Cognitive Behavioral Telehealth Treatment for Adolescents with Loss of Control Eating: A Randomized Controlled Feasibility Study

Abstract

Objective:

To develop and test the feasibility and acceptability of a telehealth treatment for adolescent loss-of-control (LOC) eating, including the feasibility of attaining clinically-meaningful outcomes. LOC-eating episodes occur when experiencing a sense of LOC while eating a subjectively or objectively large amount of food.

Method:

Participants were 55 adolescents (aged 12–17) with recurrent LOC-eating episodes and elevated weight (based on age and sex-normed percentiles) who enrolled in a feasibility trial of cognitive behavioral therapy (CBT). Five participants enrolled in the treatment development “open series” and provided feedback. Fifty participants were randomized to the subsequent randomized controlled trial and received either CBT-LOC or nutrition education (NE). Treatment fidelity was checked.

Results:

CBT-LOC for adolescents with LOC eating demonstrated treatment feasibility and acceptability. Retention (M=15.3 of 16 sessions), treatment credibility (M=5.26 of 6), and treatment fidelity (92–100% of broad topics covered) were high. Treatment satisfaction was higher in CBT-LOC than NE (ps<.001), Time-by-intervention effects revealed greater improvements for CBT-LOC than NE for binge eating, LOC eating, secretive eating, and global eating-disorder psychopathology. Weight change was minimal for both interventions. Among adolescents with weekly or greater baseline LOC-eating episodes, overvaluation of weight/shape and depression also improved.

Conclusions:

Overall, this feasibility randomized controlled trial testing CBT-LOC for adolescents experiencing LOC eating and elevated weight demonstrated feasibility and acceptability. Initial evidence demonstrated it was feasible to produce clinically-significant improvements in LOC-eating episodes, eating behaviors, and eating-disorder psychopathology. Few participants attained decreased weight, which may limit motivation to seek treatment from parents and adolescents.

INTRODUCTION

Binge-eating disorder (BED) carries substantial psychiatric, medical, and psychosocial burden (Santomauro et al., 2021) and is the most common formal eating disorder (Swanson et al., 2011; Udo & Grilo, 2018). BED is associated with functional impairment, chronic medical conditions, and psychiatric comorbidities including depression, anxiety, and suicidality (Marzilli et al., 2018; Micali et al., 2015; Swanson et al., 2011; Tanofsky-Kraff et al., 2020; Udo et al., 2019). Loss of control (LOC) while eating is regarded as the hallmark feature of binge eating. While the DSM-5 (American Psychiatric Association, 2022) definition of BED additionally requires the consumption of an unusually large quantity of food, the ICD-11 (Stein et al., 2020; World Health Organization, 2019/2021) definition does not. Both LOC-eating episodes (experiencing LOC while eating any amount of food) and binge-eating episodes (LOC while eating an unambiguously large amount of food) often emerge during adolescence (Marzilli et al., 2018; Tanofsky-Kraff et al., 2020) and have heightened psychopathology compared with overeating (without LOC) alone (Goldschmidt, 2017; Shomaker et al., 2010). Furthermore, data suggest a developmental trajectory wherein LOC eating is a precursor to binge eating (Tanofsky-Kraff et al., 2020). Together, these points suggest that targeting LOC-eating episodes is clinically warranted, particularly among youth.

Adolescent LOC treatment research is scarce compared with adult treatment research (Byrne et al., 2019; Smith & Goldschmidt, 2024; Tanofsky-Kraff et al., 2020). Cognitive-behavioral therapy (CBT) has a strong evidence base for adult BED (Grilo, 2024; National Institute for Health and Care Excellence, 2017), and CBT-Enhanced (CBT-E), a specific form of CBT for transdiagnostic eating disorders (Fairburn et al., 2009), has been adapted for adolescents (Dalle Grave et al., 2015), but there are no well-established, evidence-based treatments specific to adolescent LOC eating. Treatments for adolescent LOC eating (Tanofsky-Kraff et al., 2020) need to consider the characteristics of adolescence (Kazdin & Weisz, 1998), which include moves towards independence, increases in executive function capacity, and evolutions in social relationships and support (Erikson, 1968).

Prior studies of CBT for adolescent LOC eating demonstrate feasibility, but research is in an early stage, with a notable absence of any comparison to a credible control (Smith & Goldschmidt, 2024). Existing trials, while promising, are also limited by broad inclusion criteria such as combining adolescents with and without purging behaviors (Debar et al., 2013), combining adolescents and young adults (Hilbert et al., 2020) and combining overeating (no LOC) with LOC eating (Jones et al., 2008). Another limitation is that the emerging CBT for LOC eating in adolescents has not to this point considered weight-related motivations for treatment (Dalle Grave et al., 2015), nor been tailored to address adolescent-appropriate parental involvement (Debar et al., 2013; Jones et al., 2008) or executive function deficits and growth (Debar et al., 2013; Hilbert et al., 2020; Jones et al., 2008). Importantly, most trials to date have employed methodologically-weak controls (e.g., waitlist) (Debar et al., 2013; Hilbert et al., 2020; Jones et al., 2008), which tend to inflate effect sizes and decrease generalizability as some families are unwilling to be randomized when the control is delayed treatment (Mohr et al., 2009; Wade et al., 2018).

The present study describes the development and initial evaluation of adolescent-specific CBT for LOC eating (CBT-LOC). CBT-LOC was designed based on existing evidence-based treatments (Fairburn, 2013; Grilo et al., 2023), with developmental appropriateness taken into consideration (executive function content, focus on both short-term and long-term goals in relapse prevention planning, tangible activities such as creating a “monument of progress,” consolidation of skills and encouragement of independence by teaching parent skills). The ecologically-valid comparator was nutrition education (NE; similar to adult BED treatment trials, e.g., Masheb et al., 2011). To develop CBT-LOC, we followed the Stage Model of Behavioral Therapies Research, conducting Stage 1a (development of materials and content in response to patient, parent, and clinician feedback) and Stage 1b (feasibility randomized controlled trial [RCT]) (Carroll & Nuro, 2002; Onken et al., 2014; Rounsaville et al., 2001). Primary outcomes were feasibility of CBT-LOC for adolescents, acceptability, and an initial test of efficacy reducing LOC-eating episodes and weight. Exploratory outcomes included other eating behaviors (objective binge eating, secretive eating) and weight/shape concerns (overvaluation of weight/shape). Depression and self-esteem were additional exploratory outcomes because of their relevance to functioning and well-being in eating disorders (Lydecker & Grilo, 2022; Sutin et al., 2018)(Hamilton et al., 2011). We hypothesized that CBT-LOC would be feasible, more acceptable than NE, and that youth would experience better outcomes in CBT-LOC compared with NE.

METHODS

This study was approved by the Yale Human Investigations Committee and was pre-registered at ClinicalTrials.gov (NCT04088097). Parents provided consent and permission and adolescents provided assent prior to study activities. Eligibility was determined in an initial intake assessment that also established baseline clinical presentation. This assessment included eating-disorder psychopathology, psychiatric conditions, and all survey instruments.

Research Frame

The Stage Model (Carroll & Nuro, 2002; Onken et al., 2014; Rounsaville et al., 2001) advocates for methodology in treatment development Stages 1a and 1b that is as rigorous as major efficacy and effectiveness trials.

Treatment Development Stage 1a is material development. This involves writing a working version of a treatment manual specifying session format (individual, parent-adolescent) and session content (essential elements); developing clinical-training protocols; treatment delivery in an “open series”; refining clinical materials based on feedback; developing the assessment battery; creating comparator materials; and outcomes assessment (Carroll et al., 2000).

Treatment Development Stage 1b is feasibility testing. This involves training study clinicians to administer treatment and comparator conditions; close supervision of initial cases; recruitment; treatment delivery; soliciting feedback; and outcomes assessment (Carroll et al., 2000).

Study Design and Randomization.

Stage 1a was an “open series” in which a consecutive series of adolescents were enrolled to receive CBT-LOC and provide feedback. Stage 1b was a feasibility RCT. Randomization to the two treatments (CBT-LOC or NE) was in equal proportions without restriction or stratification, in blocks of 6. Allocations followed a sequentially numbered randomization schedule created by the study biostatistician and assigned by the PI consecutively at the time when participants were determined to be eligible. Participants were told of their allocation at the first session. Outcomes assessors were masked to intervention assignment.

Treatment Fidelity.

Audio recordings were reviewed for adherence to the CBT-LOC manual, including session structure (agenda, weighing, self-monitoring, home practice) and session-specific content. Treatment fidelity evaluated adherence to the CBT-LOC protocol as CBT-LOC was the treatment undergoing development, but adherence to NE was not specifically evaluated. Digital recordings (n=51) were selected after completion of the study to balance across treatment segment (“quarter” 1, 2, 3, 4) and type of session (parent-adolescent, adolescent). Interventionists who had training in the CBT-LOC protocol but were not study therapists for the cases they reviewed (two advanced doctoral students, one postdoctoral trainee) and the principal investigator coded treatment fidelity.

Format and Content of Treatment

Session content is detailed in Supplemental Table 1 and depicted in Supplemental Figure 1. Study interventionists were advanced doctoral students and/or postdoctoral trainees in psychology; seven interventionists provided CBT-LOC and eight interventionists provided NE.

CBT-LOC involved four months of weekly, 45–60 minute telehealth sessions between the clinician and adolescent (11 sessions) with parent-adolescent sessions occurring monthly (clinician, adolescent, parent; 5 sessions). Adolescents, parents, and clinicians provided feedback on each treatment session (how helpful they found it, how much they liked it, how likely they were to use the strategies covered). Content was presented as “quarters”: starting therapy and specific behavior strategies, coping skills, executive function skill training, and relapse prevention planning.

NE addressed eating behaviors and nutrition and involved weekly educational materials delivered by email to families over 16 weeks (e.g., USDA healthy recipes, list of local recreation areas and activities) with monthly telehealth check-in appointments that were 15–30 minutes plus the time to complete assessments. Materials were all publicly available (e.g., MyPlate.gov, Rudd Center, Mayo Clinic, healthychildren.org) and curated by the study team.

Participants

Recruitment efforts included social media advertisements, trial registries, and referrals from healthcare providers and organizations. Adolescents self-referred or were referred by providers or parents. Adolescents were eligible if they lived in the United States, were 12–17 years old, reported two LOC-eating episodes per month for three months, and were above the 85^th percentile for BMI on age- and sex-normed growth charts. Participants could not have medical or psychiatric conditions requiring intensive care or hospitalization (e.g., severe suicidality, psychosis), developmental disorders that would interfere with therapy, uncontrolled medical conditions requiring intensive care (e.g., cancer), another feeding/eating disorder, pregnancy, or concurrent treatments that influence weight or eating behavior (e.g., ADHD stimulant medications, anti-obesity medications). Other conditions not listed, including current and past psychiatric disorders, were not exclusionary.

Measures

Treatment feasibility was measured by participant recruitment and retention. Treatment acceptability was measured by a credibility survey at the end of the first treatment session and session satisfaction surveys at the end of each session.

Clinical outcomes were measured by instruments with established reliability and validity among adolescents. Eating Disorder Examination-Questionnaire (EDE-Q-5.2) (Berg et al., 2012; Fairburn & Beglin, 1994; Goldschmidt et al., 2007; Tanofsky-Kraff et al., 2004) assessed global eating-disorder psychopathology (in the current study, α=.92), overvaluation of weight/shape, and eating behaviors (objective binge eating, LOC eating [objective and subjective binge eating], and secretive eating [a putative precursor to LOC eating]). Objective binge eating is considered to be eating an unusually large amount of food while experiencing a sense of loss of control. Subjective binge eating has the same perceived loss of control but the amount of food, while considered by the individual to be overeating, is not unusually large (i.e., “Over the past 28 days, how many times have you had a sense of having lost control, but did not eat an unusually large amount of food?”). Secretive eating is queried as the frequency an individual ate in secret or furtively, and individuals are directed not to consider episodes of binge eating. We also examined whether participants experienced clinically-significant improvements in binge eating (objective binge-eating episodes) and LOC eating (objective and subjective binge-eating episodes), considering a change from “any” (at least one episode) to “no” (zero episodes) behavior to be clinically significant. Finally, we used the EDEQ global score, which has an established threshold of 2.77 (Dalle Grave et al., 2015; Hart et al., 2024), to evaluate the presence of clinically-significant eating-disorder psychopathology at end-of-treatment. We used weight measured by digital home scales and self-reported height to determine adolescents’ age- and sex-normed body mass index (BMI) z-score, BMI percentile, percent over median BMI, and percent over 95^th BMI percentile (Krukowski & Ross, 2020; Ogden et al., 2002). Patient Health Questionnaire-9 (PHQ) assessed depression (Kroenke et al., 2001; Richardson et al., 2010) (current study α=.87). Rosenberg Self-Esteem Scale (RSES) assessed self-esteem (French et al., 1995; Gray-Little et al., 1997; Rosenberg, 1965) (current study α=.90).

Data Analysis

Descriptive statistics quantified feasibility and acceptability, including treatment ratings (clinician, patient, parent) and total sessions attended, and clinical outcome variables, including both binge-eating episodes and LOC-eating episodes. Repeated measures analyses of variance (ANOVAs) tested change in variables from baseline to end-of-treatment and sensitivity analyses were performed restricted to those with a frequency threshold of ≥4 LOC-eating episodes per month. While the age-adapted definition of binge-eating disorder includes two episodes per month, full criteria specify four episodes per month; thus, this subsample analysis can provide clinically-relevant data for clinicians treating youth who meet the full-threshold frequency criterion. While power for exploratory efficacy aims was limited given the pilot nature of the study, the study was designed to have 80% power to detect between-group mean differences (d=0.8) and within-subject mean change within each group (d’=0.58) assuming a two-sided test at α=0.05 threshold.

RESULTS

Feasibility

We evaluated 109 participants for eligibility, of whom 55 were eligible and began treatment. Of these participants, 45 (82%) reported weekly LOC episodes. Five were enrolled in the open series, and 50 were randomized in the RCT. Six RCT participants dropped or were withdrawn from treatment (88% [n=44/50] treatment retention). Mean attendance was 15.3 of 16 sessions (SD=2.1). Figure 1 depicts participant recruitment and flow throughout the study.

Figure 1.

Participant flow throughout the study

*Psychiatric condition: participants reporting serious mental illness such as another eating disorder (anorexia, bulimia) and conditions requiring a higher level of care (e.g., suicidality, restrictive eating disorder). **ADHD stimulant medication, anti-obesity medication, or atypical antidepressant. ***Treatment for eating disorder or higher level of care. BMI=body mass index; CBT-LOC=Cognitive Behavioral Therapy for Loss of Control Eating [ ]=number meeting weekly frequency threshold of LOC episodes

Acceptability

At the end of the first treatment session, all participants (both conditions) thought the program made sense (M=5.26, SD=0.85; scale range 0–6, participant score range 3–6) and that their clinician seemed helpful (M=5.29, SD=0.99; scale range 0–6, participant score range 2–6). 100% of adolescents and 100% of parents were at least moderately confident that CBT-LOC would reduce binge eating (M=4.26, SD=1.04), and 94% of adolescents and 93% of parents were at least moderately confident that CBT-LOC would reduce weight (M=4.06, SD=1.21).

CBT-LOC had higher session satisfaction among patients, parents, and clinicians than NE (all ps<.001), as depicted in Figure 2A. Each individual session had high satisfaction, although there was some marginal variability, as depicted in Figure 2B.

Figure 2.

Acceptability of CBT-LOC and Nutrition Education Overall and By Session for the Full Sample (N=55)

Ado=Adolescent; CBT-LOC=Cognitive Behavioral Therapy for Loss of Control Eating. Parents attended sessions 1, 4, 8, 12, and 16.

Participant Demographic Characteristics

Participant demographic characteristics are reported in Table 1 overall and across treatment groups.

Table 1.

Baseline Demographic Characteristics of Participants

	All (N=55)		CBT-LOC (N=30)		Nutrition (N=25)
	M or n	SD or %	M or n	SD or %	M or n	SD or %
Age, M, SD	14.56	1.48	14.30	1.56	14.88	1.33
Sex
Male, n, %	22	40.0	11	36.7	11	44.0
Female, n, %	32	58.2	19	63.3	13	52.0
Non-binary, n, %	1	1.8	0	0	1	4.0
Race and Ethnicity
Non-Hispanic White, n, %	32	58.2	16	53.3	16	64.0
Non-Hispanic Black, n, %	10	18.2	5	16.7	5	20.0
Hispanic/Latinx White, n, %	5	9.1	3	10.0	2	8.0
Hispanic/Latinx Black, n, %	1	1.8	0	0	1	4.0
Multiracial or Another race, n, %	7	12.7	6	20.0	1	4.0
Sexual Orientation
Heterosexual	40	72.7	21	70.0	19	76.0
Homosexual	3	5.5	3	10.0	0	0
Bisexual	8	14.5	4	13.3	4	16.0
Another sexuality	4	7.3	2	6.7	2	8.0

Treatment Fidelity

Across all coded sessions, 80% set an agenda, 61% weighed the patient during the session, 84% reviewed self-monitoring logs, and 93% set a “home practice” goal. Average session length was 49.7 minutes. Regarding session-specific content, overall adherence to the manual was high, with 92–100% of broad session topics covered across quarters. Parent-adolescent sessions had 80–100% adherence, and individual sessions had 87–100% adherence.

Patient Outcomes

Table 2 summarizes baseline and end-of-treatment psychopathology and the effects of time and time-by-intervention interaction. Significant time-by-intervention interaction effects are depicted for the full sample (Figure 3) and the subsample of participants who engaged in weekly LOC-eating episodes at baseline (Figure 4).

Table 2.

Means, standard errors, and effects of time and the time-by-intervention interaction on patient outcomes

All Participants (N=55)
	Baseline				End-of-treatment				Time			Time * Condition
	CBT-LOC (n=30)		Nutrition Edu (n=25)		CBT-LOC (n=27)		Nutrition Edu (n=21)
	M	SE	M	SE	M	SE	M	SE	F	p	ηp2	F	p	ηp2
OBEs	9.28	1.37	6.00	1.57	1.32	0.78	3.53	0.89	33.00	<.001	.440	9.12	.004	.178
LOCs	12.28	1.65	10.79	1.90	1.88	1.43	6.05	1.64	38.62	<.001	.479	5.41	.02	.114
Secret	1.80	0.26	0.76	0.26	1.11	0.30	1.32	0.30	3.80	.06	.083	8.63	.005	.171
EDEQ Global	2.72	0.24	2.59	0.26	1.45	0.25	2.30	0.28	22.16	<.001	.340	8.73	.005	.169
EDEQ Overvaluation	3.54	0.40	3.43	0.44	1.88	0.40	2.55	0.44	19.90	<.001	.316	1.91	.17	.042
BMI-z	2.11	0.21	2.48	0.26	2.13	0.19	2.41	0.24	0.29	.59	.007	0.94	.34	.023
% median BMI	163.93	8.48	177.39	10.81	165.09	7.89	175.24	10.05	0.06	.80	.002	0.72	.40	.018
% 95th BMI percentile	118.09	6.20	129.19	7.91	118.90	5.80	127.77	7.40	0.05	.83	.001	0.63	.43	.015
Weight-z	2.09	0.15	2.42	0.19	2.12	0.15	2.38	0.19	0.05	.82	.001	1.17	.29	.028
PHQ Depression	10.71	1.35	8.15	1.39	4.52	0.91	5.30	0.94	25.88	<.001	.399	3.53	.07	.083
RSES Self-Esteem	15.38	1.27	16.25	1.31	18.52	1.43	19.25	1.47	14.95	<.001	.277	0.01	.93	<.001
Participants with Weekly LOC-eating episodes at Baseline (N=45)
	Baseline				End-of-treatment				Time			Time * Condition
	CBT-LOC (n=24)		Nutrition Edu (n=21)		CBT-LOC (n=21)		Nutrition Edu (n=17)
	M	SE	M	SE	M	SE	M	SE	F	p	ηp2	F	p	ηp2
OBEs	10.79	1.63	6.12	1.73	1.53	0.95	3.94	1.01	32.14	<.001	.486	12.33	.001	.266
LOCs	13.79	1.98	10.47	2.09	2.21	1.78	6.71	1.89	31.67	<.001	.482	8.21	.007	.195
Secret	2.53	0.45	2.00	0.47	0.84	0.32	1.41	0.34	3.91	.06	.103	8.59	.006	.202
EDEQ Global	2.78	0.27	2.43	0.28	1.73	0.28	2.31	0.30	16.66	<.001	.329	10.61	.003	.238
EDEQ Overvaluation	3.89	0.43	3.15	0.46	2.21	0.45	2.65	0.48	16.15	<.001	.322	4.75	.04	.123
BMI-z	2.05	0.21	2.48	0.26	2.08	0.19	2.49	0.23	0.22	.64	.007	0.02	.89	.001
% median BMI	163.14	8.78	177.25	10.89	164.54	8.24	178.36	10.22	0.33	.57	.010	<.001	.95	<.001
% 95th BMI percentile	116.95	6.29	129.08	7.80	117.90	5.93	130.03	7.36	0.36	.55	.011	<.001	1.00	<.001
Weight-z	2.06	0.15	2.41	0.18	2.10	0.14	2.42	0.18	0.41	.53	.013	0.13	.72	.004
PHQ Depression	11.38	1.47	7.53	1.43	5.00	1.05	5.53	1.02	19.44	<.001	.385	5.31	.03	.146
RSES Self-Esteem	15.44	1.48	17.00	1.44	18.06	1.61	19.12	1.57	9.65	.004	.237	0.11	.74	.004

Note. OBE=objective binge-eating episode; LOC=loss-of-control eating episode; Secret=secretive eating episode; EDEQ=Eating Disorder Examination Questionnaire; BMI=body mass index; PHQ=Patient Health Questionnaire; RSES=Rosenberg Self-Esteem Scale; CBT-LOC=Cognitive Behavioral Therapy for Loss of Control Eating

Figure 3.

Time-by-Intervention Interaction Effects for All Participants (N=55)

Note. OBE=objective binge-eating episode; LOC=loss-of-control eating episode; SBE=subjective binge-eating episode; EDEQ=Eating Disorder Examination Questionnaire; CBT-LOC=Cognitive Behavioral Therapy for Loss of Control Eating. All depicted interactions were statistically significant.

Figure 4.

Time-by-Intervention Interaction Effects for Participants with Weekly LOC-eating episodes at Baseline (N=45)

Note. OBE=objective binge-eating episode; LOC=loss-of-control eating episode; SBE=subjective binge-eating episode; EDEQ=Eating Disorder Examination Questionnaire; CBT-LOC=Cognitive Behavioral Therapy for Loss of Control Eating. All depicted interactions were statistically significant.

LOC-eating episodes.

There was a significant time-by-intervention interaction effect on objective binge-eating episodes (F(1,42)=9.12, p=.004, η_p²=.178), LOC-eating episodes (F(1,42)=5.41, p=.03, η_p²=.114), and secretive-eating episodes (F(1,42)=8.64, p=.005, η_p²=.171). For all eating episodes, CBT-LOC showed more improvement than NE. 52.0% [n=13/25] of CBT-LOC participants remitted LOC eating at end-of-treatment, compared to 35.0% [n=7/20] of NE participants (χ²(1)=1.30, p=.25).

The pattern of results remained the same among the subsample of participants engaging in LOC-eating episodes at least weekly at baseline: objective binge-eating episodes (F(1,34)=12.33, p=.001, η_p²=.266), LOC-eating episodes (F(1,34)=8.21, p=.007, η_p²=.195), and secretive-eating episodes (F(1,34)=8.59, p=.006, η_p²=.202). For all eating episodes, CBT-LOC showed more improvement than NE.

Eating-Disorder Psychopathology.

There was a significant time-by-intervention effect on global eating-disorder psychopathology (F(1,43)=8.73, p=.005, η_p²=.169). CBT-LOC participants improved more than NE participants in global eating-disorder psychopathology. However, the time-by-intervention interaction was not significant for overvaluation of weight/shape (F(1,43)=1.91, p=.17, η_p²=.042). For overvaluation, there was a significant main effect of time (F(1,43)=19.90, p<.001, η_p²=.316), but not intervention (F(1,43)=0.28, p=.60, η_p²=.007); CBT-LOC and NE participants both improved but did not differ significantly from each other.

Among the subsample of participants engaging in LOC-eating episodes at least weekly at baseline, there was a significant time-by-intervention effect on global eating-disorder psychopathology (F(1,34)=10.61, p=.003, η_p²=.238) and overvaluation of weight/shape (F(1,34)=4.75, p=.04, η_p²=.123). CBT-LOC participants had more improvement than NE participants in both global eating-disorder psychopathology and overvaluation of weight/shape.

Regarding global eating-disorder psychopathology, 88.0% [n=22/25] of CBT-LOC participants were below the clinical cut off for the global EDEQ scores at end-of-treatment, compared to 60.0% [n=12/20] of NE participants, which was a significantly different proportion (χ2(1)=4.72, p=.03).

Among the subsample of those who engaged in weekly LOC-eating episodes at baseline, 84.2% [n=16/19] of CBT-LOC participants were below the clinical cut off for the global EDEQ scores at end-of-treatment, versus 58.8% [n=10/17] of NE participants (χ2(1)=2.88, p=.09).

Weight Change.

There were no significant time-by-intervention effects, nor any main effect of time or intervention, on any of the weight measurements: BMIz (F(1,40)=0.94, p=.34, η_p²=.023), percent above the median BMI (F(1,40)=0.72, p=.40, η_p²=.018), percent above the 95^th BMI percentile (F(1,40)=0.63, p=.43, η_p²=.015), and weight z-score (F(1,40)=1.17, p=.29, η_p²=.028). There were also no significant effects of time, intervention, or time-by-intervention interaction in the subsample.

Depression and self-esteem.

There was not a significant time-by-intervention interaction effect on depression (F(1,39)=3.53, p=.07, η_p²=.083), nor self-esteem (F(1,39)=0.01, p=.93, η_p²<.001). However, time had a significant effect for both depression (F(1,39)=25.88, p<.001, η_p²=.399) and self-esteem (F(1,39)=14.95, p<.001, η_p²=.277). CBT-LOC and NE participants had significant improvements in depression and self-esteem but did not differ significantly from each other.

In the subsample of those who engaged in weekly LOC-eating episodes at baseline, there was a significant time-by-interaction effect on depression (F(1,31)=5.31, p=.03, η_p²=.146). CBT-LOC saw greater improvements than NE participants. The interaction of time-by-intervention was not significant for self-esteem (F(1,31)=0.11, p=.74, η_p²=.004); however, there was a significant effect of time (F(1,31)=9.65, p=.004, η_p²=.237). CBT-LOC and NE participants improved in self-esteem but did not differ significantly from each other.

DISCUSSION

This study tested the feasibility of CBT-LOC for adolescents, as well as acceptability and initial evidence of attaining clinical improvements. We followed a rigorous treatment development model (Carroll & Nuro, 2002; Onken et al., 2014; Rounsaville et al., 2001) to develop a preliminary version of the treatment and solicit adolescent, parent, and clinician feedback in an “open series” of patients, and then conduct a randomized controlled feasibility test of CBT-LOC against a credible condition comprising curated publicly-available material. Treatment credibility, satisfaction, and fidelity were high and CBT-LOC outperformed NE. Nearly all participants determined to be eligible began (90%) and completed (88%) treatment, with an average of 15.3 CBT-LOC sessions completed. Similarly, clinicians covered 92–100% of broad topics across treatment quarters.

Our controlled pilot study demonstrated potential evidence of CBT-LOC efficacy that warrants extension to a larger trial. We found that several key patient outcomes improved to a greater extent in CBT-LOC than NE, including objective binge-eating episodes, LOC-eating episodes, secretive-eating episodes, and global eating-disorder psychopathology. While our inclusion criteria used the age-adapted criteria for BED (2 LOC-eating episodes per month (Hilbert et al., 2020)), sensitivity analyses were performed restricted to those with a frequency threshold of ≥4 episodes per month (corresponding to the DSM-5 frequency threshold). The subgroup of participants with weekly or greater LOC-eating episodes at baseline had more pronounced improvements when receiving CBT-LOC compared to NE and had greater improvement in overvaluation of weight/shape and depression in CBT-LOC than NE. More research is needed to understand clinical staging of eating disorders in adolescence, as well as how to approach early intervention (Schmidt et al., 2025). In particular, the finding that overvaluation of weight/shape decreased among all participants with time and particularly improved among adolescents with weekly LOC-eating episodes is important. Overvaluation of weight/shape is considered a core aspect of eating disorders (Fairburn et al., 2003; Grilo, 2013; Wang et al., 2019) related to the persistence of disordered eating (Grilo et al., 2024). The decrease in global eating-disorder psychopathology and overvaluation of weight/shape suggests that clinically-meaningful outcomes are feasible with CBT-LOC, and CBT-LOC warrants testing in a larger-scale comparison against a viable comparator to build, potentially, empirical support.

Prior reports have suggested the feasibility of some CBT-based interventions for LOC eating specific to adolescents, albeit with some limitations. The current study extends the literature by focusing on a clinically-relevant population, adolescents (age 12–17) who are engaging in LOC eating but not purging and who also have excess weight. Our LOC eating abstinence rate (52%) is generally comparable to that reported in prior clinical trials testing CBT for LOC eating in adolescents (Hilbert et al., 2020) and adults (Grilo et al., 2025; Grilo et al., 2011). While our baseline LOC-eating episodes frequency was lower than other studies, the change from baseline to end-of-treatment (from 12.3 episodes at baseline to 1.9 episodes at end-of-treatment) was similar to DeBar and colleagues’ (2013) reported change from 19.8 to 1.8 episodes and to the changes reported by Jones and colleagues’ (2008) from 18.4 to 7.4 episodes. Moreover, considering the EDEQ global score threshold, 37% were below the clinical threshold at baseline and 88% were below the clinical threshold at end-of-treatment, compared to transdiagnostic eating disorder treatment for non-underweight adolescents, which had 26% under threshold at baseline and 82% under threshold at end-of-treatment (Dalle Grave et al., 2015). Notably, in prior work, Jones and colleagues (2008) observed a significant reduction in BMI z-score but other studies (Debar et al., 2013; Hilbert et al., 2020), like ours, did not. This may be because the Jones study considered overeating (no LOC) and LOC eating as part of the inclusion criteria, not just LOC eating, and because their study was geared towards healthy weight maintenance.

Limitations and Future Directions

Preliminary efficacy findings should be interpreted cautiously given study limitations, including the relatively small sample size and the difference in time spent with the clinician between conditions (weekly interactions versus monthly interactions with weekly materials). While this initial trial used nutrition education, a face-valid and ecologically-valid comparator, it is possible that differences in time, intensity, or format might have influenced findings in unknown ways. An important next step will be a definitive extension and test of the efficacy of CBT-LOC in a larger, adequately-powered RCT that includes a stronger comparator, such as intensive health behavior and lifestyle treatment (Hampl et al., 2023; Kelly et al., 2024). Research is needed to establish evidence for efficacy and effectiveness and to inform recommendations for adolescents and their families who may seek treatment for LOC eating, weight, or both.

While one approach to future research would be to enhance the potential of CBT-LOC to change weight, there is also value in having multiple treatments so that some patients, families, and providers can make informed and personalized decisions about treatments to pursue. It is well-established that eating disorders have low rates of help-seeking (Sanzari & Liu, 2019) and it is possible that having multiple treatment options could yield multiple pathways to treatment. Future research should also test potential mediators and moderators of treatment outcomes (e.g., psychiatric comorbidities and medical comorbidities that may influence responses to interventions) and evaluate participants for longer periods of time following the intervention to evaluate durability and relapse.

Conclusion

In conclusion, the current study was an initial evaluation of CBT-LOC tailored to adolescence, to treat LOC eating. Our sample included adolescents with racial/ethnic and gender diversity, and a substantial subgroup were engaging in weekly LOC-eating episodes. Feasibility, acceptability, and evidence supporting improvements in patient outcomes were promising. Future research should conduct definitive tests of efficacy and effectiveness, as well as test potential mediators, moderators, and mechanisms of change.

Clinical Impact Statement.

Question:

How can we treat binge eating experienced by adolescents?

Findings:

This study showed that cognitive behavioral therapy for adolescent loss-of-control eating produced greater improvements in eating behaviors than those who received nutrition education, but weight change was minimal.

Meaning:

Binge eating is a prevalent and costly mental health problem that often begins in adolescence, but while evidence-based therapies have been effective treatments for adults, there is no clear standard of care for adolescents. This study offers a promising new treatment option.

Next Steps:

Additional research, with a larger number of participants, needs to test whether the treatment works well on patients’ desired outcomes, and needs to test whether the treatment works better for some patients than others based on different personal characteristics.