Dietetic‐Led Cognitive‐Behavioral Therapy for Avoidant/Restrictive Food Intake Disorder: Feasibility, Acceptability, and Proof‐of‐Concept

ABSTRACT

Objective

Cognitive‐Behavioral Therapy for Avoidant/Restrictive Food Intake Disorder (CBT‐AR) is an emerging treatment for ARFID, but to date, it has only been delivered by mental health clinicians. Given the relevance of nutrition to treatment goals, we adapted the treatment to be dietetic‐led (D‐CBT‐AR).

Method

We conducted a pilot trial of D‐CBT‐AR. Participants self‐referred to a multidisciplinary eating‐disorder clinic in Brisbane, Australia, and attended in person or virtually. Participants rated D‐CBT‐AR on acceptability and feasibility. We also measured changes in ARFID psychopathology, Body Mass Index, and co‐morbid psychopathology.

Results

Of the 12 participants offered D‐CBT‐AR, 92% (n = 11) took up the treatment, and 72% (n = 8) completed, with a median treatment length of 33 weeks. The majority of completers rated D‐CBT‐AR above benchmarks on feasibility (89%) and acceptability (78%). Using intent‐to‐treat analyses, participants reported large and significant improvements in ARFID psychopathology on all subscales of the Pica, ARFID and Rumination Disorder Interview. They incorporated a median of 26 new foods. For those who were weight‐suppressed, Body Mass Index increased by 1.5 bands from pre‐ to post‐treatment, though effects were non‐significant. Participants also reported large and significant reductions in clinical impairment, depression, and stress. Additionally, at 6‐months post‐treatment, five participants met remission criteria.

Discussion

The trial shows proof‐of‐concept for the feasibility and acceptability of D‐CBT‐AR. Although requiring replication in larger and randomized controlled trials, our findings highlight the potential of D‐CBT‐AR to expand the types of professionals who could effectively deliver CBT‐AR, thereby increasing access to care for individuals living with ARFID.

Summary

• Eleven adults with ARFID participated in a feasibility and acceptability trial of dietetic‐led cognitive‐behavioral therapy with a completion rate of 72%.

• Participants provided high ratings of satisfaction, acceptability, appropriateness, and feasibility of dietetic‐led cognitive‐behavioral therapy for ARFID.

• In an intent‐to‐treat analysis, participants reported large and significant reductions in their ARFID psychopathology from pre‐ to post‐treatment and incorporated a median of 26 new foods. Improvements were sustained at 6‐month follow‐up.

1. Introduction

Avoidant/restrictive food intake disorder (ARFID) is a feeding and eating disorder marked by inadequate dietary volume and/or variety due to sensory sensitivity, fear of aversive consequences, and/or lack of interest in food or eating (American Psychiatric Association ²⁰²²; Thomas et al. ²⁰¹⁷). The Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM‐5, American Psychiatric Association ²⁰²²) distinguishes ARFID from traditional eating disorders, such as anorexia nervosa, bulimia nervosa, and binge‐eating disorder, as body image concerns are not the primary driver of food restriction. The variance in etiology suggests that ARFID requires a treatment approach that is distinct from other eating disorders which focus on body image and fear of weight gain (Bryant‐Waugh et al. ²⁰²¹; Fonseca et al. ²⁰²⁴; Monteleone et al. ²⁰²²). Currently, the most studied treatment for adults with ARFID is cognitive‐behavioral therapy for ARFID (CBT‐AR) (Willmott et al. 2024).

CBT‐AR has strong proof of concept (Thomas et al. 2020, 2021) with clinical integration into real world treatment settings (Hellner et al. 2025; MacDonald et al. ²⁰²⁴). CBT‐AR is a flexible manualised treatment created for children, adolescents, and adults with ARFID aged 10 years and above who can tolerate oral food intake (Thomas and Eddy 2019). Case studies and series have extended the generalizability of CBT‐AR to diverse clinical groups of individuals with possible high rates of co‐occurring disorders, such as those with autism (Lien et al. 2025; Price et al. ²⁰²⁴; Sader et al. ²⁰²⁵) and disorders of the gut‐brain interaction (Burton Murray et al. 2023; Rezaei et al. ²⁰²⁵). There is also research supporting its delivery through telehealth (Burton Murray et al. 2022; Hellner et al. ²⁰²⁵; Price et al. ²⁰²⁴) and across levels of care (MacDonald et al. ²⁰²⁴).

Thus, the aims of the present study were firstly to evaluate the feasibility and acceptability of dietitian‐led CBT‐AR (D‐CBT‐AR), and secondly to provide proof‐of‐concept for potential efficacy, hypothesizing that a dietitian could serve as the primary clinician. For aim one, we hypothesized that (a) the majority of participants offered D‐CBT‐AR would accept treatment, (b) the majority of those who initiated treatment would complete it, and (c) the majority of participants would provide ratings above predetermined benchmarks on satisfaction, acceptability, appropriateness, and feasibility.

For aim two, we hypothesized that (a) we would observe large reductions in ARFID psychopathology from pre‐ to post‐treatment, (b) participants would meet fewer DSM‐5‐TR criteria for ARFID at post‐treatment, (c) participants would incorporate several new foods from pre‐ to post‐treatment, (d) the subset of participants who were weight‐suppressed would show a significant increase in body mass index (BMI), (e) the intervention would improve ARFID symptoms without worsening eating disorder psychopathology, (f) participants would show reductions in clinical impairment, depression, anxiety, and stress from pre‐ to post‐treatment, and (g) all improvements would be sustained from post‐treatment to 6‐month follow‐up.

2. Method

2.1. Study Design

We conducted a feasibility pilot trial of D‐CBT‐AR using manualised CBT‐AR with additional dietetic resources that we created for the current study as outlined in Winten et al. (2025). These resources included macro and micronutrient educational worksheets and alterations to the plate visual image. The core CBT‐AR treatment was not modified, but we added dietetic resources to strengthen its implementation into dietetic practice. We conducted the study in accordance with the Declaration of Helsinki and received approval from the Queensland University of Technology Human Research Ethics Committee (2024‐8273‐18314). We pre‐registered the trial on the Australia and New Zealand Clinical Trials Registry (ACTRN12622000860763). Participants received evaluation and treatment without financial expense and a General Practitioner (GP) within Queensland University of Technology‐Eating Disorders Clinic also offered medical evaluation and monitoring at no cost. We observed no adverse events related to treatment during the trial.

2.2. Setting

At the time of the study, the Queensland University of Technology Eating Disorders Clinic comprised clinical psychologists, a GP, a nurse, dietitians, psychology interns, and dietetic interns. C.G.W. (the dietitian) provided D‐CBT‐AR through the clinic and prescribed appropriate oral nutrition supplements at therapeutic doses, if required. The clinic held weekly multidisciplinary team meetings to review each participant's care. Each participant's GP reviewed the requested blood tests and assessed medical stability, at a minimum of monthly, using the Queensland Eating Disorder Service inpatient admission criteria (Queensland Eating Disorders Service ²⁰²⁰). Postgraduate psychology interns, supervised by E.S., assessed psychological diagnosis and risk at the start of the study. Participants were able to attend in person, via telehealth, or a combination. For telehealth visits, we used video conferencing software (Zoom Communications, Ca, USA) and followed telehealth for eating disorder treatment recommendations outlined by Waller et al. (2020) for example, ensuring privacy when conducting telehealth visits and ensuring self‐weighing on home scales during session when appropriate.

2.3. Participants

We recruited participants between July 2023 and July 2024 through word of mouth and by advertising to eating disorder health professionals in Queensland via social media and emails. Treatment was advertised as dietetic‐led within a multidisciplinary team. Participants had to meet the following inclusion criteria: age of 18 years or older, able to provide informed consent, medically stable as per the Queensland Eating Disorders Service (2020) criteria, consent to pausing other ongoing therapies, and a confirmed diagnosis of ARFID via Structured Clinical Interview for DSM‐5 (SCID‐5) (First et al. 2015). Participants were not eligible if they had an active diagnosis of another eating disorder, another condition significantly impacting their oral nutrition (e.g., Crohn's, substance misuse), or prior CBT‐AR treatment. Additionally, as per the CBT‐AR manual, participants needed to be receiving nutrition orally and not entirely reliant on enteral feeding.

Initially, C.G.W. screened for ARFID using the Nine Item ARFID Screen (Zickgraf and Ellis 2018), and other eligibility criteria, followed by the Eating Disorders module of the SCID‐5 (First et al. 2015) conducted by a psychology intern.

Of the 16 adults who self‐referred for D‐CBT‐AR, 12 met criteria for inclusion in the trial. Details are outlined in the participant flow diagram (Figure 1).

FIGURE 1.

Flow diagram for dietetic‐led cognitive‐behavioral therapy for avoidant/restrictive food intake disorder (CBT‐AR) participant recruitment. AAN, atypical anorexia nervosa; PTSD, post‐traumatic stress disorder.

2.4. Treatment

CBT‐AR is a flexible, manualised treatment of 20–30 weekly 50‐minutes sessions. Treatment can be offered in individual or family‐supported formats; for the purpose of this study participants were provided the option of their preferred format. We followed the manualized treatment protocol (Thomas and Eddy 2019), with the addition of dietetic‐specific worksheets and an Australian‐specific food list that we added to the participant workbook (see Winten et al. ²⁰²⁵ for additional D‐CBT‐AR content). The additional content included nutrition‐specific education for Stage 2, that is, a visual of micronutrients' impact on the body to go alongside the nutrient list, and worksheets for Stage 3 that could optimize engagement, that is, a neutral words worksheet to support participants to describe a novel food for the sensory sensitivity module. Treatment occurred in four stages: Stage 1 involved psychoeducation and early behavior change (2–4 sessions), Stage 2 was treatment planning (2 sessions), Stage 3 consisted of addressing the phenotypic presentations (14–22 sessions), and Stage 4 involved relapse prevention (2 sessions).

2.5. Treatment Fidelity

Consistent with the treatment manual, participants deemed weight stable (had a BMI > 18.5 kg/m² and no symptoms of malnutrition) were offered 20 sessions, and participants deemed weight‐suppressed (i.e., current BMI < 18.5 kg/m², chronic inability to attain expected adult weight as determined by dietetic assessment, or with > 5% weight loss in the last 6 months, despite current BMI > 18.5 kg/m²) were offered up to 30 sessions. For weight‐suppressed participants, weight gain was the focus for Stages 1–2 and continued throughout Stage 3. To support weight gain, participants were encouraged to increase the volume of their preferred foods. Three participants chose family‐supported CBT‐AR as they were young adults living at home who were either weight‐suppressed (n = 2) or had a neurodevelopmental disorder which required greater parental support (n = 1). All participants were invited to involve significant others, family members, or caregivers in their individual treatment, with four participants deciding to do so. In family‐supported CBT‐AR, family actively participated in treatment for example by guiding and outlining meals and/or participating in exposures. In individual treatment the participant's support network and family were still key for support and consistency, and their level of engagement was determined by the participant. After every session, C.G.W. completed the fidelity checklist from the CBT‐AR manual to ensure she had covered all necessary session elements, and C.G.W. met with J.J.T. weekly to discuss therapy fidelity. E.S., G.H., and L.J.R. provided additional supervision to C.G.W. C.G.W. is an Accredited Practising Dietitian and Credentialed Eating Disorder Clinician‐Dietitian and completed a formal one‐day training in CBT‐AR prior to designing the study.

2.6. Measures

We sent pre‐treatment, post‐treatment, and follow‐up questionnaires to participants to complete the week before starting treatment, the week after completing treatment, and the week that participants attended the 6‐month follow‐up visit. We collected all data via REDcap (Research Electronic Data Capture), hosted at Queensland University of Technology. One participant completed paper forms, which we transcribed into the REDcap database. Participants self‐reported their demographic characteristics, including participant age, gender, ethnic background, education level, relationship status and income level at pre‐treatment. They also self‐reported psychiatric and medical co‐diagnoses previously conferred by other clinicians, current medications, and completed a screening tool to assess for autism traits, the AQ‐10 (Allison et al. 2012) at pre‐treatment and follow‐up. Scores above 6 highlight clinically significant traits that require further assessment for a possible diagnosis of autism spectrum disorder. Cronbach alpha in this sample was 0.78.

2.6.1. Feasibility and Acceptability Measures

To measure feasibility and acceptability, we calculated the number of participants who completed the treatment and also assessed the percentage who scored above established benchmarks on the Client Satisfaction Questionnaire‐8 (CSQ‐8) (Larsen et al. 1979). Feasibility of Intervention Measure (FIM) (example item: “CBT‐AR seems possible”), Acceptability of Intervention Measure (AIM) (example item: “CBT‐AR meets my approval”), and Intervention Appropriateness Measure (IAM) (example item: “CBT‐AR seems suitable”) (Weiner et al. 2017). Higher scores indicate greater participant favorability; we used the threshold of 16 and above for each measure to indicate strong favorability as suggested in Watkins et al. (2024). The CSQ is a validated measure with scores above 26 indicating high satisfaction (Attkisson and Zwick 1982; Larsen et al. ¹⁹⁷⁹). Cronbach alpha for this sample was 0.87. The Cronbach alphas in our sample were 0.97, 0.98, and 0.77, respectively.

2.6.2. ARFID Psychopathology Measures

We measured change in ARFID symptoms with the Pica, ARFID, and Rumination Disorder Interview (PARDI, Bryant‐Waugh et al. ²⁰¹⁹), at pre‐treatment, post‐treatment, and 6‐month follow‐up. The PARDI assessed overall severity, and the severity of phenotypic presentations of sensory sensitivity, lack of interest, and fear of aversive consequences. For the overall severity and phenotypes the scores range from 0 to 6 with 0 representing no symptoms and 6 representing severe symptoms. The PARDI also identifies which DSM‐5 criteria for ARFID are met. C.G.W. conducted the PARDI and J.J.T. reviewed results. We also administered the PARDI ARFID Questionnaire (PARDI‐AR‐Q, Bryant‐Waugh et al. ²⁰²²) and Food Neophobia Scale (FNS, Pliner and Hobden ¹⁹⁹²) at pre‐treatment, post‐treatment, and follow‐up. The FNS measures food neophobia, or reluctance to try novel foods. The PARDI‐AR‐Q was shortened to assess phenotype severity change across treatment and FNS altered for cultural competency as previously described in Hellner et al. (2025) and Miles et al. (2025). Cronbach alpha for the PARDI‐AR‐Q was 0.85 and FNS was 0.62 in the current sample. C.G.W. collected metric physical measures of weight, height, BMI, in person or via self‐report on telehealth. For weight‐suppressed participants, we measured weight weekly to meet the treatment goal of weight gain.

Participants recorded the increased variety in dietary intake by the number of foods being learned about, actively being incorporated into their diet, and fully incorporated on a CBT‐AR worksheet. To qualify as a new food incorporated by the end of treatment, participants needed to report that they were regularly eating the new food as part of their usual diet at an appropriate portion size.

2.6.3. Comorbid Psychopathology Measures

We assessed comorbid psychopathology using the Eating Disorders Examination Questionnaire (EDE‐Q, Fairburn and Beglin ²⁰⁰⁸), Clinical Impairment Assessment (CIA, Bohn et al. ²⁰⁰⁸), and Depression Anxiety Stress Scale‐21 (DASS‐21, Lovibond and Lovibond ¹⁹⁹⁵), at pre‐treatment, post‐treatment, and 6‐month follow‐up. We used the EDE‐Q to confirm the lack of other eating disorder pathology at baseline, and also to confirm that non‐ARFID eating disorder symptoms did not increase during the course of dietetic‐led CBT‐AR. Higher scores are more severe for the global EDE‐Q and all subscales (Mond et al. 2006; Smith et al. ²⁰¹⁷). A global score of > 3 is clinically significant, as in Australia this is a criterion for government funded supports (McLean and Fuller‐Tyszkiewicz ²⁰²⁴). Cronbach alpha for the EDE‐Q global was 0.81 in the current sample. We used the CIA, a 16‐item self‐report measure, to ascertain psychosocial impairment secondary to eating disorder pathology. The CIA cut‐off for clinical impairment indicating a likely eating disorder is 16 and above (Bohn et al. 2008). Cronbach alpha for our sample was 0.89. The DASS‐21 clinical cut‐off for mild scores of depression, anxiety and stress are above 9, 7, and 14, respectively (Lovibond and Lovibond 1995; Shaw et al. ²⁰¹⁷).

2.7. Data Analysis

We conducted intent‐to‐treat analysis, meaning that when participants did not complete treatment, we used their score at last assessment carried forward. We conducted all analyses in SPSS V30.00 (IBM, NY, USA). Due to the small sample size and skewed nature of the data, we used nonparametric measures to evaluate our hypotheses. We used Wilcoxon signed‐rank test, with scores reported as Z‐scores, with significant scores above Z = ±1.96 and p < 0.05. For effect sizes, we used biserial rank correlation (r _rb ), which we interpreted effect sizes as small (0.10), medium (0.30), or large (> = 0.50) (Cohen 2013).

First, to test the hypothesis that D‐CBT‐AR would be feasible and acceptable, (a) we calculated the percentage of participants offered treatment who accepted it, and (b) completed the treatment, as well as ascertaining the mean number of sessions and weeks of treatment. Finally, for (c) we calculated the percentage of participants who scored above predetermined benchmarks for the CSQ, FIM, AIM, and IAM.

Second, to test the hypothesis that D‐CBT‐AR would reduce ARFID psychopathology we (a) conducted Wilcoxon signed‐rank test comparing pre‐ and post‐ outcome measures of PARDI, PARDI‐AR‐Q, and FNS, and we calculated the effect size using biserial rank correlation, and (b) we reviewed the percentage of DSM‐5 criteria met as per the PARDI at pre‐ versus post‐treatment. For (c) using intent‐to‐treat data, we calculated the number of new foods incorporated post‐treatment, using the food list worksheet, with one participant who did not complete the food list worksheet, and this participant was included as zero. To test (d) we analyzed weight change in weight‐suppressed participants with BMI from pre‐ to post‐treatment using Wilcoxon signed‐rank tests with biserial rank correlation. For (e) to evaluate the possibility that the eating disorder psychopathology would increase during D‐CBT‐AR we conducted Wilcoxon signed‐rank tests with biserial rank correlation for the EDE‐Q. Further, to assess (f) pre‐ to post‐treatment improvements in clinical impairment, depression, anxiety, and stress symptoms would improve, as measured by the CIA and DASS‐21, we conducted Wilcoxon signed‐rank tests with biserial rank correlation. Finally, to assess (g) whether improvements were sustained at 6‐month follow‐up, we used Wilcoxon signed‐rank tests to compare post‐treatment versus 6‐month data.

3. Results

Table 1 displays the demographic and clinical characteristics of the 11 participants who started treatment. Participants were two‐thirds female, had a mean age of 25.4 years, and 55% reported a European ancestry. The mean duration of illness was 17.3 years (SD = 14.1). Nearly two‐thirds (n = 7, 64%) screened positive for possible autism spectrum disorder on the AQ‐10 at start of treatment.

TABLE 1.

Characteristics and demographics of participants in pilot trial of dietetic‐led cognitive behavioral therapy for ARFID (n = 11).

	M (SD) or n (%)
Age	25.4 (9.3)
Gender
Male	4 (36%)
Female	7 (64%)
Ancestry/ethnicity
European	6 (55%)
Oceanian	1 (9%)
Asian	3 (27%)
Other	1 (9%)
Relationship status
Prefer not to answer	1 (9%)
Single	5 (46%)
In a relationship	3 (27%)
Married	2 (18%)
Income level per annum ($AUD)
Prefer not to answer	1 (9%)
0–29,000	5 (46%)
30,000‐49,000	1 (9%)
50,000‐79,000	4 (36%)
Highest education level
Grade 12	5 (46%)
Diploma	2 (18%)
Apprenticeship/trade	1 (9%)
Bachelor's degree	2 (18%)
Post‐graduate degree	1 (9%)
ARFID presentation a
Sensory sensitivity only	2 (18%)
Lack of interest only	0
Fear of aversive consequences only	1 (9%)
Sensory sensitivity + lack of interest	5 (46%)
Lack of interest + fear of aversive consequences	2 (18%)
All 3 presentations	1 (9%)
Duration of ARFID (years)	17.3 (14.1)
Weight status (last 6 months)
Weight suppressed	8 (73%)
Weight stable	3 (27%)
Treatment format
Family	3 (27%)
Individual	8 (73%)
Have experienced food insecurity in past 12 months	3 (27%)

Abbreviation: M, mean.

^a From PARDI scores and clinical judgment.

3.1. Aim 1: Feasibility and Acceptability

Consistent with our hypothesis that dietetic‐led CBT‐AR would be feasible and acceptable, nearly all (i.e., 11 of 12, 92%) participants who were offered D‐CBT‐AR elected to take up the treatment. The only participant who did not accept the offer of D‐CBT‐AR cited prioritization of other concurrent psychological concerns. Of those who began treatment (n = 11), nearly three quarters (n = 8, 73%) completed it. One participant withdrew during treatment due to low personal readiness and did not proceed with Stage 4. A further two withdrew due to poor engagement in treatment, characterized by limited and inconsistent attendance. All participants (100%) who completed treatment completed the 6‐month post‐treatment assessment. The completers attended a mean of 24.4 (SD = 3.9) sessions over a mean of 32.6 (SD = 4.4) weeks.

Participants who completed all post‐treatment assessments (n = 9), outlined in Table 2, provided high satisfaction ratings, with 89% scoring above the CSQ benchmark. Participants also reported high ratings of feasibility (89% above FIM benchmark), acceptability (78% above AIM benchmark), and appropriateness (89% above IAM benchmark).

TABLE 2.

Post‐treatment measures of feasibility and acceptability for dietetic‐led cognitive‐behavioral therapy for ARFID.

Feasibility, acceptability measures (n = 9)	Post M	Post IQR
Client Satisfaction Questionnaire (CSQ)	31	26, 32
Feasibility of Intervention Measure (FIM)	18	16, 19.5
Intervention Appropriateness Measure (IAM)	18	15.5, 20
Acceptability of Intervention Measure (AIM)	16	16, 19.5
Number of sessions (completers only, n = 8)	24	21, 29
Treatment length (weeks) (completers only, n = 8)	33	29, 36

Abbreviations: IQR, interquartile range; M, median.

3.2. Aim 2: Proof‐of‐Concept of D‐CBT‐AR

3.2.1. ARFID Psychopathology

Table 3 displays pre‐treatment, post‐treatment, and 6‐month follow‐up results on ARFID psychopathology outcome measures. As hypothesized, PARDI scores decreased significantly from pre‐ to post‐treatment with large effect sizes for sensory sensitivity, lack of interest, fear of aversive consequences, and overall ARFID severity (Z = −2.52, p = 0.012, r _rb = −0.89 for each respectively). Also as hypothesized, scores reduced significantly with large effect sizes from pre‐ to post‐ treatment on the FNS (Z = −2.81, p = 0.005, r _rb  = −0.089) and for one subscale of the PARDI‐AR‐Q (lack of interest, Z = −2.05, p = 0.04, r _rb  = −0.72). The PARDI‐AR‐Q severity, sensory sensitivity and fear of aversive consequences subscales also decreased with all having a medium effect size but the change was not statistically significant. Results did not significantly change between post‐treatment to follow‐up across any of the ARFID psychopathology measures, suggesting that improvements were sustained over the 6‐month‐follow‐up period.

TABLE 3.

Pre‐treatment, post‐treatment, and 6‐month follow‐up measures for ARFID psychopathology measures for dietetic‐led cognitive‐behavioral therapy for ARFID.

ARFID psychopathology measures	Subscale	Pre M	Pre IQR	Post M	Post IQR	Z (pre‐post)	p (pre‐post)	r rb (pre‐post)	r rb 95% CI (pre‐post)	FU M	FU IQR	Z (post‐FU)	p (post‐FU)
Pica ARFID, and Rumination Disorder Interview (n = 11)	Severity	3.9	2.2, 4.1	1.6	0.8, 2.2	−2.52	0.01	−0.89	−0.9, −0.02	1.5	0.9, 2.1	−0.18	0.86
	SS	2.7	2.5, 3.6	1.3	0.5, 2.5	−2.52	0.01	−0.89	−0.9, −0.02	0.8	0.9, 2.4	−0.42	0.67
	LOI	3.2	2.5, 4.2	1.4	0.9, 2.6	−2.52	0.01	−0.89	−0.9, −0.02	2.1	0.9, 2.7	−0.30	0.77
	FOA	1.6	0.4, 4.5	0.5	0, 1.1	−2.52	0.01	−0.89	−0.9, −0.02	0.5	0, 1.1	−0.28	0.78
Pica, ARFID, and Rumination Disorder Interview—ARFID Questionnaire (n = 11)	Severity	3.0	1.5, 3.5	2.0	1.0, 2.3	−1.16	0.25	−0.41	−0.77, 0.44	2	1, 2.3	−1.28	0.20
	SS	3.0	2.7, 4.0	2.7	1.0, 3.33	−1.36	0.18	−0.48	−0.79, 0.38	2	0.7, 3.7	−0.53	0.59
	LOI	3.33	1.7, 4.7	2.0	0.3, 3.3	−2.05	0.04	−0.72	−0.86, 0.16	2	0.7, 3	−0.05	0.96
	FOA	1.3	0, 3.7	1.0	0, 2.3	−1.44	0.15	−0.51	−0.8, 0.35	0.7	0, 1.3	−4.59	0.65
Food Neophobia Scale (n = 11)		36	33, 39	30	25, 34	−2.81	> 0.01	−0.89	−0.9, −0.02	27	23, 31	−2.34	0.02
Foods incorporated (n = 11)				26	17, 44

DSM‐5 criteria met for ARFID based on PARDI (A1‐A4) (n = 11)	Pre n (%)	Post n (%)	FU n (%)
A1. Low weight (BMI < 18.5), significant weight loss, failure to grow as expected	7 (64)	2 (18)	2 (18)
A2. Nutrition deficiency (diagnosed by blood pathology)	11 (100)	6 (55)	5 (45)
A3. Dependence on nutritional supplements (oral nutrition supplements or micronutrient supplements)	4 (36)	2 (18)	1 (9)
A4. Psychosocial impairment (score 4+ on PARDI impairment items)	10 (91)	8 (72)	6 (55)
Remission (post‐treatment)		3 (27)	5 (45)

Note: Intent‐to‐treat analysis completed for n = 11, with last assessment carried forward.

Abbreviations: CI, confidence interval; FOA, fear of aversive consequences; FU, follow up; IQR, interquartile ranges; LOI, lack of interest; M, median; p, p‐values; r, biserial rank correlation; SS, sensory sensitivity; Z, Z‐score.

Using intent‐to‐treat measures with last assessment carried forward, participants met fewer DSM‐5‐TR (American Psychiatric Association ²⁰²²) criteria at post‐treatment than at pre‐treatment. Specifically, criterion A1 (weight suppression) reduced from 64% of participants meeting the criterion to 18%, criterion A2 (nutrient deficiency) reduced from 100% of participants to 55%, criterion A3 (dependence on nutrition supplement) reduced from 36% of participants to 18%, and criterion A4 (psychosocial impairment) reduced from 91% of participants to 72%. Of the 11 participants who began D‐CBT‐AR, more than one quarter (n = 3) met PARDI criteria for remission at post treatment. Also of note, participants added a median of 26 foods to their diets by end of treatment. By follow‐up, 5 (45%) participants were in remission, with 18% continuing to meet criteria A1, 45% of participants meeting criterion A2, 9% meeting criterion A3, and 55% meeting criterion A4, as per intent‐to‐treat analysis.

3.2.2. Physical Health Measures

We categorized eight participants as weight‐suppressed. Table 4 presents the BMI results which contrasted with our initial hypothesis. The median BMI of weight‐suppressed participants increased numerically by 1.5 BMI bands from pre‐ to post‐treatment; the increase was not statistically significant (Z = 1.12, p = 0.26, r _rb  = −0.04). The BMI of weight stable participants did not change significantly during treatment. At follow‐up, the weight‐suppressed median increased 1.3 bands from post‐treatment, and weight stable median 1.1 bands from post‐treatment, but neither group saw significant changes during the 6‐month follow‐up period.

TABLE 4.

Pre‐treatment, post‐treatment, and 6‐month follow‐up measure of physical health for dietetic‐led cognitive‐behavioral therapy for ARFID.

Physical measures	Pre M	Pre IQR	Post M	Post IQR	Z (pre‐post)	p (pre‐post)	r rb (pre‐post)	r rb 95% CI (pre‐post)	FU M	FU IQR	Z (post‐ FU)	p (post‐FU)
Body Mass Index, weight suppressed (kg/m2) (n = 8)	18.2	17.5, 25.5	19.7	17.6, 25.2	1.12	0.26	0.49	−0.24, 0.88	21.0	17.5, 21.3	−0.85	0.40
Body Mass Index, weight stable (kg/m2) (n = 3)	25.8	21.6, 28	25.8	21.5, 27.9	1.60	0.11	0.92	N/A	26.9	23.5, 33.2	−1.60	0.11

Note: Intent‐to‐treat analysis completed for n = 11, with last assessment carried forward.

Abbreviations: CI, confidence interval; FU, follow up; IQR, interquartile ranges; M, median; N/A, not applicable; p, p‐values; r, biserial rank correlation; Z, Z‐score.

3.2.3. Comorbid Psychopathology

Table 5 showcases the results of co‐morbid psychopathology. As hypothesized, EDE‐Q scores did not change significantly from pre‐ to post‐treatment or at follow‐up and stayed below cut‐offs for clinically significant eating disorder psychopathology at all time points. Also in line with hypotheses, clinical impairment as measured by the CIA decreased significantly from pre‐ to post‐ treatment (Z = −1.96, p = 0.05, r _rb = 0.69) with a large effect size, moving from above the clinical cut‐off at pre‐treatment (M = 18) to below the clinical cut‐off at post (M = 11) and remained below at follow‐up (M = 7). For depression, anxiety, and stress the DASS‐21 showed reduction in scores, with a significant large effect size. The subscale of depression reduced significantly with a large effect size (Z = −2.23, p < 0.05, r _rb  = 0.91). The stress subscale reduced significantly, highlighting a large effect size (Z = −1.98, p < 0.05, r _rb = 0.70). The anxiety subscale did not change pre to post treatment (Z = −0.17, p = 0.864, r _rb  = 0.06). At follow‐up the DASS‐21 subscales had not changed significantly and one less person screened positive for autism using the AQ‐10 (n = 6, 54%).

TABLE 5.

Pre‐treatment, post‐treatment, and 6‐month follow‐up measures of co‐morbid psychopathology for dietetic‐led cognitive‐behavioral therapy for ARFID.

Co‐morbid psychopathology measures	subscale	Pre M	Pre IQR	Post M	Post IQR	Z (pre‐post)	p (pre‐post)	r rb (pre‐post)	r rb 95% CI (pre‐post)	FU M	FU IQR	Z (post‐FU)	p (post‐FU)
Clinical Impairment Assessment (n = 11)		18	10, 24	11	4, 17	−1.96	0.05	−0.69	−0.86, 0.18	7	2, 18	−0.76	0.45
Eating Disorder Examination‐Questionnaire (n = 10)	Global	0.3	0.1, 0.8	0.2	0, 0.4	−1.57	0.12	−0.50	−0.75, 0.24	0.14	0.00, 0.8	−0.85	0.40
Depression Anxiety Stress Scale‐21 (n = 11)		18	7, 28	8	6, 22	−1.68	0.09	−0.60	−0.83, 0.27	9	4, 19	−0.51	0.61
	Depression	5	2, 9	3	1, 6	−2.23	0.03	−0.91	−0.70, 0.22	2	1, 4	−0.76	0.45
	Anxiety	3	2, 7	2	1, 7	−0.17	0.86	−0.06	−0.66, 0.72	4	2, 6	0.00	1.0
	Stress	7	5, 10	6	2, 10	−1.98	< 0.05	−0.70	−0.86, 0.17	5	2, 10	−0.51	0.61

Autism Quotient Score‐10 (n = 11)	Pre n (%)	Post n (%)
Met threshold (score > 6)	7 (64)	6 (54)

Note: Intent‐to‐treat analysis completed for n = 11, with last assessment carried forward.

Abbreviations: CI, confidence interval; FU, follow up; IQR, interquartile ranges; M, median; p, p‐values; r, biserial rank correlation; Z, Z‐score.

4. Discussion

This pilot study is the first evaluation of the feasibility, acceptability, and proof‐of‐concept of a dietitian delivering CBT‐AR as the primary clinician. In this initial case series of 11 adults living with ARFID, as hypothesized in aim 1, D‐CBT‐AR demonstrated high levels of initial uptake, treatment completion, and participant satisfaction, providing support for both feasibility and acceptability. As per aim 2, intent‐to‐treat analyses demonstrated proof of concept for D‐CBT‐AR. Hypothesis 2a, b, and c were supported with significant reductions in ARFID psychopathology, including participants meeting fewer DSM‐5‐TR criteria and increased food variety. Hypothesis 2d was not supported as BMI for weight‐suppressed participants increased with a large effect size, but was not statistically significant. Importantly, as hypothesized in 2e and f, we found no evidence that D‐CBT‐AR was associated with increases in eating disorder psychopathology or other types of co‐morbid psychopathology, which is reassuring given that dietitians are not mental health clinicians. Importantly, all findings were sustained at 6‐month follow‐up, with 45% meeting remission criteria as per intent‐to‐treat analysis at 6‐month follow‐up, as hypothesized in 2g. The findings from this study provide preliminary support that dietitians could have the capability to work successfully as the primary clinician in the treatment of ARFID when following a manualized cognitive‐behavioral therapy as also suggested for interprofessional practice in Bray et al. (2023). The results of the study also affirm the need for larger trials of D‐CBT‐AR to evaluate the efficacy and effectiveness of the approach and, ultimately, the possibility of leveraging dietitians to increase the workforce of clinicians who can alleviate the burden of ARFID in adults.

The feasibility of D‐CBT‐AR showcases the utility of dietitians in the treatment of ARFID, with 11 of 12 participants taking up the treatment and completers reporting high ratings of satisfaction, feasibility, acceptability, and appropriateness of a dietitian providing ARFID treatment. The feasibility and acceptability of dietitian delivery were notable given the chronicity (median duration of ARFID was over 17 years at treatment presentation) and psychiatric complexity (nearly two‐thirds screened positive for possible autism spectrum disorder on the AQ‐10) of the sample.

The reductions in ARFID psychopathology are supportive of CBT‐AR as a potentially efficacious treatment for adults living with ARFID. The PARDI outcomes showed pre‐ to post‐reduction in all three clinical phenotypes (i.e., sensory sensitivity, lack of interest in eating, and fear of aversive consequences) as well as overall ARFID severity. The effect size was consistent at r = 0.89 but more variable for the comparable Thomas et al. (2021) study which ranged from d = 0.37 to d = 1.95. Interestingly, the number of new foods added (n = 26) was slightly higher than in prior studies of psychologist‐delivered CBT‐AR (n = 16, Thomas et al. ²⁰²⁰; n = 18, Thomas et al. ²⁰²¹), highlighting a potential advantage of dietitian delivery that is worthy of investigation in future studies. Potential advantages include the use of the dietetic scope of practice with the use of medical nutrition therapy such as including specific recommendations to address nutrient deficiencies or excesses or prescribing oral nutrition supplements. In contrast, pre‐post changes on the PARDI‐AR‐Q showed medium effect sizes but were non‐significant for all but the lack of interest subscale. Differential findings for PARDI‐AR‐Q and PARDI interview may reflect low levels of participant insight at pre‐treatment; on average, participants rated their own symptoms as less severe on the PARDI‐AR‐Q than did the dietitian on the investigator‐based PARDI interview. Lastly, the nonsignificant change in BMI for weight‐suppressed participants could have been related to the variability of weight trajectories within the context of the small sample size.

Reductions to co‐morbid psychopathology occurred, and interestingly the DASS‐21 scores reduced, particularly stress and depression that had larger effect sizes. Compared to Thomas et al. (2021) the effect sizes for the CIA were of similar magnitude (d = 0.77 to our r = −0.69) and anxiety (d = 0.11 to our r = −0.06), but depression was stronger for our study at r = −0.91, compared to d = 0.47 of Thomas et al. (2021). Additionally, the effect‐size directions differed, but both measures reflected reductions in symptomatology. Of note, the median scores across the DASS stayed below levels indicating clinical concern (Shaw et al. 2017). While CBT‐AR does not directly target these measures of psychopathology, nor is it in a dietitian's clinical skillset to provide general psychotherapy, the observed changes in the scores could potentially reflect improvements in nutritional status (Feillet et al. 2019; García‐Montero et al. ²⁰²²) and/or reductions in ARFID psychopathology. Further research is needed to understand the reasons for the improvement of these secondary outcome measures. Of note, one participant who screened positive for autism spectrum disorder at baseline no longer screened positive at follow‐up after improvements with their ARFID symptoms. This finding raises the possibility that avoidant/restrictive eating may contribute, at least in part, to the high rates of overlap (Sader et al. 2025) observed between ARFID and ASD. Findings from one participant are not generalizable but warrant more research to explore the possibility.

The 6‐month follow‐up data presents a positive outlook for the longevity of CBT‐AR supporting reduced symptomology and remission from ARFID from the start of treatment. This aligns with other eating disorder treatments (de Jong et al. ²⁰²⁰; Schmidt et al. ²⁰¹²) and suggests participants continue to use skills learned in treatment and improve. Interestingly, at 6 months, of completers, only one participant had engaged in further therapy.

This trial is the first to use dietetic‐led CBT‐AR. The introduction of dietetic assessment, before starting D‐CBT‐AR, supported diagnostic accuracy, particularly for the identification of weight suppression, where a dietitian can diagnose malnutrition above and beyond standard BMI cut‐offs, and can assess if the participant is meeting specific macro and micro‐nutrient requirements. The core components of the CBT‐AR manual are not modified but augmented with dietetic education and skill. Dietitians provide a unique skillset of medical nutritional management, counseling, and education that is highly valued by lived experience (Yang et al. 2023) but under‐utilized in evidence‐based treatments (McMaster et al. ²⁰²¹). Further exploration is required to ascertain if certain phenotypes are more appropriate for a dietitian as the primary clinician, but participants of D‐CBT‐AR did not indicate negative responses to receiving treatment primarily from a dietitian. We provided D‐CBT‐AR with psychological and medical oversight for the safety and wellbeing of participants, recognizing that treatment for any eating disorder should not be done by a singular clinician (Heruc et al. 2020).

Limitations of this trial include the small sample size, which increases statistical heterogeneity. Further, there is bias from having a singular dietitian for treatment delivery and data collection. Further, we cannot ascertain whether any individuals declined to self‐refer to dietetic‐led treatment due to a preference for another type of practitioner. We recommend that further trials recruit larger samples with multiple clinicians to address these concerns. Further exploration is needed to identify if there are specific presentations of individuals that are not appropriate for dietitians to treat. Additional studies with larger cohorts will be able to provide further clarity. The FNS demonstrated marginal reliability, suggesting that the altered FNS warrants review to ensure it remains robust.

5. Conclusion

This study provided preliminary evidence that dietitians with advanced practice in eating disorders can be the primary clinician, within an interdisciplinary team, treating adults with ARFID. Feasibility, acceptability, and satisfaction were all rated highly from participant feedback. The pilot trial demonstrated proof‐of‐concept for reductions in ARFID symptoms and improvements in psychosocial outcomes, both of which were maintained at 6‐month follow‐up. The findings from this study provide preliminary support for larger trials testing the efficacy of a dietetic‐led CBT‐AR with more diverse ARFID populations.

Author Contributions

Copeland G. Winten: conceptualization, writing – original draft, writing – review and editing, investigation, funding acquisition, methodology, validation, software, formal analysis, project administration, data curation, resources, visualization. Kamryn T. Eddy: conceptualization, writing – review and editing, resources. Joanne Avery: writing – review and editing, methodology, formal analysis, data curation, investigation, software. Esben Strodl: writing – review and editing, project administration, supervision, resources, data curation, formal analysis, methodology. Jennifer J. Thomas: conceptualization, methodology, validation, writing – review and editing, supervision, resources, data curation. Lynda J. Ross: investigation, writing – review and editing, methodology, validation, formal analysis, supervision, data curation, resources. Gabriella Heruc: conceptualization, writing – review and editing, supervision, resources, validation.

Lived Experience Involvement Statement

Persons with lived experience, including members of the authorship team, were involved in the study design, execution, and the preparation of this manuscript.

Funding

C.G.W. acknowledges funding from the Rubenstein Charitable Foundation for a Summer Research Fellowship with the Massachusetts General Hospital Eating Disorders Clinical and Research Program. J.J.T. acknowledges funding from the National Institute of Mental Health (K24MH135189). C.G. Winten another grant is the Australian Government Research Training Program Scholarship.

Disclosure

Use of AI was not involved in any component of preparation of this manuscript.

Conflicts of Interest

J.J.T. and K.T.E. acknowledge royalties from the sale of their books on avoidant/restrictive food intake disorder from Cambridge University Press and consulting fees from Equip Health. The other authors declare no conflicts of interest.

Winten, C. G. , Strodl E., Heruc G., et al. 2026. “Dietetic‐Led Cognitive‐Behavioral Therapy for Avoidant/Restrictive Food Intake Disorder: Feasibility, Acceptability, and Proof‐of‐Concept.” International Journal of Eating Disorders 59, no. 9: 2013–2024. 10.1002/eat.70126.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.