A Systematic Review and Meta‐Analysis of Psychological Therapies for Avoidant/Restrictive Food Intake Disorder (ARFID) in Adolescents and Adults

ABSTRACT

Objective

The efficacy of psychological therapies for adolescents and adults with avoidant/restrictive food intake disorder (ARFID) has yet to be rigorously analyzed through systematic review or meta‐analysis.

Method

We identified articles from seven databases that presented psychological therapies for adolescents and adults with ARFID. First, our systematic review explored characteristics of psychological interventions, clinical team composition, changes in dietary intake, and methodological quality. Second, we conducted a meta‐analysis to quantify effect sizes for two primary outcomes (ARFID psychopathology and weight gain) and two secondary outcomes (anxiety and depression) from pretreatment to posttreatment.

Results

Forty articles included in the systematic review identified a variety of psychological interventions that led to increased dietary variety. Most studies were of high quality. Of these, 17 articles met the inclusion criteria for the meta‐analysis. Across studies, adolescents and adults with ARFID showed significant improvements from pretreatment to posttreatment including medium‐size reductions in ARFID psychopathology (g = 0.63, p < 0.001), large increases in weight (g = 1.26, p < 0.001), medium‐size reductions in anxiety (g = 0.42, p < 0.001), and small reductions in depression (g = 0.34, p < 0.05).

Discussion

This systematic review and meta‐analysis highlights promising preliminary outcomes of psychological therapies for adolescents and adults with ARFID across multiple domains. However, the current evidence base is small and primarily reliant on case studies and case series, with very few quasi‐experimental or randomized designs. As such, studies with larger sample sizes utilizing randomized control trial designs are needed to provide a more rigorous evaluation of psychological therapies for ARFID.

Summary

• This is the first meta‐analysis to provide evidence supporting the efficacy of psychological therapies for ARFID in people aged 10 years and above.

• Our systematic review identified 40 studies, primarily case studies and case series, most of which were of high quality. Our meta‐analysis of 17 studies found that psychological therapies for adolescents and adults with ARFID led to moderate and significant reductions in ARFID psychopathology, large and significant increases in weight, moderate and significant reductions in anxiety, and small but significant reductions in depression from pretreatment to posttreatment.

• Although our systematic review and meta‐analysis highlight preliminary evidence for the efficacy of psychotherapies for adolescents and adults with ARFID, further research is required to more rigorously confirm efficacy through randomized controlled trials.

1. Background

Avoidant/restrictive food intake disorder (ARFID) is a psychological disorder characterized by significant psychosocial, medical, and/or nutritional symptoms. These symptoms may be life‐threatening (Leach et al. 2024) and reduce quality of life (Ruiz Fischer and Starr 2024). ARFID‐specific treatments typically aim to target the diagnostic criteria, which include correcting low weight, resolving nutritional deficiencies, addressing dependence on nutritional supplements, and reducing psychosocial impairment. ARFID presents with three phenotypes: fear of aversive consequences of eating, lack of interest in eating food, or sensory sensitivity (Thomas et al. 2017). The three phenotypes can overlap but have individual characteristics leading to avoidance or restriction of food intake. Fear of aversive consequences can present as fear of choking, vomiting, or nausea. Lack of interest manifests as eating feeling like a chore, forgetting to eat, or poor hunger/fullness cues. Finally, sensory sensitivity involves concerns around taste, texture, smell, appearance, or tactile features of food.

Untreated ARFID can lead to severe health consequences. For example, undernutrition from long‐term nutrient deficiencies is a substantial concern (Schmidt et al. 2021), with the literature describing case studies of people with ARFID developing nerve damage from B12 deficiency (Chandran et al. 2015), blindness of various degrees from vitamin A deficiency (Chia et al. 2024), and scurvy from vitamin C deficiency (Benezech et al. 2020). Nitsch et al. (2023) reported that people with ARFID who present with severe malnutrition and low weight have similar medical risks to those with anorexia nervosa, including electrolyte abnormalities, bradycardia, and mood disturbances. As ARFID presents across the weight spectrum, not all individuals exhibit low body mass index (BMI) (James et al. 2024), meaning identification of malnutrition could be overlooked. The psychological impacts of ARFID can be reflective of symptoms of starvation syndrome with increased anxiety and depression (Spettigue et al. 2025). In turn, the impact of social isolation from ARFID leads to poor psychological outcomes (Norris et al. 2021) and the inability to connect with culture and/or resolve family conflict (Kerna et al. 2024).

Since ARFID's introduction to DSM‐5 (American Psychiatric Association 2013), there has been a noteworthy increase in clinical research to understand and treat ARFID. While there has been considerable research (Estrem et al. 2022, 2025; Sanchez‐Cerezo et al. 2023) on pediatric treatment resulting from the historical diagnosis of feeding disorder in infancy or early childhood (SAMHSA 2016), there has been less research on the treatment of ARFID in adolescents and adults (Willmott et al. 2024). This may be related to the relative recency of ARFID's introduction as a diagnosis in 2013 (APA 2013) in conjunction with the considerable time needed to conduct research to produce high‐quality evidence‐based treatments (Hariton and Locascio 2018). However, the lack of evidence to inform treatment recommendations is concerning, given the growing number of adolescents and adults requiring treatment (Soffritti et al. 2019).

Whereas pediatric feeding disorders have historically been treated in intensive medical settings with multidisciplinary involvement (Estrem et al. 2025), the leading treatments for eating disorders are outpatient psychological therapies such as family‐based treatment (Couturier et al. 2013) and cognitive‐behavioral therapy (CBT) (Öst et al. 2024). Current evidence‐based psychological therapies for eating disorders may not be appropriate for treating ARFID due to the difference in core etiological mechanisms; individuals with ARFID are not driven to restrict their intake due to weight and shape concerns (Bourne et al. 2020; Dalle Grave et al. 2019). Validated pathology tools, such as the Eating Disorder Examination Questionnaire (EDE‐Q), show individuals with ARFID score below the clinically significant cutoffs for weight and shape concerns and dietary restraint (Becker et al. 2020). In addition, individuals with ARFID present with higher usage and reliance on enteral nutrition (Grunewald et al. 2023; Nitsch et al. 2023). All these characteristics distinguish ARFID from other eating disorders, such as anorexia nervosa (Kambanis et al. 2024). Therefore, traditional versions of cognitive behavior therapy for eating disorders (Fairburn 2008; Waller et al. 2007), the Maudsley model of anorexia nervosa treatment for adults (Schmidt et al. 2018), and specialist supportive clinical management (McIntosh et al. 2023) are not appropriate for treating ARFID.

There are limited reviews on ARFID interventions; those published mainly pertain to pediatric care (Cucinotta et al. 2023; Di Cara et al. 2023; James et al. 2024), and none have focused solely on adolescents or adults. Willmott et al. (2024) presented a scoping review of treatment in ARFID for all age groups, but the review precluded a meta‐analysis due to the limited availability of eligible studies at the time of publication. As such, the average effect sizes of psychological therapies for ARFID remain unknown. This is regrettable, given that the lack of evidence has resulted in the absence of international guidelines for outpatient treatment of ARFID (Hilbert et al. 2017).

The aim of this study, therefore, was to conduct a systematic review and meta‐analysis of studies evaluating the preliminary efficacy of psychological therapies for adolescents and adults with ARFID, using pre‐ and post‐data, to contribute to the evidence base and support clinicians to use evidence‐informed treatments.

We acknowledge the limitations of conducting a meta‐analysis using pre‐ and post‐data (Cuijpers et al. 2016), given that confounders (such as the passage of time or attention from the interventionist) cannot be disentangled from the intervention itself without a proper comparison group. Further, preliminary effect size findings from a pre‐ and post‐meta‐analysis can provide a starting point for sample size calculations for future randomized control trials that are urgently needed. Therefore, this systematic review and meta‐analysis will inform clinicians on the emerging evidence base for ARFID treatment and encourage the evolution of research in ARFID treatment for adolescents and adults.

First, in a systematic review, we aimed to summarize the characteristics of current treatments for adolescent and adult ARFID (e.g., treatment modalities, health professionals engaged in treatment, assessment of nutritional intake, and methodological quality). Second, via meta‐analysis, we aimed to determine the strength and direction of effect sizes on two primary outcomes (reductions in ARFID psychopathology, increases in weight or BMI) and two secondary outcomes (reductions in anxiety and depression).

2. Methods

This systematic review and meta‐analysis followed the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) (Page et al. 2021) to guide research collection. Our PRISMA diagram is displayed in Figure 1. We preregistered this study in the International Prospective Register of Systematic Reviews (PROSPERO CRD42022339231).

FIGURE 1.

PRISMA 2020 flow diagram for 40 studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder. Note: ARFID = Avoidant/restrictive food intake disorder, WHO = World Health Organisation, PEG = Percutaneous Endoscopic Gastrostomy.

2.1. Data Collation, Extraction, and Analysis

We conducted a two‐pronged analysis: (1) a systematic review of all included studies and (2) a meta‐analysis of the subset of these studies that contained sufficient data to assess the degree of change from pretreatment to posttreatment across at least one of four outcome measures. For the systematic review, we analyzed data using narrative and quantitative assessment. The narrative assessment included determining whether treatment was provided by a singular therapist or a multidisciplinary team, the modalities and interventions used in treatment, how nutrition changes were reported and changes in oral intake reported via food groups as per the Recovery from Eating Disorders for Life (REAL) food pyramid (Hart et al. 2018), and study quality as per JBI (formerly “Joanna Briggs Institute”) critical appraisal tools (Moola et al. 2020). We reviewed all articles for quantitative data to include in our meta‐analyses across four outcomes: ARFID psychopathology, weight gain, anxiety, and depression.

The primary outcome measures for the meta‐analysis were change in ARFID psychopathology and improvement in weight status through weight gain. The secondary outcome measures were changes in the severity of anxiety and depression. We used the software comprehensive meta‐analysis (CMA) (Biostat Inc., NJ, USA) to calculate effect size from pre‐ to post‐data, using a random effects model. For randomized controlled trials with waitlist or control arms (which comprised a very small minority of studies in the current meta‐analysis), we only included data from the intervention arm. The analysis considered the study design and sample size when calculating effect size, making appropriate adjustments using a pair‐wise t‐test to assess raw datasets. If pre‐ and post‐correlation were required and could not be calculated, we set correlation at a conservative 0.7 (Jané et al. 2024). We chose Hedge's g as our measure of effect size because it adjusts for small sample size and heterogeneity (Higgins et al. 2024). We interpreted Hedge's g as small (0.2), medium (0.5), or large (0.7 and above) (Cohen 2013) and used p value to determine significance, at p < 0.05. To assess heterogeneity of effect sizes for each of our four outcomes, we used the Cochran Q, I ², and ${\tau }^2$ statistics. We set the Q‐test at p < 0.10 to appropriately identify outliers in our sample size (Borenstein 2009), used I ² to identify the proportion of variability due to heterogeneity, and applied ${\tau }^2$ to estimate the variance of true effects (Borenstein 2009). To calculate subgroup effect sizes, we used Z‐scores or ANOVAs with significance determined by p < 0.05. We conducted intercoder agreement for manuscript inclusion with double coding of all articles between C.G.W., E.S., and L.J.R., effect sizes through double coding with C.G.W. and P.E.K., and quality and bias from the JBI critical appraisal checklists between C.G.W. and E.S. We evaluated intercoder agreement using Cohen's kappa (κ), which we interpreted as none (0–0.2), minimal (0.21–0.39), weak (0.40–0.59), moderate (0.60–0.79), strong (0.80–0.90), or almost perfect (0.90) (McHugh 2012). We analyzed intercoder agreement for effect sizes using intraclass correlation (ICC), which we interpreted as poor (< 0.5), moderate (0.5–0.75), good (0.75–0.9), or excellent (> 0.9) (Koo and Li 2016).

2.2. Search Curation

We searched six databases to identify peer‐reviewed studies reporting on treatment outcomes for adolescents and/or adults with ARFID: PsycINFO, PubMed, CINAHL, Embase, Medline, and Web of Science; and searched a seventh database (ProQuest dissertations and thesis database) to identify unpublished gray literature. We conducted the initial search in October 2022 and conducted a follow‐up search in November 2025. As ARFID was introduced as a diagnosis in 2013, the timeframe for studies was April 2013 until November 2025.

We used the following search terms in each database, with variations for specific database requirements: ((“Avoidant Restrictive Food Intake Disorder” OR “ARFID”) OR (“picky eat*” OR “fussy eat*” OR “selective eat*” OR “food neophobia”)) AND (“intervention” OR “therapy” OR “treatment”).

Four authors (C.G.W., L.J.R., E.S., and J.J.T.) agreed on the search strategy, with C.G.W. executing the strategy and placing the resulting findings into Rayaan (Rayyan, MA, USA), an online systematic review software to create a database of records to be screened. C.G.W. was removed and manually checked for duplicates. Three authors (C.G.W., E.S., and L.J.R.) screened titles and abstracts using Rayyan's blinded autonomous feature to reach a consensus. Finally, we manually reviewed articles for full‐text eligibility and sought agreement on finalized papers with intercoder agreement of moderate strength at κ = 0.61. Any initial disagreements on study inclusion were resolved through consensus discussion. The finalized papers were reviewed to identify if they met the criteria to be included in the meta‐analysis by two authors, C.G.W. and E.S. To be included in the meta‐analysis, the paper needed to have a sample size of at least three participants and to present quantitative pre‐ and post‐intervention data on one of the four outcomes: ARFID psychopathology, weight gain, anxiety, or depression. We produced a codebook to extract data and recorded it onto an Excel spreadsheet, following a comprehensive code that was reproducible by all authors, which collated the information for this meta‐analysis and systematic review. Authors aligned with the codebook by reviewing how to code each item and holding interrater reliability meetings.

2.3. Inclusion Criteria

We included both published and unpublished studies written in or translated into English. To be included in our systematic review, the study needed to include participants who were aged 10 years and above as per the WHO definition of adolescents and adults (World Health Organization 2025), and who met criteria for an ARFID diagnosis per DSM‐5 (American Psychiatric Association 2013) or DSM‐5‐TR (American Psychiatric Association 2022) criteria. Studies also needed to focus on a psychological treatment involving at least some outpatient care, including day programs and hospital programs with outpatient components. We included studies where at least part of the sample was aged 10 years and above. When a study included some participants younger than 10 years old, we requested that the study authors share data on the sample subset who were 10 years and older for inclusion in the meta‐analysis. We omitted studies that involved an ineligible study design or publication type (systematic review, opinion piece, editorials, commentaries, conferences, workshops, or qualitative reviews); cell or animal studies, or acute interventions. We omitted medical or dietetic management only studies (including the use of medications or tube feeding), as we considered this outside of the psychological therapeutic scope. We also omitted studies where ARFID was not the primary diagnosis or the sample was not explicitly described as having ARFID. We did not exclude articles based on demographic data other than age.

2.4. Demographics and ARFID Diagnosis

We collated the demographic characteristics of the sample for each study, where possible. This included the number of participants meeting inclusion criteria, age, gender, socioeconomic status, co‐diagnoses, ethnicity/race, and country of authorship. We collated the measures used to ascertain the four meta‐analysis outcomes to identify if they are viable for the meta‐analysis. Further, we reviewed each paper for details on the ARFID diagnostic process, including the tool used to diagnose, qualitative measures of ARFID symptomology, and phenotypic presentation.

2.5. Intervention Characteristics

We reviewed all studies on the intervention method and the therapeutic modality. Two authors (C.G.W. and E.S.) coded and placed interventions into healthcare settings, modality, length of treatment, and treatment intensity.

2.6. Clinical Team Composition

We reviewed all studies to identify the health professionals reported as having delivered the treatment. We excluded clinicians who were only part of the research team. Author (C.G.W.) reviewed all papers and identified health professionals, with confirmation by authors (L.J.R. and E.S.). For ease of identification across countries, clinicians with mental health psychotherapy training were grouped as mental health clinicians.

2.7. Dietary and Food Intake

To assess nutritional outcomes and changes in dietary intake, we collated dietary intake data and reviewed measures such as changes in caloric intake, food preferences, percentage of meals completed, bites completed, number of foods added, or nutrient intake. The first author (C.G.W.) assessed each study based on her Accredited Practicing Dietitian qualifications, and the second author (L.J.R.) reviewed the assessment. After identifying the papers and determining how dietary intake was reported (C.G.W.), the researcher analyzed the details reported in each paper to identify foods linked to core food groups using predefined terminology outlined in the REAL food pyramid (Hart et al. 2018). According to the REAL food pyramid, these groups included carbohydrate foods, calcium foods, vegetables, fruits, protein foods, nuts/oils/fats, fun/social foods, and filler foods. C.G.W. analyzed each paper and assigned the foods described in each study to a food group.

2.8. Methodological Quality

We used the JBI (formerly “Joanna Briggs Institute”) critical appraisal checklists to assess the overall methodological quality of the identified papers with respect to trustworthiness, relevance, results, and bias. JBI has created tailored critical appraisal tools for the relevant research conducted, for case reports (Moola et al. 2020), case series (Munn et al. 2020), cohort studies (Moola et al. 2020), quasi‐experimental studies (Barker et al. 2024), and randomized controlled trials (Barker et al. 2023). Each tool has 8–13 domains that assess various areas of methodological bias, including selection, performance, detection, and reporting bias (Moola et al. 2020). Each domain is rated as “yes,” “no,” or “unclear.” As per the recommendation of Barker et al. (2024) and Munn et al. (2020), we reviewed the domains to evaluate their importance in relation to the aims of our paper; then, we decided to weight each domain equally and presented an overall rating for methodological quality based on the percentage of criteria met as “yes.” To operationalize this assessment across all tools, we defined quality as “high” when papers scored “yes” in > 70% of domains. We defined quality as “moderate” when papers scored “yes” in ≤ 70% of domains. Authors (C.G.W. and E.S.) reviewed each paper independently using the appropriate appraisal tool for the article. Cohen's Kappa was moderate (κ = 0.75), and all initial discrepancies were resolved by consensus agreement.

2.9. Measures of ARFID Psychopathology

We calculated effect sizes for ARFID psychopathology from the measures reported in the studies including the Pica, ARFID, and Rumination Disorder Interview‐ARFID‐Questionnaire (PARDI‐AR‐Q) (Bryant‐Waugh et al. 2022); Nine Item ARFID Screen (NIAS) (Burton Murray et al. 2021); Food Neophobia Scale (FNS) (Pliner and Hobden 1992); Children's Eating Behavior Questionnaire (CEBQ) (Wardle et al. 2001); Fear of Food Questionnaire (FFQ) (Zickgraf et al. 2022); and Food Acceptance/Fears Survey (FAFS) (Lane‐Loney et al. 2022). If a study used multiple measures of ARFID psychopathology, we used CMA software to combine effect sizes to create a single effect size per study. We prioritized measures completed by participants (rather than parents). Four studies had parent‐only completed measures (Billman et al. 2022; Breiner et al. 2024; Dumont et al. 2019; Lock, Sadeh‐Sharvit, et al. 2019; Shimshoni and Lebowitz 2020). Scores from Dumont et al. (2019) and Lane‐Loney et al. (2022) studies used dichotomous measures of food acceptance or rejection. To align with other studies in the meta‐analysis reporting increased food intake, we analyzed only food acceptance scores. Authors (C.G.W. and P.E.K.) achieved excellent (ICC = 1.0) intercoder agreement for effect sizes of ARFID psychopathology. Factors contributing to the excellent intercoder agreement included the authors meeting to agree on which measures to extract in advance and the frequent presentation of data as effect sizes in the studies themselves, which did not require further mathematical calculation.

2.10. Measures of Weight Gain

We collated weight gain effect size data for the meta‐analysis. Studies provided weight measures of varying types, including BMI, weight (kilograms and pounds), maximum body mass percentage, percentage of expected body weight, z‐score of weight and height, and percentage goal weight. We requested raw data from the authors of nine studies (Billman et al. 2022; Burton Murray et al. 2023; Knatz Peck et al. 2021; Lane‐Loney et al. 2022; Lien et al. 2025; Lock, Sadeh‐Sharvit, et al. 2019; Norris et al. 2021; Shimshoni and Lebowitz 2020; Volkert et al. 2021) to obtain the effect size of weight changes. Since adolescent participants may still have been growing in height during the course of treatment, we prioritized change in absolute weight as the measure of choice over change in BMI. This ensured that in instances where the patients gained weight and height, resulting in no change in BMI from pretreatment to posttreatment, weight change still reflected treatment progress. We used BMI in two studies where participants were adults and presumed to have finished growing (Burton Murray et al. 2023; MacDonald et al. 2024). We used a single measure for each individual study. Authors (P.E.K. and C.G.W.) independently calculated effect sizes for weight change, and intercoder agreement was excellent (ICC = 0.94). For the post hoc moderator analysis, we allocated studies (or sample subsets within studies where data were presented separately) into non‐weight gain or requiring weight gain, as defined in each individual study. We determined this through reviewing each study's inclusion criteria, goals, and raw data (if provided).

2.11. Measures of Depression and Anxiety

We produced effect sizes for depression and anxiety symptoms separately, from validated measures. For anxiety, the measures included the Depression Anxiety Stress Scale (DASS‐21) (Lovibond and Lovibond 1995), State Trait Anxiety Inventory (STAI) (Sydeman 2018), Beck Anxiety Inventory (BAI) (Beck et al. 1988), Revised Children's Manifest Anxiety Scales (RCMAS) (Reynolds 2014), Health Anxiety Inventory (HAI) (Salkovskis et al. 2002), Multidimensional Anxiety Scale for Children (MASC) (March et al. 1997), and the anxiety subscale from the Hospital Anxiety and Depression Scale (HADS) (Stern 2014). For depression, the measures included the DASS‐21 (Lovibond and Lovibond 1995), Children's Depression Inventory (CDI) (Kovacs 1981), Beck Depression Inventory (BDI) (Beck 1961), and the depression subscale from the HADS.

No studies reported on multiple measures for depression or anxiety, although some reported multiple subscales within these measures, such as the DASS‐21. In those instances, we used the subscales when available, or the entire scale when subscale scores were not reported. Initially, authors (C.G.W. and P.E.K.) achieved moderate intercoder agreement for depression (ICC = 0.56) and poor intercoder agreement for anxiety (ICC = 0.16). However, in both cases, the authors met to agree on each effect size through author consensus. In these meetings, they realized that the initially low ICCs were related to human input error (e.g., one coder entering the incorrect sample size or effect direction into the CMA software). Once these errors were corrected, agreement was excellent (ICC = 1.0) for both depression and anxiety.

2.12. Lived Experience

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

3. Results

3.1. Study Selection

The screening and selection results are outlined in the PRISMA diagram (Figure 1). The dataset comprised 40 articles, including 15 case reports, 22 case series, one quasi‐experimental study, and two randomized controlled trials. Of the 40 studies, we contacted 12 authors to request additional data for inclusion in the meta‐analysis; 9 of these consented and provided additional data, and 3 were excluded. The investigators provided the data in raw, de‐identified form, or as statistical output, such as effect sizes. Of the 40 studies, 17 had data available and met the inclusion criteria for the meta‐analysis. For the meta‐analysis, two articles (Lane‐Loney et al. 2022; Ornstein et al. 2017) used the same dataset and are encompassed under the Lane‐Loney study for the calculation of effect sizes.

3.2. Systematic Review

3.2.1. Demographic and Clinical Characteristics

A total of 963 participants were included in the 40 studies analyzed for the systematic review (Table 1), with 932 participants from 17 studies included in the meta‐analysis. Research was conducted primarily (65%) in the United States, with the remaining studies conducted in Australia, Canada, the United Kingdom, Türkiye, Italy, the Netherlands, and Denmark. The mean age of participants across studies was 17.4 years (SD = 13.4), and 56% identified as female. From the 18 studies that reported race or ethnicity, the predominant classification was Caucasian or white (74%). The majority of studies (93%) were disseminated between 2018 and 2025. A total of 31 papers provided or reported on comorbid medical and/or psychiatric diagnoses.

TABLE 1.

Characteristics of 40 studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

First author surname, year	Demographic characteristics						Treatment outcome measures
	N	Age (SD)	Gender	Ethnicity/race	Socioeconomic status	Country	ARFID psychopathology	Weight	Depression	Anxiety
Aloi et al. (2018)	1	24	100% M	NR	NR	Italy	NR	wt, BMI	NR	NR
Bergonzini et al. (2022)	1	15	100% M	NR	NR	Italy	NIAS	BMI	NR	NR
Billman et al. (2022)*	18 a	13 (2)	33% M	77% C	NR	USA	NIAS, Child Eating Behavior Questionnaire	% MBMI	NR	NR
			67% F	6% HL
				6% A
				11% MR
Blalock et al. (2020)	3 a	31 (4)	33% F	100% IAN	NR	USA	NR	NR	NR	NR
			67% NB
Blalock et al. (2025)	7 a	24 (6)	43% F	86% C	NR	USA	NR	NR	PHQ‐9	NR
			43% M	14% NR
			14% NB/T
Breiner et al. (2024)*	11 a	11 (1)	36% F	100% C	Household income b	USA	PARDI‐AR‐Q	NR	NR	NR
Brown and Hildebrandt (2020)	1	12	100% M	100% C	NR	USA	Always Sometimes Never	wt	BDI‐2	NR
Bryant‐Waugh (2013)	1	13	100% M	NR	NR	UK	NR	wt, percentiles	NR	NR
Burton Murray et al. (2023)*	14 a	40 (18)	71% F	100% C	NR	USA	NIAS, PARDI‐AR‐Q, Fear of Food Questionnaire, FNS	wt	Hospital Anxiety and Depression Scale
			29% M
Burton Murray et al. (2022)	1	16	100% M	NR	NR	USA	FNS, NIAS, PARDI‐AR‐Q	wt	NR	NR
Datta et al. (2023)	2 a	11 (1)	100% M	50% C	NR	USA	PARDI	wt	NR	NR
				50% HL
Dumont et al. (2019)*	11	14 (3)	63% M	NR	NR	Netherlands	Food selectivity test, FNS	Standard deviations of wt and height	NR	Visual analogue scale
			37% F
Fischer et al. (2015)	1	16	100% M	NR	NR	USA	Foods presented, bites consumed	wt	NR	NR
Görmez et al. (2018)	1	27	100% F	NR	NR	Türkiye	NR	wt, BMI	Hamilton Depression Scale	Hamilton Anxiety Rating Scale
Hellner et al. (2025)*	631 a	18 (8)	56% F	75% C	NR	USA	PARDI	wt restoration%	PHQ‐9	GAD‐7
			34% M	5% HL
			6% NB	4% A
			3% TM	3% AA
			1% TF	10% MR
				3% NR
Hooper et al. (2025)	4	27 (7)	75% F	NR	NR	USA	NIAS, Food Fear Questionnaire, PARDI	Meeting DSM‐5 ARFID A1 criteria	NR	NR
			25% M
Jensen (2020)	3	12 (1)	67% M	NR	NR	USA	Food preferences	NR	NR	NR
			33% F
King et al. (2022)	1	17	100% M	100% HL	NR	USA	Percentage food consumption	wt	NR	NR
Knatz Peck et al. (2021)*	6 a	19 (1)	83% F	75% C	NR	USA	NR	BMI	NR	STAI‐Trait
			17% M	25% A
Lane‐Loney et al. (2022)*	79 a	12 (2)	71% F	85% C	NR	USA	Food Acceptance/Fears Survey	BMI, % MBMI	CDI	RCMAS
			29% M	6% HL
				3% A
				1% AA
				5% other
Lien et al. (2025)*	27 a	13 (2)	52% F	100% C	NR	Denmark	NR	wt	NR	NR
Lock, Robinson, et al. (2019)	1 a	11	100% F	100% HL	NR	USA	PARDI	wt, % EBW	NR	NR
Lock, Sadeh‐Sharvit, et al. (2019)*	5 a	11 (1)	60% M	60% C	Household socioeconomic status b	USA	PARDI, Parent versus ARFID	wt, % EBW	BDI	Beck Anxiety Inventory
			40% F	40% MR
MacDonald et al. (2024)*	42 a	26 (7)	62% F	64% C	Financial support b	Canada	NR	BMI	NR	NR
			21% M	7% A
			17% NB	5% IAN
				17% AA
				5% MR
Norris et al. (2021)*	26	16 (2)	62% F	NR	NR	Canada	NR	BMI, z‐score BMI, % target goal weight	NR	NR
			38% M
Ornstein et al. (2017)* c	79 a	12 (2)	71% F	85% C	NR	USA	NR	BMI, % MBMI	NR	RCMAS
			29% M	6% HL
				3% A
				1% AA
				5% other
Price et al. (2024)	1	26	100% F	100% C	NR	UK	Clinical Impairment Assessment‐ARFID	BMI	PHQ‐9	GAD‐7
Proctor et al. (2024)	1	10	100% M	NR	NR	USA	NR	wt, BMI	NR	NR
Rienecke et al. (2020)	2 a	12 (3)	100% M	100% C	NR	USA	NR	wt, BMI	CDI	Multidimensional Anxiety Scale for Children
Ripple et al. (2022)	1	12	100% F	100% C	NR	USA	Percentage of food consumed	wt, wt percentile	NR	NR
Shimshoni and Lebowitz (2020)*	6 a	13 (1)	100% M	NR	Annual household income b	USA	NIAS, food flexibility	wt, % EBW	NR	Anxiety Disorders Interview Schedule Children and Parent versions
Spettigue et al. (2018)	3 a	13 (1)	100% F	NR	NR	USA	NR	wt	NR
Steen and Wade (2018)	1	42	100% M	NR	NR	Australia	NR	BMI	Depression Anxiety Stress Scale‐21
Taylor et al. (2019)	1	13	100% F	NR	“High”	Australia	Percentage consumption	NR	NR	NR
Taylor (2021)	1	11	100% M	NR	NR	Australia	Percentage consumption	NR	NR	NR
Thomas et al. (2020)*	20	13 (2)	55% M	90% C	NR	USA	PARDI, number of foods incorporated, FNS	wt, BMI	CDI‐2	STAI‐C
			45% F	5% A
				5% MR
Thomas et al. (2021)*	15	25 (10)	67% F	100% C	NR	USA	PARDI, number of foods incorporated, FNS	wt, BMI	BDI‐2	STAI
			33% M
Volkert et al. (2021)*	9 a	12 (2)	100% M	100% C	NR	USA	Mealtime observations	% BMI, z‐score BMI, weight status	NR	NR
Winten et al. (2025)	2	21 (3)	50% F	50% C	NR	Australia	PARDI, PARDI‐AR‐Q	wt	DASS‐21	DASS‐21
			50% M	50% A
Yasar et al. (2019)	2	19 (1)	100% F	100% C	NR	Türkiye	NR	NR	NR	NR

Note: standard deviation (SD) is only available in studies larger than n = 1. *article included in meta‐analysis.

Abbreviations: %EBW, percentage estimated mean body weight; %MBMI, percentage median BMI; A, Asian; AA, African or African American; BDI, Beck Depression Inventory; BMI, body mass index; C, Caucasian or white; CDI, Children's Depression Inventory; F, female; FNS, Food Neophobia Scale; GAD‐7, General Anxiety Disorder‐7; HL, Hispanic or Latino; IAN, Indian or Alaskan Native; M, male; MR, multiracial; NB, nonbinary; NIAS, Nine Item ARFID Screen; NR, not reported; PARDI, Pica; ARFID and Rumination Disorder Interview; PARDI‐AR‐Q, PARDI‐ARFID Questionnaire; PHQ‐9, Patient Health Questionnaire‐9; RCMAS, Revised Children's Manifest Anxiety Scale; T/TF/TM, trans male or female; wt, weight.

^a Partial sample due to including participants aged ≥ 10 from a larger sample with a broader age range; and/or including participants with ARFID diagnosis only from a study including multiple eating disorder diagnoses.

^b Data unavailable for sample reported.

^c Uses the same dataset as Lane‐Loney et al. (2022).

A total of 37 studies included samples that were exclusively ARFID, with 3 studies (Blalock et al. 2020, 2025; Knatz Peck et al. 2021) diagnosing ARFID within a larger eating disorders sample. A total of 26 of the 40 studies (n = 237) reported on ARFID phenotypes; pooling across those studies, 30% of participants presented with fear of aversive consequences; 16% with lack of interest and sensory sensitivity; 15% with lack of interest alone; 12% as sensory sensitivity alone; 12% undefined multiple phenotypes; 5% lack of interest and fear of aversive consequences; 5% fear of aversive consequences and sensory sensitivity; and 5% all phenotypes. ARFID remission was inconsistently described, and multiple measures were referred to as markers of remission, such as no longer meeting DSM‐5 criteria; decreased scores on the NIAS, PARDI‐AR‐Q, or PARDI; reduction in ARFID symptoms; or weight gain. A total of 20 published and unpublished papers (Bergonzini et al. 2022; Billman et al. 2022; Breiner et al. 2024; Brown and Hildebrandt 2020; Bryant‐Waugh 2013; Burton Murray et al. 2022, 2023; Datta et al. 2023; Dumont et al. 2019; Hooper et al. 2025; Knatz Peck et al. 2021; Lane‐Loney et al. 2022; Lock, Robinson, et al. 2019; Proctor et al. 2024; Shimshoni and Lebowitz 2020; Steen and Wade 2018; Thomas et al. 2020, 2021; Winten et al. 2025; Yasar et al. 2019) reported on meeting versus not meeting ARFID diagnostic criteria at posttreatment, or perceived remission status.

3.2.2. Intervention Characteristics

The psychological therapies evaluated were heterogeneous across studies. Intervention modalities varied and were frequently therapies adapted for ARFID. Treatments included evidence‐based modalities such as CBT (Aloi et al. 2018; Billman et al. 2022; Bryant‐Waugh 2013; Görmez et al. 2018; Jensen 2020; Ornstein et al. 2017; Steen and Wade 2018), CBT with behavioral therapy (Dumont et al. 2019; Fischer et al. 2015), and CBT in conjunction with Eye Movement Desensitization (Yasar et al. 2019). For ARFID‐specific treatments, nine studies used CBT for ARFID (CBT‐AR) (Burton Murray et al. 2022, 2023; Hellner et al. 2025; Hooper et al. 2025; MacDonald et al. 2024; Price et al. 2024; Thomas et al. 2020, 2021; Winten et al. 2025), and five papers reported on family‐based therapy for ARFID (FBT‐ARFID) (Hellner et al. 2025; Lien et al. 2025; Lock, Robinson, et al. 2019; Lock, Sadeh‐Sharvit, et al. 2019; Spettigue et al. 2025). Behavioral interventions were reported in various formats from multidisciplinary teams with applied behavioral analytic or described as behavioral therapy (Blalock et al. 2020, 2025; King et al. 2022; Norris et al. 2021; Proctor et al. 2024; Rienecke et al. 2020; Ripple et al. 2022; Taylor 2021; Taylor et al. 2019; Volkert et al. 2021). Other studies specifically developed interventions for ARFID, including Supportive Parenting for Anxious Childhood Emotions (SPACE)‐ARFID (Shimshoni and Lebowitz 2020), Parent Training Protocol (PTP)‐ARFID (Breiner et al. 2024), and psychoeducation and motivation treatment (Datta et al. 2023). More generalized eating disorder treatments that included ARFID were young adults temperament‐based treatment with supports (YA‐TBTS) (Knatz Peck et al. 2021). Hellner et al. (2025) reported on two separate treatments administered to participants: CBT‐AR and FBT‐ARFID. A total of 34 interventions were delivered exclusively in an outpatient setting; the other 6 were delivered as a mix across outpatient and inpatient settings. In the six studies (Bergonzini et al. 2022; Görmez et al. 2018; MacDonald et al. 2024; Norris et al. 2021; Rienecke et al. 2020; Ripple et al. 2022) that utilized a mixed outpatient‐inpatient setting, the therapeutic strategies typically changed across settings; for example, supportive meal therapy was more likely to be used in the inpatient setting (vs. outpatient). The outpatient interventions also involved a mix of face‐to‐face and telehealth delivery modes. Common intervention strategies drawn from article descriptions included psychoeducation, family sessions, cognitive interventions, self‐monitoring, and exposure therapy. The descriptors of strategies and interventions exhibited considerable variation. There was diversity in the intensity and length of treatments, ranging from two sessions within 2 weeks (Breiner et al. 2024) to 100 sessions provided across an unknown timeframe (Ripple et al. 2022). Two studies (Bryant‐Waugh 2013; Norris et al. 2021) did not outline the number of sessions or the length of time.

3.2.3. Clinical Team Composition

Interventions were provided by a mental health clinician in all 40 studies, with only one study also employing a peer support worker alongside the mental health clinician (Hellner et al. 2025). A total of 21 studies reported involving a multidisciplinary team, as outlined in Table 2. Inpatient programs were more likely to involve a multidisciplinary team (e.g., mental health clinicians, dietitians, psychiatrists, and physicians).

TABLE 2.

Health professionals engaged in providing treatment for 40 studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder. ^a

First author surname, year	Mental health clinician	Psychiatrist	Dietitian	Nurse	Physician	Speech pathologist	Peer support
Aloi et al. (2018)	×	×
Bergonzini et al. (2022)	×	×	×		×
Billman et al. (2022)	×	×	×	×	×
Blalock et al. (2020)	×	×	×		×
Blalock et al. (2025)	×		×
Breiner et al. (2024)	×
Brown and Hildebrandt (2020)	×				×
Bryant‐Waugh (2013)	×
Burton Murray et al. (2022)	×
Burton Murray et al. (2023)	×
Datta et al. (2023)	×
Dumont et al. (2019)	×
Fischer et al. (2015)	×
Görmez et al. (2018)	×		×	×
Hellner et al. (2025)	×		×		×		×
Hooper et al. (2025)	×		×	×	×
Jensen (2020)	×
King et al. (2022)	×
Knatz Peck et al. (2021)	×	×
Lane‐Loney et al. (2022)	×	×	×	×	×
Lien et al. (2025)	×
Lock, Robinson, et al. (2019)	×
Lock, Sadeh‐Sharvit, et al. (2019)	×
MacDonald et al. (2024)	×	×	×	×
Norris et al. (2021)	×	×	×	×	×
Ornstein et al. (2017)	×	×	×		×
Price et al. (2024)	×
Proctor et al. (2024)	×		×	×		×
Rienecke et al. (2020)	×
Ripple et al. (2022)	×		×		×	×
Shimshoni and Lebowitz (2020)	×
Spettigue et al. (2018)	×
Steen and Wade (2018)	×				×
Taylor et al. (2019)	×
Taylor (2021)	×
Thomas et al. (2020)	×				×
Thomas et al. (2021)	×				×
Volkert et al. (2021)	×		×	×	×	×
Winten et al. (2025)	×		×	×	×
Yasar et al. (2019)	×

Note: x = health professional was present in study. dietitian = dietitian or nutritionist, nurse = nurse or nurse practitioner, mental health clinician = occupational therapist, psychologist (of all discourses) or Board Certified Behavior Analyst or undefined therapist or social worker, physician = pediatrician, adolescent medicine physician, physician, neurologist, general practitioner or local medical provider.

^a Engagement in treatment is defined as a health professional providing care that aligns with the method described for treatment or specifically outlined within the paper as health professionals involved during treatment.

3.2.4. Dietary and Food Intake

Measures of food and dietary intake varied considerably. across the studies. A total of 16 studies did not report on dietary intake or foods consumed (Billman et al. 2022; Blalock et al. 2020, 2025; Breiner et al. 2024; Burton Murray et al. 2023; Hellner et al. 2025; Hooper et al. 2025; Knatz Peck et al. 2021; Lane‐Loney et al. 2022; Lien et al. 2025; Lock, Robinson, et al. 2019; Lock, Sadeh‐Sharvit, et al. 2019; MacDonald et al. 2024; Norris et al. 2021; Ornstein et al. 2017; Spettigue et al. 2018). A total of 24 papers reported on dietary intake in various methods, including the number of foods incorporated (Thomas et al. 2020, 2021; Winten et al. 2025), caloric intake (Bergonzini et al. 2022; King et al. 2022; Volkert et al. 2021), food sensory preferences (Yasar et al. 2019), flexibility of oral intake (Shimshoni and Lebowitz 2020), or percentage consumed (King et al. 2022; Ripple et al. 2022; Taylor 2021; Taylor et al. 2019). Definitions of food and diet varied across studies, being dependent on how intake was measured. A total 16 papers (Aloi et al. 2018; Bergonzini et al. 2022; Brown and Hildebrandt 2020; Bryant‐Waugh 2013; Burton Murray et al. 2022; Datta et al. 2023; Dumont et al. 2019; Fischer et al. 2015; Görmez et al. 2018; King et al. 2022; Price et al. 2024; Proctor et al. 2024; Rienecke et al. 2020; Ripple et al. 2022; Steen and Wade 2018; Taylor 2021; Taylor et al. 2019; Volkert et al. 2021) had sufficient descriptive data to be analyzed for pretreatment food group intake based on the REAL food pyramid (Hart et al. 2018) (Table S1). We found that, across studies, carbohydrates were the most consumed food group, while fruit and vegetables were the least. Additionally, protein and calcium intake varied across studies, with some reporting high and others negligible intake. Protein intake tended to be of low‐quality proteins, such as processed meats. Saturated fats, categorized as fun/social foods, accounted for the majority of fat intake. Intake of filler foods such as energy drinks or sugar‐free foods was not frequently described. Posttreatment oral intake was inconsistently reported; all studies described improvements in oral intake. Seven of these studies (Aloi et al. 2018; Burton Murray et al. 2022; Fischer et al. 2015; King et al. 2022; Taylor 2021; Taylor et al. 2019; Volkert et al. 2021) provided enough descriptive information to assess changes in nutritional intake posttreatment, and all showed an increase in volume of nutrition and intake of food categories, including protein, fat, fruits, vegetables, and carbohydrates.

A majority of studies were based on Western food systems. Methods of measuring nutrition in the studies included caloric intake, percentage of bites consumed, and the number of new foods introduced to participants' diets. One study reported on the extent to which the sample met caloric recommendations pretreatment (Volkert et al. 2021), and two reported on micronutrient requirements pretreatment and posttreatment (Volkert et al. 2021; Winten et al. 2025). Of the 24 studies that reported on dietary intake, 6 included a dietitian on the treatment team (Bergonzini et al. 2022; Görmez et al. 2018; Proctor et al. 2024; Ripple et al. 2022; Volkert et al. 2021; Winten et al. 2025).

All participants included in the review were eating food orally, and 22 studies reported on the consumption of oral nutrition supplements as part of the diagnostic criteria. Of these, eight studies (Datta et al. 2023; Dumont et al. 2019; Proctor et al. 2024; Spettigue et al. 2018; Thomas et al. 2020, 2021; Volkert et al. 2021; Winten et al. 2025) reported that a subset of participants were consuming oral nutrition supplements at pretreatment. None included participants who were exclusively tube‐fed at baseline (consistent with our systematic review's exclusion criteria), although historical information reported that a subset of participants had been previously tube‐fed (Dumont et al. 2019; Proctor et al. 2024; Volkert et al. 2021). Studies frequently reported that participants were malnourished, but only one study (Winten et al. 2025) specified how malnutrition was defined or diagnosed.

Several studies reported on micronutrient deficiencies and excesses at pretreatment. However, most relied on self‐report, such as those that used the PARDI or PARDI‐AR‐Q (Breiner et al. 2024; Burton Murray et al. 2022, 2023; Hellner et al. 2025; Lien et al. 2025; Lock, Robinson, et al. 2019; Lock, Sadeh‐Sharvit, et al. 2019; Thomas et al. 2020, 2021; Volkert et al. 2021), which includes items on self‐reported nutrition deficiencies. Only four studies (Aloi et al. 2018; Bryant‐Waugh 2013; Görmez et al. 2018; Winten et al. 2025) reported micronutrient deficiencies evaluated by blood tests. All studies reported an improvement in nutritional intake; however, the nature of the improvement was directed by the intervention and measures used to ascertain dietary change. For example, Taylor et al. (2019) reported on the percentage of bites consumed specific to food type to report volume and variety of dietary intake change.

3.2.5. Methodological Quality

We assessed methodological quality of studies for case reports (n = 15; Table 3), case series (n = 22; Table 4), quasi‐experimental studies (n = 1; Table 5), and randomized controlled trials (n = 2; Table 6). Most case reports were of high quality (n = 14), with one rated as moderate; most case series were high quality (n = 18), with five rated as moderate. Outcome measurement bias was identified as a limitation to the quality of case reports and case series studies, with further clarity needed on participant details. One study was identified as quasi‐experimental and of high quality. Of the two randomized control trials, one was rated high (Lock, Sadeh‐Sharvit, et al. 2019) and the other moderate (Breiner et al. 2024).

TABLE 3.

JBI ^a Critical Appraisal Tool for assessment of methodological quality and report clarity of 15 case reports included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

First author, year	Domains								Quality
	1	2	3	4	5	6	7	8
Aloi et al. (2018)	Yes	Yes	Yes	Unclear	Yes	Yes	Unclear	Yes	High
Bergonzini et al. (2022)	Yes	Yes	Yes	Yes	Unclear	Yes	Yes	Yes	High
Brown and Hildebrandt (2020)	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	Yes	High
Bryant‐Waugh (2013)	Yes	Yes	Yes	Yes	Yes	Unclear	Unclear	Yes	High
Burton Murray et al. (2022)	Yes	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	High
Fischer et al. (2015)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Görmez et al. (2018)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
King et al. (2022)	Yes	Yes	Yes	Unclear	Yes	Yes	Yes	Yes	High
Lock, Robinson, et al. (2019)	Yes	No	Yes	Yes	Yes	Unclear	No	Yes	Moderate
Price et al. (2024)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Proctor et al. (2024)	Yes	Yes	Yes	Unclear	Yes	Yes	Yes	Yes	High
Ripple et al. (2022)	Yes	Yes	Yes	Unclear	Yes	Yes	Yes	Yes	High
Steen and Wade (2018)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Taylor et al. (2019)	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	High
Taylor (2021)	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	High

Note: We defined quality as “high” when case reports scored “yes” in > 70% of domains. We defined quality as “moderate” when case reports scored “yes” in ≤ 70% of domains. The eight domains for case reports included the following: Domain 1. Were patient's demographic characteristics clearly described? Domain 2. Was the patient's history clearly described and presented as a timeline? Domain 3. Was the current clinical condition of the patient on presentation clearly described? Domain 4. Were diagnostic tests or assessment methods and the results clearly described? Domain 5. Was the intervention(s) or treatment procedure(s) clearly described? Domain 6. Was the post‐intervention clinical condition clearly described? Domain 7. Were adverse events (harms) or unanticipated events identified and described? Domain 8. Does the case report provide takeaway lessons?

^a Formerly Joanna Briggs Institute.

TABLE 4.

JBI ^a Critical Appraisal Tool for assessment of methodological quality and report clarity of 22 case series studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

First author, year	Domains										Quality
	1	2	3	4	5	6	7	8	9	10
Billman et al. (2022)	Yes	Yes	Yes	Yes	Yes	Unclear	Unclear	Yes	Yes	Yes	High
Blalock et al. (2020)	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	No	Yes	No	Moderate
Burton Murray et al. (2023)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Datta et al. (2023)	Unclear	Yes	Yes	No	No	Yes	Yes	Yes	Yes	Unclear	Moderate
Dumont et al. (2019)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Hellner et al. (2025)	Yes	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Hooper et al. (2025)	Yes	Yes	Yes	Unclear	Yes	No	Yes	Yes	Yes	Yes	High
Jensen (2020)	Unclear	Yes	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Knatz Peck et al. (2021)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Lane‐Loney et al. (2022)	Yes	Yes	Unclear	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	High
Lien et al. (2025)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
MacDonald et al. (2024)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Norris et al. (2021)	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Ornstein et al. (2017)	Yes	Yes	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Rienecke et al. (2020)	Yes	No	No	Unclear	No	Yes	Yes	Unclear	Yes	Yes	Moderate
Shimshoni and Lebowitz (2020)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Spettigue et al. (2018)	Yes	Yes	Unclear	No	Yes	Yes	Yes	Yes	Yes	Yes	High
Thomas et al. (2020)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Thomas et al. (2021)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High
Volkert et al. (2021)	NA	NA	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Moderate
Winten et al. (2025)	Yes	Yes	Yes	Unclear	No	Yes	Yes	Yes	Yes	Yes	High
Yasar et al. (2019)	Unclear	Yes	No	Unclear	Unclear	Yes	Yes	Unclear	Yes	Yes	Moderate

Note: We defined quality as “high” when case series scored “yes” in > 70% of domains. We defined quality as “moderate” when case series scored “yes” in ≤ 70% of domains. The 10 domains for case series included the following: Domain 1. Were there clear criteria for inclusion in the case series? Domain 2. Was the condition measured in a standard, reliable way for all participants included in the case series? Domain 3. Were valid methods used for identification of the condition for all participants included in the case series? Domain 4. Did the case series have consecutive inclusion of participants? Domain 5. Did the case series have complete inclusion of participants? Domain 6. Was there clear reporting of the demographics of the participants in the study? Domain 7. Was there clear reporting of clinical information of the participants? Domain 8. Were the outcomes or follow‐up results of cases clearly reported? Domain 9. Was there clear reporting of the presenting site(s)/clinic(s) demographic information? Domain 10. Was statistical analysis appropriate?

Abbreviation: NA, not applicable.

^a Formerly Joanna Briggs Institute.

TABLE 5.

JBI ^a Critical Appraisal Tool for assessment of methodological quality and report clarity of quasi‐experimental study included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

First author, year	Domains												Quality
	1	2	3	4	5	6	7	8	9	10	11	12
Blalock et al. (2025)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High

Note: We defined quality as “high” when quasi‐experimental studies scored “yes” in > 70% of domains. We defined quality as “moderate” when quasi‐experimental scored “yes” in ≤ 70% of domains. The 12 domains for quasi‐experimental studies included the following: Domain 1. Is it clear in the study what is the “cause” and what is the “effect” (i.e., there is no confusion about which variable comes first)? Domain 2. Was there a control group? Domain 3. Were participants included in any comparisons similar? Domain 4. Were the participants included in any comparisons receiving similar treatment/care, other than the exposure or intervention of interest? Domain 5. Were there multiple measurements of the outcome, both pre and post the intervention/exposure? Domain 6. Were the outcomes of participants included in any comparisons measured in the same way? Domain 7. Were outcomes measured in a reliable way? Domain 8. Was follow‐up complete and if not, were differences between groups in terms of their follow‐up adequately described and analyzed? Domain 9. Was appropriate statistical analysis used?

^a Formerly Joanna Briggs Institute.

TABLE 6.

JBI ^a Critical Appraisal Tool for assessment of methodological quality and report clarity of two randomized controlled trial studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

First author, year	Domains													Quality
	1	2	3	4	5	6	7	8	9	10	11	12	13
Breiner et al. (2024)	Yes	No	Unclear	No	Unclear	NA	No	Yes	Yes	Yes	Yes	Yes	Yes	Moderate
Lock, Sadeh‐Sharvit, et al. (2019)	Yes	Yes	Yes	Unclear	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	High

Note: We defined quality as “high” when randomized controlled trials scored “yes” in > 70% of domains. We defined quality as “moderate” when randomized controlled trials scored “yes” in ≤ 70% of domains. The 13 domains for randomized controlled trials included the following: Domain 1. Was true randomization used for assignment of participants to treatment groups? Domain 2. Was allocation to treatment groups concealed? Domain 3. Were treatment groups similar at the baseline? Domain 4. Were participants blind to treatment assignment? Domain 5. Were those delivering the treatment blind to treatment assignment? Domain 6. Were treatment groups treated identically other than the intervention of interest? Domain 7. Were outcome assessors blind to treatment assignment? Domain 8. Were outcomes measured in the same way for treatment groups? Domain 9. Were outcomes measured in a reliable way? Domain 10. Was follow‐up complete and if not, were differences between groups in terms of their follow‐up adequately described and analyzed? Domain 11. Were participants analyzed in the groups to which they were randomized? Domain 12. Was appropriate statistical analysis used? Domain 13. Was the trial design appropriate and any deviations from the standard RCT design (individual randomization, parallel groups) accounted for in the conduct and analysis of the trial?

Abbreviation: NA, not applicable.

^a Formerly Joanna Briggs Institute.

3.3. Meta‐Analysis

3.3.1. ARFID Psychopathology

A total of 11 studies evaluated the change in ARFID psychopathology from pretreatment to posttreatment. Figure 2 shows the overall effect size, which revealed a statistically significant medium‐sized reduction in ARFID psychopathology (Hedges' g = 0.63; 95% CI = [0.53, 0.73]; Z = 12.42, p < 0.001). Additionally, the Q statistic did not indicate significant heterogeneity across studies (Q (10) = 14.50, p = 0.152; I ² = 31%; τ ² = 0.008), suggesting only modest variability in effect sizes for ARFID psychopathology.

FIGURE 2.

Change in ARFID psychopathology from pretreatment to posttreatment for ARFID (k = 11).

3.3.2. Weight Gain

A total of 13 studies in the meta‐analysis included measures of weight gain from pretreatment to posttreatment (Figure 3). The overall effect size revealed a statistically significant, large increase in weight from pretreatment to posttreatment (Hedge's g = 1.26, 95% CI = [0.97, 1.56]; Z = 8.44, p < 0.001). Additionally, there was significant heterogeneity across studies (Q (12) = 25.61, p < 0.05; I ² = 53%; τ ² = 0.13), indicating substantial variability in effect sizes. The level of heterogeneity was influenced by one study—Lien et al. (2025) (g = 3.20). With this study removed, the Q statistic was no longer significant (Q (11) = 18.58, p > 0.05; I ² = 41%; τ ² = 0.075), as all other studies reported Hedge's g within a narrower range (from 0.40 to 1.91). We conducted a post hoc moderator analysis to identify any potential contributors to the observed heterogeneity by categorizing each study into a weight‐gain treatment versus a non‐weight‐gain treatment. The results identified that there was no significant difference in the magnitude of weight change between treatments focused on weight gain versus those not focused on weight gain (Z = 0.38, p = 0.706); in other words, patients showed weight increases from pretreatment to posttreatment whether weight gain was a treatment goal or not.

FIGURE 3.

Change in weight from pretreatment to posttreatment for ARFID (k = 13).

3.3.3. Anxiety

Seven studies measured change in anxiety from pretreatment to posttreatment (Figure 4). The overall effect size revealed a statistically significant, medium reduction in anxiety symptoms (Hedge's g = 0.42, 95% CI = [0.26, 0.57], Z = 5.35, p < 0.001), with no significant heterogeneity in effect sizes across studies (Q (6) = 7.47, p = 0.28; I ² = 20%; τ ² = 0.008).

FIGURE 4.

Change in anxiety from pretreatment to posttreatment for ARFID (k = 7).

3.3.4. Depression

Nine studies measured change in depression from pretreatment to posttreatment (Figure 5). The overall effect size revealed a statistically significant, small reduction in depression symptoms (Hedge's g = 0.34, 95% CI = [0.09, 0.59], Z = 2.66, p < 0.05). Of note, there was significant heterogeneity among the studies (Q (8) = 22.61, p < 0.01; I ² = 65%; τ ² = 0.08), indicating substantial variability in effect sizes. Two studies showed negative effect sizes, indicating an increase in depressive symptoms from pretreatment to posttreatment (Knatz Peck et al. 2021; Shimshoni and Lebowitz 2020). Although we identified significant heterogeneity in effect sizes for depression, we did not conduct a post hoc moderator analysis given the small sample size (k = 9).

FIGURE 5.

Change in depression from pretreatment to posttreatment from ARFID (k = 9).

4. Discussion

This study is the first systematic review and meta‐analysis to investigate the evidence base for psychological therapies for adolescents and adults with ARFID and provides preliminary evidence of efficacious treatments that warrant continued investigation through rigorous randomized controlled trials. The majority of studies were case reports or case series, with all data reviewed as pretreatment and posttreatment outcomes. Our systematic review revealed that most studies evaluated cognitive‐behavioral or family‐based therapies, or a mix of multiple therapies. There were common activities and strategies utilized across interventions, independent of modality or intensity, and the majority of providers were mental health clinicians. Our meta‐analysis of 17 studies highlighted that psychological therapies, on average, led to significant improvements in ARFID psychopathology (medium effect), weight gain (large effect), anxiety (medium effect), and depression (small effect).

Our findings highlight the substantial promise of psychological therapies for adolescents and adults with ARFID for improving both primary ARFID symptoms (i.e., ARFID psychopathology and weight gain) and comorbid psychopathology (i.e., anxiety and depression). As the majority (n = 37) of the treatments evaluated were designed or adapted specifically for ARFID, it is encouraging that they exerted significant effects on the primary outcomes of ARFID psychopathology and weight gain. Weight gain was a more commonly measured outcome and had a larger effect size than ARFID psychopathology. ARFID psychopathology exhibits inherent heterogeneity due to variability in severity and phenotypic presentations. Our meta‐analytic heterogeneity analysis captured heterogeneity only among effect sizes between studies, rather than within the studies. Therefore, the results are reflective of study‐level differences and do not account for within‐study variability. Weight gain was the most commonly reported measure (k = 13) and may have been an easier measure to capture than ARFID psychopathology because ARFID is in its infancy compared to other eating disorders, and some studies were disseminated before the validation of newer measures designed to meticulously quantify ARFID psychopathology (e.g., PARDI and NIAS). Notably, our moderator analysis identified that weight gain occurred even for participants who were not targeted for weight gain. A possible explanation for this finding is that adolescents gained weight because the treatment increased variety, thereby improving growth. Similarly, across the age spectrum, limited dietary variety is indicative of malnutrition, and additional oral intake may have been required to meet micronutrient needs. Beyond physical health improvements, the treatments also present secondary benefits of overall positive effects on anxiety and depression symptomology. Interestingly, the effect sizes for anxiety and depression in our meta‐analysis were larger than the effect sizes from a previous meta‐analysis for bulimia nervosa and binge eating disorder (Monteleone et al. 2022). This needs further investigation to assess the level of interaction of ARFID with depression and anxiety compared to other eating disorders.

The results of the review highlight varied research methodologies from foundational case studies (e.g., Bryant‐Waugh 2013) to large naturalistic trials (e.g., Hellner et al. 2025) to randomized controlled trials (e.g., Breiner et al. 2024), reflecting the increasing sophistication of ARFID treatment research since its introduction to the diagnostic nomenclature in 2013. That said, the majority of the current evidence base for psychological therapies for ARFID relies on case studies and case series, with only two small randomized controlled trials. This means we cannot rule out that ARFID could benefit from alternative interventions, including the passage of time (i.e., no intervention). That said, in both the randomized controlled trials, the treatment condition outperformed usual care (Lock, Sadeh‐Sharvit, et al. 2019) or wait list (Breiner et al. 2024), and naturalistic nontreatment studies suggest that, without intervention, ARFID typically follows a chronic course (Breithaupt et al. 2022; Kambanis et al. 2025). Future research is needed to compare psychological therapies in general—and specific psychological therapies such as CBT and FBT—to plausible comparison treatments, to control for nonspecific factors such as time and attention (Cuijpers et al. 2016). Further research is also required to identify the active components of these treatments. Most studies included at least some core therapeutic interventions such as psychoeducation, family engagement, exposures, and behavioral experiments. Future studies should employ factorial or multiphase optimization designs to identify the interventions associated with the strongest effects.

Although our meta‐analysis provided evidence of weight increases and decreases in ARFID symptoms, depression, and anxiety from pretreatment to posttreatment, conclusions about the likelihood of psychological therapies leading to ARFID remission remain elusive. This may be due to the lack of consensus on how to define remission from ARFID and the relative recency of available tools to assess remission, such as the PARDI. Indeed, several diagnostic criteria for the disorder (e.g., nutritional deficiencies, supplement dependence, and weight loss) lack a consensus definition (Eddy et al. 2019; Harshman et al. 2019). A potential reason for the lack of consistent nutrition‐related outcomes is that 50% of the treatments were delivered by mental health clinicians, the majority of whom were psychologists, who do not typically receive training in nutritional assessment as part of their scope of practice. Therefore, our review also identified the need for consistency in assessing and reporting dietary patterns, nutrition deficiencies, reliance on enteral nutrition, and changes in oral intake, to align with diagnostic criteria and support a consensus definition of remission. Given the complexity of objectively measuring dietary intake, research quality could be strengthened by using dietetic measures such as blood biochemistry for nutrient deficiencies or analysis of oral intake and assessment of malnutrition. For example, Archibald and Bryant‐Waugh (2023) suggest the use of a 3‐day dietary record for assessing nutritional outcomes. Coming to consensus as a field on the assessment of malnutrition would also be beneficial for evaluating treatment outcomes, as much of the DSM‐5 diagnostic criteria overlaps with the definition of malnutrition (World Health Organization 2024), and this was inadequately defined and assessed in the studies we reviewed.

The statistically meaningful medium reduction in anxiety and depression symptomology highlights that treating ARFID might also have a small positive effect on these symptoms and possibly impact co‐diagnosed disorders of anxiety or depression. Further research investigating the impact of nourishment on depression and anxiety symptoms in individuals with ARFID is required, as well as the impact of medication in conjunction with therapy. We acknowledge that two studies reported an increase in symptoms of depression. This may be due to Shimshoni and Lebowitz's (2020) treatment being pediatric‐focused, with parents completing surveys and reporting perceived mood. Furthermore, the Knatz Peck et al. (2021) treatment (YA‐TBTS) was provided to individuals with all eating disorder diagnoses, and prior lived experience studies of ARFID have suggested that individuals with ARFID who receive general eating‐disorder treatments may feel less understood by providers (LaMarre et al. 2023), which could impact mood. Furthermore, both studies had small sample sizes (n = 6).

4.1. Limitations

This study has a number of limitations that are important to highlight when interpreting findings. Firstly, given there are currently no published large randomized controlled trials for ARFID treatment, this review relied mainly on studies with predesign and postdesign with no control groups. A range of interventions was identified, and the appropriateness of the interventions to specific age groups, developmental stages, and available supports needs further investigation. Further, the data emanate predominantly from individual studies with small sample sizes, N < 50. The small sample sizes and predesign and postdesign limit the generalizability of the findings and reduce the robustness of population‐based inferences. In addition to the limited sample sizes, the limited reporting of phenotypic presentations did not allow for a moderator analysis to provide effect sizes for phenotypes. The low intercoder agreement reflects the initial pre‐consensus rating, conducted on 100% of double‐coded articles before consensus was reached.

4.2. Recommendations for Future Research

There is a strong need for evidence‐based treatments for ARFID. For example, in Australia, the lack of evidence has meant the exclusion of ARFID from government‐funded multidisciplinary support, which other eating disorders can access (McLean and Fuller‐Tyszkiewicz 2024). Further, clinicians are unable to make evidence‐based decisions for treatment options as no guidelines provide recommendations. ARFID differs from other eating disorders and requires specialist support (Kambanis et al. 2024). Given the complexity of ARFID, interdisciplinary practice should be considered in future treatment and research. It is possible that involving clinicians and interventions from other disciplines could increase the observed effects on ARFID psychopathology, which were medium in the current study, rather than large, suggesting room for improvement. Furthermore, determining whether psychological therapies could be implemented by a nonmental health clinician, such as a dietitian (Winten et al. 2025), is an important next step, given the shortage of mental health clinicians and barriers in access to care (LaMarre et al. 2023).

Future studies should aim for greater consistency in describing interventions, including the number of sessions and treatment intensity, and whether a specific manual was followed. In this way, continued research into various treatment modalities and their specific components, particularly those with promising effect sizes, should be encouraged to refine and optimize therapeutic strategies for ARFID. Further interventions and therapeutic strategies should be analyzed with consideration of co‐diagnoses and symptomology beyond anxiety and depression, such as autism, to ensure effectiveness.

In addition, future research should prioritize the use of consistent and validated ARFID‐specific measurement tools to ensure comparability across studies and for replication of use in non‐research settings. Further research on the sensitivity of these measures pre‐ to posttreatment is required (Bryant‐Waugh et al. 2022; Burton Murray et al. 2021). Using validated tools to characterize ARFID phenotypic presentations at pretreatment can reduce heterogeneity by evaluating effect sizes separately for each phenotype. By doing so, the most important interventions for each phenotype can be identified. Using more comprehensive measures of anthropometric and nutritional status beyond weight or BMI, such as malnutrition, would provide a clearer picture of treatment efficacy. A further recommendation is the use of validated tools specific to ARFID and its phenotypic presentation, such as the PARDI or NIAS.

While ARFID diagnoses extend beyond childhood, much of the literature on treatment is focused on pediatrics. Research on clinical interventions should expand to include a broader range of age groups, particularly adults, to help strengthen the evidence base for these populations. Delineating adolescents and adults separately is a recommended next step, given the distinct nutritional and psychological needs of growing adolescents across developmental stages, while ensuring that treatment is not completely separated so that findings from research can be translated across age groups. Research into the ARFID pediatric population should consider the inclusion of pediatric feeding disorder, given the overlap in diagnostic criteria (Estrem et al. 2025).

5. Conclusion

This systematic review and meta‐analysis provide an initial contribution to our understanding of the current efficacy of psychological therapies for ARFID for adolescents and adults. Preliminary findings indicate that existing psychological therapies are overall associated with significant improvements in both primary (i.e., reductions in ARFID psychopathology and increases in weight) and secondary (i.e., anxiety and depression) outcomes. However, the current evidence base is small, of moderate quality, and heavily reliant on pre‐ and post‐case series. As such, additional studies with larger sample sizes utilizing randomized controlled trial designs are needed to provide a more rigorous evaluation of psychological therapies for ARFID compared with alternative interventions. There is also a clear need for standardized measurement tools and consistent reporting to build a more substantial evidence base. The evaluation of augmentation with interdisciplinary approaches and further exploration of novel treatment modalities will also be essential as the field progresses.

Author Contributions

Copeland G. Winten: conceptualization, data curation, formal analysis, writing – original draft, methodology, investigation, project administration, writing – review and editing, software, validation, resources, visualization. Esben Strodl: writing – review and editing, formal analysis, data curation, supervision, methodology. P. Evelyna Kambanis: data analysis, data review, writing – review and editing. Lynda J. Ross: writing – review and editing, data analysis, and supervision. Jennifer J. Thomas: conceptualization, supervision, methodology, validation, writing – review and editing.

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

Ms. Copeland G. Winten acknowledges funding for her Eating Disorders Clinical and Research Program summer research fellowship from the Rubenstein Charitable Foundation and the Australian Government Research Training Program Stipend. Dr. Jennifer J. Thomas acknowledges funding from the National Institute of Mental Health (K24MH135189).

Conflicts of Interest

Drs. P. Evelyna Kambanis and Jennifer J. Thomas receive consulting fees from Equip Health. Dr. Jennifer J. Thomas receives royalties for the sale of her books on feeding and eating disorders from Cambridge University Press and Oxford University Press. The other authors declare no conflicts of interest.

Supporting information

Table S1: Changes in dietary intake from the Recovery from Eating Disorders for Life (REAL) food groups (Hart et al. 2018) and the method used to ascertain food group consumption of 40 studies included in a systematic review of psychological therapies for avoidant/restrictive food intake disorder.

Winten, C. G. , Strodl E., Kambanis P. E., Ross L. J., and Thomas J. J.. 2026. “A Systematic Review and Meta‐Analysis of Psychological Therapies for Avoidant/Restrictive Food Intake Disorder (ARFID) in Adolescents and Adults.” International Journal of Eating Disorders 59, no. 7: 1403–1425. 10.1002/eat.70086.

Data Availability Statement

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