Examining Within-session and Between-session Change in Exposure-Meal-related Distress During Intensive Outpatient Treatment for Eating Disorders

Abstract

There has been increasing theoretical interest in adapting exposure therapy for eating disorders (EDs), and preliminary data tentatively support the utility of meal-based exposures for targeting eating-related distress. Drawing upon existing work examining habituation processes and theories of inhibitory learning in anxiety disorders, we evaluated patterns of between- and within-session change in subjective units of distress (SUDS) during meal exposures in a mixed diagnostic sample of individuals enrolled in an intensive outpatient program for EDs. Participants (N=178) who were enrolled in an intensive outpatient program completed SUDS at regular intervals during exposure meals over the course of treatment, M(SD)_{TreatmentDuration} = 13.67(10.27) weeks. Results from multilevel models suggested significant between-session decreases in SUDS; no significant changes in within-session SUDS occurred over the course of treatment. Models examining associations between average self-reported eating disorder symptoms and global functioning and SUDS trajectories indicated that individuals with higher global functioning reported slightly lower within-session decreases in SUDS compared to those with lower global functioning. Overall, results provide an initial descriptive examination of habituation processes associated with the use of meal-based exposures for EDs and contribute to a growing body of evidence suggesting benefits of adapting exposure-based intervention approaches for EDs.

Summary:

• Participants (N=178) enrolled in an intensive outpatient program completed SUDS at regular intervals during exposure meals over the course of treatment

• There were significant between-session decreases in SUDS across exposure meals in an intensive outpatient eating disorder (ED) sample

• There were no significant changes in within-session SUDS over the course of treatment.

• Higher average ED pathology was associated with decreased SUDS across meal exposures.

• Higher average global functioning was associated with decreased SUDS across exposure meals.

• Individuals with higher average global functioning reported slightly lower within-session decreases in SUDS.

Eating disorders (EDs) are associated with decreased quality of life (Winkler et al., 2014), increased risk for suicidality (Maguire et al., 2025), and severe medical consequences (Amiri & Khan, 2024), with low remission rates (pooled transdiagnostic rate: 46%; Solmi et al. 2024). Accordingly, identifying active mechanisms of ED treatment may facilitate improved ED treatment outcomes. Because exposure-based interventions for anxiety disorders have robust support (Abramowitz, 2013), and anxiety-related processes underlie ED psychopathology (Essayli et al., 2023; Koskina, et al., 2013; Steinglass et al., 2011), adapting exposure-based interventions commonly used in anxiety disorders has been recommended (Reilly et al., 2017).

Recently, researchers have proposed adapting exposure-based interventions, commonly used in anxiety disorders, as an adjunctive intervention approach in EDs (e.g., Reilly et al., 2017). The theoretical rationale for increasing the emphasis of exposure-based interventions in ED treatments includes well-established, significant comorbidity between anxiety and EDs, shared temperamental features, and analogous clinical characteristics, such as behavioral avoidance (i.e., of foods) and safety behaviors (i.e., extreme caloric restriction or compensatory behaviors that function to offset caloric intake that could result in feared weight gain; Koskina et al., 2013; Steinglass et al., 2011). When considering exposure therapy for EDs, meal-based food exposures —consuming avoided food(s) — target fears common across ED presentations. Although adding meal-based exposures to ED treatments has preliminary support (Ferrer-Garcia et al., 2019; Steinglass et al., 2012), few studies have examined mechanisms associated with food exposure outcomes. A significant body of research, however, has examined mechanisms underlying exposure for anxiety-related disorders. For example, several studies have explored reductions in subjective units of distress (SUDS) as a proxy for habituation, or reduction of expressed fear via repeated exposures over time (e.g., Bluett et al., 2014; Hoeboer et al., 2022; Meuret et al., 2012; Norton et al., 2011). The term “habituation” has been frequently debated, based on the premise that a reduction in distress (i.e., within and between repeated exposure sessions) is more accurately described as an inhibitory learning process in which old learned associations (in this context, threat-based associations) are inhibited by new learned associations (in this context, safety-based associations; Craske et al., 2008), rather than a non-learning process in which habituation occurs simply due to repeated exposure to a distressing stimulus over time. For the purposes of the current paper, which aims to provide an initial examination of habituation patterns during meal exposures for EDs, observed reductions in distress within, and over the course of, repeated exposure sessions will be referred to as evidence of within- and between-session habituation, respectively. Foa and Kozak’s (1986) emotional processing theory initially posited that repeated exposure to a feared stimulus results in extinction, evidenced by within-session habituation (e.g., significant reduction within an exposure session) and between-session habituation (e.g., significant reduction in SUDS across repeated exposure sessions), and that between-session habituation was dependent on the presence of within-session habituation (Foa et al., 2006).

Greater between-session habituation is predictive of treatment response for anxiety-related disorders (e.g., Benito et al., 2018; Bluett et al., 2014; Gallagher & Resick, 2012; Kircanski et al., 2014; Zoellner, 2023). However, others have demonstrated mixed associations among between-session habitation and symptom change (e.g., Baker et al., 2010; Meuret et al., 2012; Hoeboer et al., 2022; Elsner et al., 2022; McGlade et al., 2023). Within-session habituation may relate to between-session habituation (van Minnen & Hagenaars, 2002) and improved treatment outcomes (Norton et al., 2011; van Minnen & Hagenaars, 2002; Hoeboer et al., 2022; Elsner et al., 2022); however, some studies failed to support a relationship between within-session habituation and treatment outcome (van Minnen & Foa, 2006; McGlade et al., 2023) or found that less within-session habituation predicted favorable outcomes (e.g., Culver et al., 2012; Sripada & Rauch, 2015). These mixed findings have prompted theoretical and empirical developments in the understanding of mechanisms underlying exposure therapy (e.g., Craske et al., 2008; Craske et al., 2014; Sripada & Rauch, 2015) but are less well-studied in EDs.

Several existing studies have examined habituation processes during food exposures in ED samples. Two small case series among women with AN explored the relationship of between-session SUDS change and treatment outcome but did not assess within-session changes. Whereas between-session decreases in SUDS across exposure were significantly associated with greater post-treatment test meal intake (Steinglass et al., 2012), another study found increased between-session SUDS ratings across exposure meal sessions despite improvements in eating concern, food restriction, ED-related preoccupation, anxiety, and increased BMI (Cardi et al., 2018). Discrepancies in findings may result from differing rationales provided to patients across those studies, as Cardi et al. (2018) used an inhibitory learning framework with clients, such that they were taught that new learned associations would override existing learned associations with repeated exposure, but that those experiences may not reduce within-session distress. Contrastingly, for Steinglass et al.’s (2012) exposure and response prevention framework, within-session and between-session habituation was presumed to promote symptom change. Additional research using larger, more well-powered samples that examines within-session change during exposure is needed to better understand those results.

Two recent better-powered studies examined within- and between-session changes in exposure for EDs. Butler et al. (2024) compared imaginal (N = 143) to imaginal plus in vivo (N = 26) exposure to a core ED fear (not required to be food-related) in a mixed clinical ED sample. Findings indicated that between-session habituation to imaginal exposures was associated with greater reduction in ED symptoms in the larger sample. However, they failed to find such an association when imaginal exposure was combined with in vivo exposure in the smaller sample. In both studies, within-session habituation was not significantly related to improved ED symptoms.

To date, one study has examined both within- and between-session changes in SUDS during exposure meals (Essayli et al., 2023) among adolescent partial hospitalization patients. Findings indicated that between-session habituation (measured as decreases in pre-exposure SUDS) significantly predicted weight gain and decreased ED psychopathology. In contrast, within-session habituation was not associated with treatment outcome. However, this investigation assessed pre- and post-exposure SUDS difference scores to assess within-session habituation, limiting the ability to assess dynamic trajectories of distress throughout each exposure session. For instance, if peak distress occurred mid-exposure, it would not be reflected by this difference score. Second, their sample was comprised of adolescents who predominantly had anorexia nervosa-spectrum diagnoses (96.3%). It remains unclear if their results would generalize in more demographically and diagnostically diverse samples. Finally, they assessed ED psychopathology, leaving the relationship between habituation and global functioning unexplored.

To address limitations and existing knowledge gaps, the current study sought to characterize within- and between-session trajectories of distress during mealtime exposures among a mixed diagnostic sample of patients with EDs, including adolescents and adults, in intensive outpatient program (IOP) treatment. Understanding whether within-session and between-session habituation are both associated with change in ED and global functioning outcomes is important because it affects how clinicians would frame and conduct exposure sessions. For example, if within-session habituation were critical, practitioners should continue the exposure until a client’s SUDS decrease during each session. If only between-session SUDS change matters, practitioners may inform clients that they may not feel better during the session, but likely will over time. Past work in anxiety-related disorders demonstrated robust between-session decreases in anxiety and distress (Bluett et al., 2014; Gallagher & Resick, 2012; Jaycox et al., 1998), with more mixed data supporting within-session habituation (e.g., Norton et al., 2011; van Minnen & Hagenaars, 2002; Baker et al., 2010; Bluett et al., 2014). These findings coincide with Essayli et al.’s (2023) results in adolescent partial hospitalization patients, for whom between-session but not within-session SUDS reductions were associated with ED symptom improvements. Thus, we hypothesized that between-session SUDS ratings would decrease over the course of treatment (i.e., evidence between-session habituation) and would be associated with better self-reported symptoms over the course of treatment, such that patients with lower average ED symptoms and higher average global functioning would demonstrate greater between-session SUDS decreases. Noting consistently mixed findings for within-session SUDS, an exploratory approach was adopted when examining patterns for within-session SUDS, or associations between within-session SUDS trends and average ED symptoms or global functioning.

2. Methods

2.1. Participants and Procedures

Participants comprised consecutive unique admissions to an IOP for EDs in the northeast US between May 2013-October 2018. Data were collected as part of IOP routine clinical outcome monitoring. Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5; American Psychiatric Association, 2013) diagnosis and appropriateness for IOP was determined by doctoral-level clinicians who conducted initial intake evaluations via unstructured interviews, as part of routine clinical care. Following initial evaluation, patients were admitted to the IOP if they (1) had an ED diagnosis, (2) were determined, by the evaluating clinician and the patient and/or patient’s carer(s), to require more structure and accountability than typical outpatient care, and (3) were medically and psychologically stable. Exclusion criteria for the IOP program included: (1) medical instability requiring hospitalization, (2) self-harm or suicide risk requiring immediate attention, (3) unmanaged psychotic or bipolar disorder, (4) current moderate or severe substance abuse disorder. Additional details about the sample are reported in Walker et al. (2021). The study protocol was approved by Union College’s human subjects review board. The need for consent forms was waived, as data were collected as part of routine clinical care and later were aggregated and de-identified for analysis for the current study.

IOP treatment groups provided evidence-based therapy modalities, including cognitive behavioral therapy, dialectical behavior therapy, and supervised exposure meals (Fairburn, 2008; Waller et al., 2007). All groups were led by doctoral or masters’ level licensed clinicians or advanced graduate students supervised by licensed clinicians. Each treatment day included one exposure meal, during which patients were asked to eat a meal that included feared foods and/or feared amounts of food, while completing an exposure meal form, in which they reported: their feared outcome(s) pre-exposure, a thought or belief to help them through the exposure practice, their SUDS throughout the meal, and, post-exposure, what was learned.

Patients attended up to three IOP treatment days per week from 2013-2015 and up to four days per week from 2015-2018. Each week, one “challenge” meal was catered, mirroring commonly avoided and/or feared foods for individuals with EDs (i.e., pizza, macaroni and cheese, chicken/eggplant parmesan with pasta, Chinese-American take-out). For other meals, patients brought their own meals according to the dietary exchanges included in their individualized meal plans; they planned these meals at home or in individual sessions with the staff dietician. Patients had one CBT goal-setting group per week that focused on behavioral challenges, which included selecting specific challenging foods to either bring to IOP meals or eat at home. Trained IOP staff checked patients’ meals to ensure it met their meal plans; portions and macronutrient content were adjusted by staff, with options from the IOP kitchen substituted to fill missing dietary exchange(s), as needed. Two IOP staff members ate with, and provided support to patients, during the meals. Occasionally, patients who needed one-on-one support would be pulled out and would sit with the case manager or another available practitioner to help them finish their meal. Patients who completed 0-<50% of their meal supplemented with two nutritional drinks, and 50%-<100% meal-completion supplemented with one nutritional drink. Patients who did not complete their meals or nutritional drink supplements for more than three consecutive treatment days were considered for admission to a higher level of care. Because of this protocol, patients reported 100% meal completion for almost all meals in the current sample. During each meal, from May 2013- May 2016, IOP patients rated SUDS every five minutes, throughout a 60-minute exposure meal. Due to change in clinic assessment protocol, from May 2016-October 2018, patients rated SUDS every 15 minutes during a 45-minute meal. SUDS frequency was reduced in response to patient feedback about intrusiveness of the original protocol (See Supplement for both Exposure Meal Form versions).

2.2. Measures

2.2.1. Subjective Units of Distress (SUDS)

SUDS, ranging from 0 (no distress) to 100 (maximal distress; Wolpe & Lazarus, 1966), were used to assess momentary distress and model habituation throughout and between exposure meals. Within-session habituation was operationalized as a decrease in SUDS over the course of an exposure meal session. Between-session habituation was operationalized as a decrease in SUDS across exposure meal sessions.

2.2.2. Eating Attitudes Test-26 (EAT-26; Cronbach’s α = .91)

The EAT-26 (Garner et al., 1982), a transdiagnostic measure of global ED psychopathology (Mintz & O’Halloran, 2000), is a 26-item, six-point truncated Likert-type scale, rated on a scale of 0 (never, rarely, and sometimes), 1 (often), 2 (usually), and 3 (always). The EAT-26 assessed weekly levels of ED attitudes and behaviors throughout treatment; higher scores indicate greater ED psychopathology (range = 0-78). The EAT-26 distinguishes between those with and without EDs and exhibits good convergent validity (Garner et al., 1982) and internal consistency.

2.2.3. Outcome Rating Scale (ORS; Cronbach’s α = .84)

The ORS (Miller et al., 2003) comprises four visual analog scales assessing individual, interpersonal, social, and overall well-being, which assessed weekly global functioning throughout treatment. Items are summed; higher scores reflect greater adaptive functioning (range = 0-40).

2.3. Analytic Approach

Data were double-entered and analyzed using R (R Core Team, 2021). Hierarchical linear models were estimated using the lme4 package (Bates et al., 2015). Data were screened for univariate and multivariate outliers, as well as violations of normality and homogeneity assumptions prior to analyses. The primary outcome measure, SUDS, was screened for extreme outliers (> 3 SDs from group mean SUDS score). All other variables were screened for outliers, problematic skewness (>2.5), and kurtosis (>7) prior to analyses. Assumptions of normality, sphericity, homogeneity of variance-covariance matrices, linearity, and multicollinearity were assessed via Box’s M, Levene’s test, and Mauchley’s test. Preliminary analyses assessed for patterns of missingness. When appropriate, variables were grand mean centered, prior to multilevel model (MLM) analyses. Descriptive analyses explored demographic variables assessed at IOP intake, including participant age, body mass index (BMI; kg/m²), and race/ethnicity (see Table 1).

Table 1.

Demographic and baseline clinical variables of interest.

Variable Name	n (%)	Range	Mean	SD
Demographic Variables
Age (years)		13-62	25.1	10.8
BMI (kg/m2)		14.5-46.7	21.8	5.6
Gender
Female	157 (88.2)
Male	19 (10.7)
Transgender	1 (0.6)
No data/Other	1 (0.6)
Race/Ethnicity
White/Caucasian	164 (92.1)
Asian	5 (2.8)
Multiple race/ethnicities	2 (1.1)
Hispanic/Latinx	2 (1.1)
Black/African American	1 (0.6)
No data/Other	4 (2.2)
Clinical Characteristics (baseline assessment)
Eating disorder diagnosis
Anorexia nervosa	109 (61.2)
Bulimia nervosa	39 (21.9)
OSFED	20 (11.2)
Binge eating disorder	8 (4.5)
ARFID	2 (1.1)
Comorbid DSM-5 diagnosis	108 (60.7)
Predictor Variables of Interest (baseline assessment)
EAT-26		1-67	34.6	17.0
ORS		0-31.6	16.9	7.3

Note. Total N = 178. OSFED = Other specified feeding or eating disorder; ARFID = Avoidant-restrictive food intake disorder; Comorbid DSM-5 diagnosis = endorsement of one or more comorbid DSM-5 psychiatric disorder(s); EAT-26 = Eating Attitudes Test 26 total score; ORS = Outcome Rating Scale total score.

2.3.1. Primary Analyses

We used MLM to evaluate changes in SUDS within and between exposure meal sessions. Linear trends in SUDS ratings throughout the exposure meal were assessed using three-level MLM models, with within-session SUDS ratings represented at level 1, between-session SUDS ratings represented at level 2, and participant-level predictors (intercept, covariates, average EAT-26 scores, average ORS scores) represented at level 3. Data from participants were combined to estimate average slopes of SUDS ratings within (time) and between (session) exposure meals for each individual. Random effects were used for intercept, within-session SUDS slope, and between-session SUDS slope, and an unstructured (UN) covariance parameter was selected. Statistical significance for all tests was set at p<.05.

Each MLM model utilized a sequential model-building approach to examine the linear effect of time on SUDS ratings within the exposure meal session, as well as between exposure meal sessions. In Step 1, an unconditional model was fitted with fixed slope and random intercept. Within-session time (“time” variable measured in 5-minute increments corresponding to the points at which SUDS measurements were acquired) was added in Step 2 as a level 1, within-session predictor of SUDS ratings. In Step 3, exposure meal “session” was added to the model as a level 2 between-session predictor of SUDS ratings. Two additional models were conducted to evaluate whether inclusion of main and interaction effects for level 3, within-person change in ED-specific symptoms (EAT-26) and global functioning (ORS) significantly improved model fit for the trend in SUDS ratings. To assess model parsimony and improvement in fit, Chi-square tests of independence were conducted on the difference in −2 log likelihood (−2LL) values between successive steps in the multilevel model. Pre-meal SUDS assessment time point was designated as Time 0 and initial IOP session was designated as Session 0. Continuous covariates and participant-level means for the EAT-26 and ORS were grand mean centered prior to analysis. Categorical covariates, including self-reported sex, racial/ethnic identification, and meal type (IOP-provided meal, patient-provided meal), were dummy coded, such that the majority category of the sample was coded as the reference group (e.g., female = 0, White/Caucasian = 0, patient-provided meal = 0). Analyses were based on all available data. Full information maximum likelihood (FIML) estimation methods were used for all models. A sensitivity analysis was conducted to examine whether SUDS trajectory results differed after the protocol changed from assessing SUDS every 5 minutes (5-minute protocol = 0) to every 15 minutes (15-minute protocol = 1).

3. Results

3.1. Sample Descriptives

Participant demographics and clinical characteristics at baseline are presented in Table 1. We applied the interquartile range (IQR) rule to screen for extreme values of treatment duration and removed nine outliers. The final data set consisted of N = 178 participants, who were predominantly female, White, ranged in age from 13-62 years, M(SD) = 25.1(10.8) (see Table 1). Participants’ initial meal exposure SUDS data were collected within 0-6 days from their initial intake assessment within the IOP program. Participants’ total treatment duration in the IOP treatment program ranged between 1-45 weeks in treatment, M(SD) = 13.67(10.27), yielding a final dataset consisting of 7,429 exposure meals and 43,188 SUDS ratings. Beyond the nine participants excluded for extreme treatment duration, preliminary analyses identified no other extreme outliers, problematic skewness, or kurtosis.

3.2. Primary Results

MLM results are reported, below, for models examining (1) the within-session (time) and between-session dose (session) effects on SUDS trajectories, (2) average ED symptom severity and symptom change (EAT-26*time and EAT-26*session) effects on SUDS trajectories, and (3) average global functioning and functioning change (ORS*time and ORS*session) effects on SUDS trajectories among the current sample.

3.2.1. Within-session and between-session change in SUDS

After accounting for covariate effects, results from the unconditional model and model including linear within-session and between-session time effects with random slope and random intercept demonstrated significant variability in slope and intercept for SUDS ratings. Shown in Table 2, the inclusion of linear within-session time effect within Step 2 significantly improved model fit from Step 1; however, the within-session time effect was not significant, such that SUDS did not significantly decrease within the course of an exposure meal session, b = −0.05, 95% CI [−0.11, 0.01]. The inclusion of between-session time effects within the final model (Step 3), significantly improved model fit, χ² = 642.46, p < .05). Between-session time effects were significant predictors of SUDS ratings, such that SUDS ratings decreased as meal exposure “dose” increased (i.e., SUDS decreased by 0.16 units, on average, with every meal exposure session), b = −0.16, 95% CI [−0.21, −0.11].

Table 2.

Summary of multilevel model specification for within-session time and exposure session effects on subjective units of distress (SUDS).

Model	Effect	Fixed effects			Random effects			Model AIC	Χ 2 Change
		Est.	SE	95% CI	Time Slope	Session Slope	Intercept
Step 1 (unconditional model)							696.0*	415476.5	--
	Level 3: Intercept*	46.40	2.22	42.05, 50.75
Step 2 (time-in-session added)					0.12*		643.02*	414840.0	642.46*
	Level 3: Intercept*	47.26	2.16	43.03, 51.50
	Level 1: Time	−0.05	0.03	−0.11, 0.01
Step 3 (exposure session added)					0.11*	0.08*	709.04*	409937.1	4913*
	Level 3: Intercept*	54.57	2.27	50.11, 59.02
	Level 2: Session*	−0.16	0.02	−0.21, −0.11
	Level 1: Time	−0.05	0.03	−0.11, 0.01

Note. Est. = Estimate; SE = Standard Error; CI = Confidence Interval. Asterisk (*) denotes significance at p < .05. AIC = Akaike information criterion; χ²_Change = Chi-square test of independence, based on difference between −2 loglikelihood values; SUDS = dependent variable, assessed using visual analog scale (0-100mm) rating of momentary subjective units of distress during exposure meal sessions. Models controlled for covariates: age at IOP intake, race/ethnicity, gender, exposure meal type.

3.2.2. ED symptom severity: Predictor of within-session and between-session SUDS change

Results for the MLM model examining the effect of ED symptom severity (EAT-26) are presented in Table 3. The main effect for EAT-26 small, but significant; greater mean EAT-26 scores (averaged across all time points and sessions) were significantly associated with greater average SUDS levels, b = 0.90, 95% CI [0.65, 1.15]. However, average EAT-26 scores did not interact with within-session time to significantly predict slope of SUDS within meal exposure sessions; nor did average EAT-26 scores significantly interact with between-session time to predict slope of SUDS across meal exposure sessions. Further, even accounting for ED severity, session significantly predicted decreased SUDS, b = −0.20, 95% CI [−0.26, −0.14]. Therefore, average EAT-26 scores were not significantly associated with SUDS trajectories.

Table 3.

Full MLM model results: Within-session time, session dose, and global symptom change effects on subjective units of distress (SUDS).

Model	Effect	Fixed effects			Random effects			Model AIC
		Est.	SE	95% CI	Time Slope	Session Slope	Intercept
Impact of Time and Session on SUDS					0.11*	0.08*	709.04*	409937.1
	Intercept*	54.57	2.27	50.11, 59.02
	Session*	−0.16	0.02	−0.21, −0.11
	Time	−0.05	0.03	−0.11, 0.01
Impact of Eating Disorder Symptom Change on SUDS					0.13*	0.10*	487.92*	302306.4
	Intercept*	51.67	2.16	47.44, 55.90
	EAT-26*	0.90	0.13	0.65, 1.15
	Session*	−0.20	0.03	−0.26, −0.14
	Session x EAT-26	0.001	0.002	−0.002, 0.005
	Time	−0.05	0.04	−0.12, 0.02
	Time x EAT-26	0.003	0.002	−0.001, 0.008
Impact of Global Functioning Change on SUDS					0.12*	0.10*	577.38*	302321.9
	Intercept*	52.08	2.33	47.51, 56.65
	ORS*	−1.54	0.30	−2.13, −0.95
	Session*	−0.20	0.03	−0.26, −0.14
	Session x ORS	−0.003	0.004	−0.012, 0.004
	Time	−0.05	0.03	−0.12, 0.02
	Time x ORS*	−0.01	0.004	−0.021, −0.004

Note. Est. = Estimate; SE = Standard Error; CI = Confidence Interval. Asterisk (*) denotes significance at p < .05. AIC = Akaike information criterion; SUDS = dependent variable, assessed using visual analog scale (0-100mm) rating of momentary subjective units of distress during exposure meal sessions. EAT-26 = Eating Disorder Attitudes Test; ORS = Outcome Rating Scale; Models controlled for covariates: age at IOP intake, race/ethnicity, gender, exposure meal type

3.2.3. Global functioning symptom severity: Predictor of within-session and between-session SUDS change (Table 3)

In the final step of the MLM model examining the effect of averaged weekly global functioning (ORS) on within-session and between-session SUDS trends, results indicated a main effect of ORS scores, such that those with higher average global functioning reported lower SUDS, b = −1.54, 95% CI [−2.13,-0.95], and a main effect of session, such that as session decreases, SUDS increase, b = −0.20, 95% CI [−0.26, −0.14], even after controlling for participants’ global functioning. Additionally, within-session time interacted with average ORS level to predict within-session SUDS trajectories; higher average ORS levels were associated with small but significantly lower within-session decreases in SUDS, b = −0.01, 95% CI [−0.021, −0.004]. ORS scores did not interact with between-session time to predict significant variance in slope of SUDS across meal exposure sessions.

3.2.4. Sensitivity Analysis Assessing the Effect of Protocol on Average SUDS (See supplemental table)

There were significant main effects of age, gender, and meal type on average SUDS scores, such that older individuals reported significantly higher distress, b = 0.42, p = .026, men reported significantly lower SUDS, b = −20.58, p = .002, and catered challenge meals were associated with significantly higher distress, b = 8.25, p <.001. There were significant direct effects of the dummy-coded protocol, b = −7.91, 95% CI [−11.20, −4.63], p <.001, such that when the protocol switched from SUDS ratings every 5 minutes to every 15, overall average SUDS scores decreased by almost 8 points (0-100 scale). Additionally, protocol significantly interacted with session, b = 0.14, 95% CI [0.10, 0.17], p <.001, such that rating every 15 minutes led to significantly slower between-session habituation than rating every 5 minutes.

4. Discussion

In the current study, we analyzed trajectories of momentary subjective distress ratings within- and between-exposure meals at an IOP for EDs in a mixed-diagnostic, mixed-age sample. Similar to prior research in anxiety disorders (Bluett et al., 2014; Gallagher & Resick, 2012; Jaycox et al., 1998; Kircanski et al., 2014; Sripada & Rauch, 2015; Zoellner, 2023), we found significant decreases in subjective distress ratings between exposure exercises, indicating an average decrease in meal-related distress across sessions. In contrast, we found no significant decreases in state subjective distress within exposure meals. Further, higher global functioning was associated with slightly lower within-session decreases in subjective distress. Notably, this was not observed for self-reported ED psychopathology, which did not relate to within- or between-session SUDS trajectories. This is the first study examining both within- and between-session changes in state distress during meal exposures and their associations with global functioning and ED psychopathology in a naturalistic, treatment-seeking ED sample.

The current study provides new information on SUDS trajectories during meal-related exposures within ED treatment and aligns with prior research examining SUDS trajectories in ED treatment-related exposures (Butler et al., 2023; Essayli et al., 2024; Steinglass et al., 2012) in finding decreased between-session distress. However, we failed to find a significant relationship among between-session SUDS trajectories and average levels of ED psychopathology, in contrast to prior research (Essayli et al., 2023; Steinglass et al., 2012). The current study includes a larger sample size (N = 178), with more exposure meals (7,429) and SUDS ratings during exposure meals (43,188). Further, these findings appear to coincide with some prior findings from the anxiety disorders literature that suggest that within-session habituation is not a prerequisite for between-session habituation (Craske et al., 2008; Sripada & Rauch, 2015) or overall symptom improvement (Kircanski & Peris, 2015).

The lack of significant within-subject reductions in distress during exposure meals provide preliminary evidence that processes underlying exposure meals are not consistent with those outlined in emotional processing theory, which assumes between-session habituation is dependent on within-session habituation (Foa & Kozak, 1986; Foa et al., 2006). Instead, these effects may be more consistent with the inhibitory learning model, which posits that between-session reductions in distress and anxiety occur via corrective learning of non-fear/distress-related associations that replace fear/distress-based associations regardless of significant within-session fear/distress reduction (see Craske et al., 2014 for further discussion). Specifically, our findings are consistent with the assumption that meal exposures may reduce eating-related distress via new learning that an individual can tolerate distress associated with feared food and meal consumption. However, these statements are tentative, as no data were collected to evaluate patients’ pre- versus post-exposure beliefs about their ability to tolerate the challenge meal content. Therefore, while the present investigation provides initial insight into the processes underlying meal exposures for EDs, future work is needed to further elucidate the mechanisms of change associated with learning during meal exposures in ED treatments.

The current study was the first study to examine the relationship between SUDS trajectories during food exposures and global functioning. We found a main effect of global functioning, such that higher average global functioning was associated with lower distress during meals. We found that higher average global functioning related to slightly lower decreases in SUDS within meal exposure sessions possibly because those with less psychopathology had lower distress ratings to begin with, and, as a result, less room for distress to change compared with those with lower global functioning, who had higher meal-related distress in the current sample. Further, average global functioning did not predict between-session changes in SUDS. While several prior investigations have found associations between between-session reductions in subjective distress and treatment outcomes in non-ED clinical samples (e.g., Bluett et al., 2014; Gallagher & Resick, 2012; Jaycox et al., 1998; Kircanski et al., 2014; Sripada & Rauch, 2015) and ED samples (Essayli et al., 2023; Steinglass et al., 2012), other work has documented significant associations between within-session reductions in SUDS and positive treatment outcomes (e.g., van Minnen & Hagenaars, 2002; Norton et al., 2011) that are broadly consistent with initial models of exposure therapy (Foa & Kozak, 1986).

In contrast to global functioning, SUDS trajectories and average ED psychopathology were not significantly related in the current sample. The main effect of ED symptom severity on SUDS levels was statistically significant, such that higher average ED severity related to higher SUDS scores. However, hypotheses regarding SUDS trajectories and overall ED psychopathology were not supported: there were no significant associations with ED psychopathology and within- or between-session changes in SUDS scores. These findings also ran counter to our predictions based on prior ED research (Burton et al., 2024; Essayli et al., 2023; Steinglass et al., 2012). It is possible that the lack of significant findings is secondary to the assessments selected. Specifically, the EAT-26 was not designed for use as a clinical outcome measure (Lebow, 2015). Further, the truncated Likert-rating scale may limit variability in total score outcomes (Papini et al. 2022), potentially obscuring associations between SUDS trajectories and ED symptoms. Selection of the measures used for the IOP’s routine outcome monitoring were made prior to development of some assessments that may provide greater ability to understand how trajectories of change relate to ED psychopathology. Future research should explore these questions with measures of eating pathology that readily allow examination of different types of cognitive and behavioral symptoms (e.g., the Eating Pathology Symptom Inventory [EPSI, Forbush et al., 2013]). Further, to better understand the current findings, latent profile analysis may be a useful methodology by which to examine whether certain profiles exist within ED patients who are more or less likely to experience within- and/or between-session change during exposure meals and overall symptom improvement. Although the lack of significant associations between EAT-26 and within- and between-session SUDS trajectories may be secondary to measurement choices, lack of association between within-session SUDS and symptom outcome is not unprecedented within the anxiety disorders field (e.g., Badour et al., 2017; Rauch et al., 2018).

Lastly, our sensitivity analysis demonstrated that the change in exposure meal protocol from rating SUDS every 5 to every 15 minutes influenced average and between-session decreases in SUDS. Specifically, rating SUDS every 15 minutes was associated with almost 8 points lower average SUDS. It is possible that rating SUDS every 5 minutes increased momentary distress, caused by reactivity to frequently thinking about one’s distress levels, but it is equally plausible that reducing granularity in SUDS ratings decreases a full picture of the SUDS trajectory and yields lower average SUDS. For example, if a person had SUDS ratings every 5 minutes and experienced an early spike in anxiety that was already in the process of decreasing by the 15-minute mark, that early SUDS elevation would not be evident when rating SUDS before, during, and at the end of the meal, in 15-minute increments. Although the changed protocol in which pre-meal, mid-meal, and post-meal exposure rating provided greater granularity than prior studies (e.g., Essayli et al., 2024), the current results suggest potential utility of more frequent SUDS ratings during exposure meals. Further, there was a session by protocol interaction, which showed that rating SUDS every 15 minutes led to significantly slower between-session habituation than when distress was rated every 5 minutes, supporting that more frequent distress ratings during exposure meals may be optimal.

4.1. Strengths and Limitations

Strengths of the current study include a relatively large sample size, more subjective distress ratings during exposure meals enabling better articulation of SUDS trajectories, and the transdiagnostic, mixed-age sample. Additionally, FIML is a robust method when working with missing data (Niu, 2024). Further, data were collected naturalistically and demographic characteristics and psychiatric comorbidities of the patient sample likely increase generalizability to other “real life” ED treatment settings. Nonetheless, the current study’s limitations also stem from data that were collected as part of routine clinical practice. Specifically, patient diagnosis was determined using unstructured clinical interviews, limiting the potential reliability and validity of clinical diagnoses. Additionally, we do not have information on patients’ other-specified feeding or eating disorder diagnoses to better understand sample characteristics. Further, our results may not generalize to settings where meal plan adherence is not enforced or settings in which avoidance, lack of meal completion, and/or compensatory behaviors are possible.

Additionally, the current IOP comprised several components (e.g., exposure meals, DBT skills groups, CBT skills groups), and clients also met with individual psychotherapists and dieticians. Thus, multiple therapeutic mechanisms may have influenced changes in ED symptoms and global functioning. Therefore, these questions must be explored within the context of more highly controlled study designs and/or should examine the effects of exposure meals delivered as the primary intervention. Participants were also not randomly assigned to receive meal exposures, obfuscating the ability to determine whether changes in SUDS between sessions and changes to ED psychopathology and global functioning were due to engagement in exposure meals or other treatment-related intervention components. Further, these factors may vary across patients, with some responding more to exposure meals and others to other parts of the IOP programming, and it does not allow us to understand what kinds of individuals may respond best to meal exposures and why. Another limitation of the transdiagnostic real-world treatment-seeking sample is the significant variability (age, diagnosis, eating behaviors, duration of treatment, number of exposure sessions completed, etc.) and the change in exposure session protocol partway through the data collection period. Future research should examine whether there are differences across ED diagnoses and other factors that typically vary in real-world treatment settings in response to exposure meals and its relationship to outcome variables. The symptoms-related outcome measurements (i.e., EAT-26; ORS) are limited in their scope. Future research should examine the relationship between within- and between-session habituation on outcome measurements that capture the full spectrum of ED behaviors (e.g., objective measures of food consumption/food avoidance, comprehensive assessments of ED psychopathology), to better understand mechanisms of change in meal exposure for EDs.

4.2. Conclusions

Overall, despite compelling theoretical rationale to support use of exposure therapy for the treatment of EDs, limited empirical data has examined its use, particularly to understand mechanisms driving outcomes. The current study provides an initial description of between- and within-session habituation processes associated with meal-based exposures during IOP in a mixed diagnostic ED sample. Results are consistent with past work highlighting the importance of between-session changes in distress across exposure sessions. Our findings highlight exposure exercises as a promising therapeutic technique for individuals with EDs and suggest potential utility for having clients complete distress ratings more frequently during exposure meals.