating Disorders Examination version 17.0 (EDE; Fairburn et al., 2014). The EDE is a widely used and well validated semi-structured interview that assesses eating disorder symptoms and behaviors. Participants were oriented to the events of the past 3-months using a calendar to help prompt participants to recall specific episodes of binge eating, purging, and exercise engagement. The diagnosis of bulimia nervosa was made based on behaviors reported during the EDE interview. Baseline dietary restraint was measured using the restraint subscale of the EDE interview and global eating pathology was measured using the EDE global score. LOC eating frequency was calculated as the sum of all objective and subjective binge episodes reported on the EDE over the past 28 days. Total LOC episodes was chosen as the outcome of interest given evidence that objective and subjective binge episodes are associated with similar levels of distress and clinical impairment (Colles, Dixon, & O’Brien, 2008; Goldschmidt et al., 2016; Niego, Pratt, & Agras, 1997). Purging frequency was calculated as the sum of all purging-type compensatory behaviors (i.e., self-induced vomiting, taking laxatives, and taking diuretics) over the past 28 days. A calendar of the last 28 days was shared with participants during the EDE to help orient participants to this specific time period and recall episodes and behaviors occurring within this time frame. In addition to the standard EDE questions assessing maladaptive exercise, baseline exercise type was determined using a modified EDE item assessing compulsive exercise engagement (see supplementary materials). Participants were first asked to provide examples of the types of exercise they have engaged in (e.g., walking, yoga, weight training) and the number of total times they exercised over the past month. Assessment of compensatory exercise frequency used the pre-existing EDE item, “Over the past four weeks have you exercised as a means of controlling your weight, altering your shape or amount of fat, or burning off calories?” Assessment of compulsive exercise used the item, “When you have exercised over the past 4 weeks, have you felt driven or compelled to exercise?” Adaptive exercise frequency was calculated by subtracting instances of maladaptive exercise from the total exercise frequency. Baseline exercise type was coded such that 0 = no exercise engagement reported over the past month, 1 = only compensatory exercise reported, 2 = only compulsive exercise reported, 3 = both compulsive and compensatory exercise reported, 4 = only adaptive exercise engagement reported. Exercise groups were defined as mutually exclusive (i.e., a person endorsing any level of maladaptive exercise was categorized as such, even if they reported some adaptive episodes).

Statistical Analyses

All data analysis was conducted in R version 4.1.3 (R Core Team, 2022). The criteria p < 0.05 was used to make inferences of statistical significance across all models tested. The current analyses were pre-registered (https://osf.io/a4bqm/). Upon reviewer suggestion, we deviated from the pre-registration by using analysis of covariance (ANCOVA) models for both cross-sectional and longitudinal aims, which allowed us to also obtain effect estimates for those not engaging in any exercise at baseline.

Power analysis.

Power analyses using G-power for linear regression models (which would require the most observations to be fully powered in the context of the planned analyses) indicated a necessary sample size of 105 participants to reach 80% power to detect a medium effect (d = 0.2). Given this, the current analyses are fully powered with 106 participants.

Missing Data.

Data were assumed missing at random (i.e., baseline exercise type did not significantly predict post-treatment missingness). We used the mice package (Van Buuren, Groothuis-Oudshoorn, & Robitzsch, 2019) to perform multiple imputation on the missing data.

Cross-sectional and Longitudinal Models.

We used ANCOVA models to examine whether exercise type was associated with baseline and post-treatment ED pathology (i.e., EDE global score, dietary restraint, LOC eating, and purging frequency). All models covaried for parent study enrollment, age, sex, and body mass index, and longitudinal models also covaried for baseline eating pathology. For any significant omnibus results, Tukey’s HSD post-hoc tests were used to identify group differences.

Results

Participant Demographics & Missing Data

Figure 1 depicts the number of participants included in the current analyses from each of the parent trials. Information on participant demographics by each individual clinical trial can be found in Table 2. The mean age of the full sample was 37.35 (SD = 12.95) years. The full sample included primarily individuals who identified as female (87.74%), White (68.87%), Black (8.49%), Latinx (8.49%), Asian (6.60%), more than one race (6.60%), other (4.72%), unknown (1.89%), and American Indian/Alaska Native (0.94%). Across the whole sample, the average baseline EDE global score was 3.23 (SD = 0.95), and average baseline EDE restraint score was 2.82 (SD = 1.31). Participants endorsed 27.09 LOC eating episodes on average (SD = 20.72) over the past month at baseline. Twenty-three participants were categorized as non-exercisers, 13 as adaptive-only exercisers, 13 as compensatory-only exercisers, 31 as compulsive-only exercisers, and 26 as both compulsive and compensatory exercisers. Average BMI was 30.37 (SD = 7.64) in the non-exercise group, 30.15 (SD = 9.20) in the compensatory-only group, 29.04 (SD = 6.08) in the compulsive-only group, 32.41 (SD = 6.53) in the both compulsive and compensatory group, and 26.55 (SD = 6.67) in the adaptive-only group. Average BMI did not significantly differ by exercise group (F(1) = 0.227, p = .634). Only 14.74% of post-treatment observations were missing within the data set and baseline exercise type was not significantly associated with missingness.

Figure 1.

Overall number of subjects included and excluded across parent studies.

Table 2.

Participant descriptives by study and overall sample.

	COMPASS	Acquire	REBOOT	SenseSupport	Full Sample
Number participants	5	53	27	21	106
Age: M (SD)	30.7 (9.3)	38.3 (14.0)	37.2 (12.7)	36.0 (11.1)	37.5 (12.9)
Body Mass Index (BMI): M (SD)	32.63 (10.19)	29.55 (6.80)	31.35 (8.29)	29.52 (5.20)	30.13 (7.01)
Female: N (%)	1 (20.0)	46 (86.8)	26 (96.3)	20 (95.2)	93 (87.7)
Race: N (%)
White	2 (40.0)	33 (62.3)	21 (77.8)	17 (81.0)	73 (68.9)
Black	1 (20.0)	5 (9.4)	2 (7.4)	1 (4.8)	9 (8.5)
Asian	0 (0.0)	4 (7.5)	2 (7.4)	1 (4.8)	7 (6.6)
American Indian/Alaska Native	0 (0.0)	1 (1.9)	0 (0.0)	0 (0.0)	1 (0.9)
Hawaiian/Pacific Islander	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
More than one	0 (0.0)	4 (7.5)	1 (3.7)	2 (9.5)	7 (6.6)
Unknown/Prefer not to say	0 (0.0)	1 (1.9)	1 (3.7)	0 (0.0)	2 (1.9)
Other	0 (0.0)	5 (9.4)	0 (0.0)	0 (0.0)	5 (4.7)
Hispanic/Latinx: N(%)	0 (0.0)	4 (7.5)	4 (14.8)	1 (4.8)	9 (8.5)
Baseline Eating Disorder Behaviors (past 3-months): M (SD)
LOC Eating Frequency	20.0 (13.1)	26.1 (23.8)	31.6 (21.3)	25.6 (10.7)	27.1 (20.7)
EDE Global Score	2.47 (0.41)	3.27 (0.97)	3.23 (1.05)	3.31 (0.83)	3.23 (0.95)
EDE Restraint Score	1.84 (1.77)	3.02 (1.39)	2.55 (1.16)	2.92 (1.04)	2.82 (1.31)
Purging Frequency	2.67 (5.78)	15.33 (26.26)	11.69 (21.91)	11.95 (18.99)	13.14 (23.21)
Exercise Engagement (days in past month): M (SD)
Any Exercise	11.27 (6.00)	9.17 (9.10)	13.44 (10.59)	13.76 (8.90)	11.25 (9.48)
Maladaptive Exercise	10.00 (14.05)	6.98 (7.82)	8.63 (9.72)	10.00 (11.04)	8.14 (9.28)

Baseline Comparisons

Baseline EDE global score.

Those engaging in adaptive-only exercise at baseline had significantly lower baseline EDE global scores compared to those engaging in compensatory-only exercise (Est = −1.493, S.E. = 0.458, p = .014, estimated M_diff = −0.97); Table 3, Figure 2). Those engaging in compulsive-only, compensatory-only, both compulsive and compensatory, and no exercise at baseline had similar baseline EDE global scores (see Table 3, Figure 2).

Table 3.

Associations of exercise type with baseline eating pathology.

Baseline EDE Global Score
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	9.531	1	11.870	< 0.001*	--
Exercise type	9.981	4	3.114	.021*	0.14
Study	0.089	1	0.111	.741	0.01
Age	0.238	1	0.297	.588	< 0.001
Sex	0.033	1	0.041	.841	< 0.001
Body Mass Index	1.344	1	1.678	.200	0.03
Pairwise Comparisons – Baseline EDE Global Score
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	0.757	0.391	1.937	.303
	Compulsive	−0.079	0.301	−0.261	.999
	Both	0.193	0.359	0.538	.983
	Adaptive	−0.736	0.431	−1.706	.430
Compensatory	Compulsive	−0.836	0.384	−2.178	.196
	Both	−0.564	0.389	−1.451	.591
	Adaptive	−1.493	0.458	−3.258	.014*
Compulsive	Both	0.272	0.347	0.783	.933
	Adaptive	−0.657	0.420	−1.564	.519
Both	Adaptive	−0.929	0.363	−2.563	.087
Baseline EDE Restraint Score
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	7.616	1	4.513	.038*	--
Exercise type	8.324	4	1.233	.306	0.07
Study	0.413	1	0.245	.622	< 0.001
Age	2.855	1	1.692	.198	0.03
Sex	0.263	1	0.156	.694	< 0.001
Body Mass Index	0.638	1	0.378	.541	< 0.001
Pairwise Comparisons – Baseline EDE Restraint Score
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	0.285	0.567	0.503	.986
	Compulsive	0.106	0.438	0.243	.999
	Both	−0.061	0.521	−0.117	.999
	Adaptive	−1.033	0.626	−1.649	.465
Compensatory	Compulsive	−0.179	0.557	−0.322	.998
	Both	−0.346	0.564	−0.614	.971
	Adaptive	−1.318	0.665	−1.981	.281
Compulsive	Both	−0.167	0.504	−0.332	.997
	Adaptive	−1.139	0.610	−1.866	.339
Both	Adaptive	−0.971	0.526	−1.846	.350
Baseline LOC Frequency
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	707.1	1	1.776	.187	--
Exercise type	5002.3	4	3.140	.020*	0.15
Study	17.9	1	0.045	.833	< 0.001
Age	166.7	1	0.419	.520	< 0.001
Sex	3.4	1	0.009	.927	< 0.001
Body Mass Index	864.4	1	2.170	.146	0.03
Pairwise Comparisons – Baseline LOC Frequency
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	−13.818	8.174	−1.586	.505
	Compulsive	−22.420	6.721	−3.336	.012*
	Both	−15.128	8.007	−1.889	.327
	Adaptive	−5.762	9.617	−0.599	.974
Compensatory	Compulsive	−8.602	8.553	−1.006	.848
	Both	−1.310	8.665	−0.151	.999
	Adaptive	8.056	10.216	0.789	.931
Compulsive	Both	7.929	7.743	0.942	.876
	Adaptive	16.658	9.374	1.777	.389
Both	Adaptive	9.366	8.082	1.159	.769
Baseline Purging Frequency
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	316	1	0.058	.810	--
Exercise type	22,438	4	1.037	.395	0.07
Study	1,638	1	0.303	.584	< 0.001
Age	317	1	0.059	.810	< 0.001
Sex	9,247	1	1.710	.196	0.03
Body Mass Index	363	1	0.067	.796	< 0.001
Pairwise Comparisons – Baseline Purging Frequency
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	−19.330	32.114	−0.602	.973
	Compulsive	−49.389	24.769	−1.994	.275
	Both	−25.502	29.509	−0.864	.906
	Adaptive	−15.991	35.443	−0.451	.991
Compensatory	Compulsive	−30.058	31.521	−0.954	.871
	Both	−6.172	31.933	−0.193	.999
	Adaptive	3.340	37.649	0.089	.999
Compulsive	Both	23.886	28.537	0.837	.916
	Adaptive	33.398	34.546	0.967	.866
Both	Adaptive	9.512	29.786	0.319	.998

Figure 2.

Mean baseline EDE global score, dietary restraint score, LOC frequency, and purging frequency by exercise type reported at baseline. *p < 0.05

Baseline dietary restraint.

Those engaging in no exercise, compensatory-only, compulsive-only, adaptive-only, and both compulsive and compensatory exercise at baseline did not significantly differ in baseline levels of dietary restraint (see Table 3, Figure 2).

Baseline LOC eating.

Those engaging in compulsive-only exercise at baseline endorsed significantly lower LOC eating frequency compared to those engaging in no exercise at baseline (Est = −22.420, S.E. = 6.721, p = .012, estimated M_diff = −12.50; Table 3, Figure 2). Those engaging in no exercise, compensatory-only, adaptive-only, and both compulsive and compensatory exercise did not significantly differ in baseline LOC eating frequency (Table 3).

Baseline purging frequency.

Those engaging in no exercise, compensatory-only, compulsive-only, adaptive-only, and both compulsive and compensatory exercise at baseline did not significantly differ in baseline frequency of purging (see Table 3, Figure 2).

Treatment Outcomes

Post-treatment EDE global score.

Those engaging in compulsive-only exercise at baseline reported significantly lower post-treatment EDE global scores compared to those not engaging in any exercise (Est = −0.856, S.E. = 0.290, p = .032, estimated M_diff = −0.64; Table 4, Figure 3). Those engaging in compulsive-only exercise at baseline also reported significantly lower post-treatment EDE global scores compared to those engaging in both compulsive and compensatory exercise (Est = 0.895, S.E. = 0.295, p = .026, estimated M_diff = −1.08; Table 4, Figure 3).

Table 4.

Associations of exercise type with post-treatment eating pathology.

Post-treatment EDE Global Score
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	0.728	1	0.8139	.370	--
Exercise type	13.545	4	3.787	.007*	0.16
Study	0.068	1	0.076	.784	< 0.001
Age	1.557	1	1.742	.191	0.02
Sex	1.271	1	1.421	.237	0.03
Body Mass Index	0.022	1	0.025	.875	< 0.001
Baseline EDE Global	22.211	1	24.840	< 0.001*	0.24
Pairwise Comparisons – Post-treatment EDE Global Score
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	−0.732	0.351	−2.084	.234
	Compulsive	−0.856	0.290	−2.951	.032*
	Both	0.039	0.307	0.128	.999
	Adaptive	−0.780	0.377	−2.068	.241
Compensatory	Compulsive	−0.124	0.332	−0.374	.996
	Both	0.771	0.350	2.201	.187
	Adaptive	−0.049	0.415	−0.117	1.000
Compulsive	Both	0.895	0.295	3.034	.026*
	Adaptive	0.076	0.351	0.215	.999
Both	Adaptive	−0.819	0.381	−2.149	.207
Post-treatment EDE Restraint Score
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	0.298	1	0.221	.639	--
Exercise type	9.743	4	1.811	.135	0.08
Study	0.963	1	0.716	.400	0.03
Age	2.670	1	1.984	.163	0.03
Sex	0.929	1	0.690	.408	0.01
Body Mass Index	0.368	1	0.274	.602	< 0.001
Baseline EDE Restraint	16.694	1	12.409	.007*	0.14
Pairwise Comparisons – Post-treatment EDE Restraint Score
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	0.054	0.430	0.128	1.000
	Compulsive	−0.602	0.356	−1.691	.441
	Both	0.209	0.377	0.556	.981
	Adaptive	−0.629	0.459	−1.369	.645
Compensatory	Compulsive	−0.657	0.409	−1.608	.492
	Both	0.155	0.428	0.361	.996
	Adaptive	−0.683	0.499	−1.371	.644
Compulsive	Both	0.812	0.362	2.240	.173
	Adaptive	−0.026	0.428	−0.061	1.000
Both	Adaptive	−0.838	0.464	−1.806	.374
Post-treatment LOC Frequency
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	0.90	1	0.016	.899	--
Exercise type	386.90	4	1.752	.147	0.10
Study	64.60	1	1.169	.283	0.01
Age	27.60	1	0.499	.482	< 0.001
Sex	119.90	1	2.171	.145	0.02
Body Mass Index	7.90	1	0.142	.707	< 0.001
Baseline LOC frequency	68.20	1	1.236	.270	0.02
Pairwise Comparisons – Post-treatment LOC Frequency
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	−3.129	2.771	−1.129	.788
	Compulsive	−3.488	2.382	−1.464	.585
	Both	0.191	2.464	0.078	1.000
	Adaptive	−6.247	2.913	−2.144	.209
Compensatory	Compulsive	−0.359	2.638	−0.136	1.000
	Both	3.320	2.750	1.207	.744
	Adaptive	−3.118	3.208	−0.972	.865
Compulsive	Both	3.679	2.324	1.583	.509
	Adaptive	−2.759	2.815	−0.980	.862
Both	Adaptive	−6.438	3.018	−2.133	.214
Post-treatment Purging Frequency
Parameter	Sum Sq	df	F	p	Partial η 2
Intercept	67.00	1	0.187	.667	--
Exercise type	1015.90	4	0.708	.589	0.05
Study	1.10	1	0.003	.957	< 0.001
Age	257.10	1	0.716	.400	0.02
Sex	102.40	1	0.285	.595	0.02
Body Mass Index	10.50	1	0.029	.864	0.02
Baseline purging freq.	7936.90	1	22.112	< 0.001*	0.22
Pairwise Comparisons – Post-treatment Purging Frequency
Pairwise Comparison		Est	S.E.	t	p
No Exercise	Compensatory	−6.743	7.048	−0.957	.872
	Compulsive	−5.715	5.900	−0.969	.866
	Both	−3.323	6.175	−1.609	.983
	Adaptive	−11.972	7.438	0.154	.492
Compensatory	Compulsive	1.028	6.668	0.489	1.000
	Both	3.419	6.999	−0.647	.988
	Adaptive	−5.229	8.087	0.403	.966
Compulsive	Both	2.391	5.930	0.837	.994
	Adaptive	−6.257	6.845	−0.914	.889
Both	Adaptive	−8.648	7.509	−1.152	.776

Figure 3.

Mean post-treatment EDE global score, dietary restraint score, LOC frequency, and purging frequency by exercise type reported at baseline. *p < 0.05

Post-treatment dietary restraint.

Those engaging in no exercise, compensatory-only, compulsive-only, adaptive-only, and both compulsive and compensatory exercise at baseline did not significantly differ in post-treatment levels of dietary restraint (Table 4, Figure 3).

Post-treatment LOC eating frequency.

Those engaging in no exercise, compensatory-only, compulsive-only, adaptive-only, and both compulsive and compensatory exercise at baseline did not significantly differ in post-treatment levels of LOC eating (Table 4, Figure 3).

Post-treatment purging frequency.

Those engaging in no exercise, compensatory-only, compulsive-only, adaptive-only, and both compulsive and compensatory exercise at baseline did not significantly differ in post-treatment levels of purging (Table 4, Figure 3).

Discussion

In this study we investigated how engagement in different types of exercise at baseline (i.e., no exercise, adaptive-only exercise, compulsive-only exercise, compensatory-only exercise, and both compulsive and compensatory exercise) was associated with baseline ED pathology and treatment outcomes in a sample of 106 individuals with BN. Overall, findings indicate that those engaging in compulsive, compensatory, adaptive, and no exercise exhibit different patterns and severity of BN pathology. Partially consistent with hypotheses, we found that adaptive-only exercise at baseline was associated with significantly lower baseline global eating pathology compared to compensatory-only exercise engagement. Interestingly, we also found that both compulsive and compensatory exercise and no exercise engagement at baseline were associated with significantly higher post-treatment global eating pathology compared to baseline compulsive-only exercise engagement.

Baseline Comparisons

Between groups, there were no significant differences in baseline dietary restraint or purging frequency. These results may indicate that domains of dietary restraint and purging frequency are not impacted by exercise type. As expected, individuals who engaged in adaptive-only exercise at baseline also reported significantly lower baseline global eating pathology in comparison to those engaging in compensatory-only exercise at baseline. Interestingly, those who engaged in compulsive-only exercise at baseline had lower baseline LOC frequency in comparison to those not engaging in any exercise at baseline. Based on these findings, individuals with BN who engage in baseline adaptive-only exercise and baseline compulsive-only exercise may exhibit lower levels of baseline domain specific eating pathology compared to other groups. Findings that those engaging in baseline compulsive-only exercise were relatively higher in global eating pathology and lower in LOC frequency may indicate that compulsive exercise engagement is more strongly associated with cognitive ED symptom severity compared to behavioral ED symptoms such as LOC eating. This is consistent with prior research linking compulsive exercise engagement to cognitive ED symptoms (Scharmer, Gorrell, Schaumberg, & Anderson, 2020). It is also interesting to note that those engaging in no exercise at baseline look fairly similar to those engaging in maladaptive exercise at baseline across baseline domains of eating pathology, with the exception of LOC frequency. These findings (especially in the context of global eating pathology scores) may indicate that no engagement in exercise might be just as reinforcing of ED pathology as engagement in maladaptive exercise (e.g., lack of compensatory exercise following a binge episode might increase an individual’s tendency to engage in dietary restriction just as much as engagement in compensatory exercise following a binge could negatively reinforce binge eating itself, thus keeping them trapped in a binge-restrict cycle). Further these findings underscore recent work highlighting the positive impacts of adaptive exercise on a broad range of mechanisms maintaining ED symptoms (Kolar & Gorrell, 2021; Mathisen, Sundgot‐Borgen, Bulik, & Bratland‐Sanda, 2021).

Treatment Outcomes

There were no significant differences in baseline or post-treatment dietary restraint between baseline exercise types in individuals with BN. Within the literature, associations between maladaptive exercise and dietary restraint have been found in populations with anorexia nervosa (AN) (Holtkamp et al., 2003), general ED populations (Dalle Grave et al., 2008), and within rodent models (Exner et al., 2000; Pirke et al., 1993). One potential rationale for why we did not see differences in dietary restraint across the exercise types is that past research has been done in primarily AN samples (Dalle Grave et al., 2008; Holtkamp et al., 2003). Our findings may indicate that the association between dietary restraint and maladaptive exercise are more prominent in AN than BN symptom pathology. We also did not see any significant differences in post-treatment frequency of LOC eating or purging by baseline type of exercise engagement. The lack of findings in these domains could be due to the overall low engagement in these behaviors at post-treatment, given than many individuals with BN reach remission from behavioral ED symptoms (i.e., LOC eating, purging) while relatively fewer reach remission from cognitive ED symptoms (Atwood & Friedman, 2020).

Counter to our hypothesis, we also found that those engaging in compulsive-only exercise at baseline also reported significantly lower post-treatment global eating pathology compared to those engaging in no exercise at baseline. Past research has found that compulsive exercise negatively impacts treatment outcomes (Monell et al., 2018), where others have found that compulsive exercise engagement has no effect on treatment outcomes (Dalle Grave et al., 2008). Our findings add to this discourse by demonstrating that individuals who do not engage in any exercise at baseline reported higher post-treatment global eating pathology relative to individuals who engage in compulsive-only exercise at baseline. This finding may indicate that there is some inherent benefit to exercise engagement regardless of exercise type. For example, both maladaptive and adaptive exercise have shown similar effects on fears of weight gain and body dissatisfaction (Lampe et al., 2022; Srivastava et al., 2022) in ED samples, both of which make up a large part of the EDE global score. Further, exercise is generally thought of as mood-regulating regardless of type (Kolar & Gorrell, 2021; Peluso & Andrade, 2005), which could contribute to improved ED treatment outcomes. While adaptive exercise engagement is certainly preferable to maladaptive exercise engagement due to its positive effects on ED pathology and treatment outcomes, these findings may shed light on the advantages of any exercise engagement over the course of treatment for EDs and underscore the need for interventions which can increase adaptive exercise engagement specifically.

As hypothesized, we found that baseline engagement in compulsive-only exercise was associated with significantly lower post-treatment global eating pathology compared to those engaging in both compulsive and compensatory exercise. Further, those engaging in both compulsive and compensatory exercise at baseline reported the highest post-treatment global eating pathology on average, indicating that there may be a compounding effect of engagement in both compensatory and compulsive exercise which interferes with treatment progress. Specifically, CBT-E may be well suited to intervene on either compulsive or compensatory exercise independently. However, it may be more difficult for therapists to adequately target both types of exercise at once given the limited treatment duration and many other foci of intervention (e.g., regular eating, overvaluation of weight and shape). Tailored clinical interventions may be required for individuals with BN who report both compulsive and compensatory exercise.

Limitations and Future Directions

Findings from the current study should be interpreted in light of several limitations. First, this research was carried out in a BN sample; however, maladaptive exercise is transdiagnostic across EDs with up to 60% of treatment-seeking individuals with an ED reporting engagement in maladaptive exercise (Monell et al., 2018).

Second, the exercise types in the current study were mutually exclusive, and a person endorsing any level of maladaptive exercise was categorized as such, even if they reported some adaptive episodes. While an important preliminary step, this approach may not capture the true heterogeneity in exercise behavior in the context of EDs (Lampe, Trainor, et al., 2021), and it is critical to conduct future studies aimed at understanding the impact of concurrent/mixed adaptive and maladaptive exercise engagement on ED severity and treatment outcome.

Third, 87.74% of participants in our study identified as female and extant studies examining compulsive exercise have historically focused on female populations (Dalle Grave et al., 2008; Holtkamp, Hebebrand, & Herpertz‐Dahlmann, 2004; Taranis & Meyer, 2011). Due to the small cell size, we could not examine the effect of gender, however future research should investigate gender differences in the association of exercise types with ED pathology in BN. It is also possible that there are other demographic/baseline features affecting these relationships besides sex, age, and BMI. Future studies should control for additional variables such as duration of illness and medication use.

Fourth, we do not know the stability of these relationships and whether these behaviors might change in the longer term past treatment completion. Future research should examine how baseline exercise type impacts stability of eating disorder symptoms following treatment termination and how change in exercise engagement over treatment (e.g., from maladaptive to adaptive) impacts treatment outcomes.

Fifth, the modified EDE questions assessing frequency of compensatory-only, compulsive-only, and both compulsive and compensatory exercise have not been independently validated. Previous studies using EMA have demonstrated that 94.44% of individuals who endorse these modified exercise items also reported at least one episode of EMA-measured compensatory and/or compulsive exercise engagement over the following 7–14 days (Lampe et al., 2023). In this study, the correlation between assessment of exercise via modified EDE items and EMA was r = 0.49 (p < 0.001). While the associations of these items with ED severity and treatment outcome identified in this paper offer some preliminary evidence for construct validity, future studies should aim to validate these modified items to ensure that we are optimally and accurately assessing exercise behavior in the context of EDs.

Sixth, despite efforts to help participants recall details occurring within a specific time frame during the EDE assessment, accurate recollection and self-reported responses should be acknowledged as a potential limitation of the current study. For example, while participants were asked to provide examples of various types of exercise, individuals with EDs are known to discount less strenuous activities and to generally under-report exercise behavior (Bratland‐Sanda et al., 2010).

Seventh, though we can establish associations, we cannot determine whether baseline exercise type causes decreased ED symptom severity at post-treatment.

Finally, this study included aggregated data from four clinical trials using CBT-E. Of the four trials, one used CBT-E with no adjunctive interventions (participants in Project COMPASS were only included if they were not randomized to receive any acceptance-based components), two included CBT-E plus just-in-time-adaptive interventions (JITAIs) and one included CBT-E plus an additional form of intervention (i.e., inhibitory control training). Additionally, two of the four parent trials were 16-week interventions and two were 12-week interventions. All models covaried for parent study enrollment and we found no significant impact of parent study enrollment either baseline or post-treatment ED pathology. However, aggregating data across clinical trials may have impacted results (e.g., potential demand characteristics based on different foci of individual studies).

Conclusions

Future research is necessary to investigate tailored treatment interventions. We found that those engaging in baseline compensatory-only exercise had significantly higher baseline global eating pathology than those engaging in adaptive-only exercise at baseline. Considering our findings, interventions targeting compulsivity more broadly (e.g., increasing cognitive flexibility, exposures) may be better able to intervene on both compulsive exercise engagement and cognitive ED symptoms among this group. Additionally, we found that engagement in no exercise or compulsive and compensatory exercise at baseline was associated with higher post-treatment eating pathology. Future research should investigate tailored treatment implementations for BN that target both compulsive and compensatory exercise as well as methods to increase adaptive exercise.

Public Significance Statement.

No research to date has examined whether those who engage in adaptive, compulsive, and/or compensatory exercise exhibit differences in BN pathology or treatment outcome compared to those not engaging in exercise, limiting targeted intervention efforts. We found that those engaging in compulsive, compensatory, and adaptive exercise exhibit different patterns of BN pathology and that adaptive exercise engagement was related to lower cognitive eating disorder symptoms at baseline.

Data availability:

De-identified data from this study are not available in an a public archive. De-identified data from this study will be made available (as allowable according to institutional IRB standards) by emailing the corresponding author.