Clinical Outcomes in a Large Sample of Youth and Adult Patients Receiving Virtual Evidence‐Based Treatment for ARFID: A Naturalistic Study

ABSTRACT

Objective

Treatment outcomes research for avoidant/restrictive food intake disorder (ARFID) has been limited to small, mixed‐age feasibility trials in face‐to‐face care settings. This study aims to examine clinical characteristics and treatment outcomes in a large sample of youth and adult patients receiving virtual multidisciplinary team treatment for ARFID.

Method

The sample included N = 783 patients (532 youth and 251 adults) diagnosed with ARFID. Patients received cognitive behavioral therapy for ARFID (CBT‐AR) or family‐based treatment for ARFID (FBT‐ARFID) enhanced by specialized support from a multidisciplinary team. Patients (or caregivers) completed a number of measures assessing ARFID and mood‐related symptoms upon admission and throughout treatment.

Results

Youth patients on weight restoration (56%) started treatment around 85% [84%, 86%] of their target weight, and increased to 94% [93%, 96%] by week 35. Adults on weight restoration (47%) started at 85% [84%, 87%] and reached 92% [90%, 94%]. Scores improved for both groups on all PARDI‐AR‐Q subscales: (sensory sensitivity: b = −0.25 [−0.33, −0.16]; lack of interest: b = −0.08 [−0.16, −0.00]; fear of aversive consequences: b = −0.12 [−0.19, −0.04]). Both youth and adults demonstrated reliable improvements in willingness to try new foods (b = −0.64 [−0.89, −0.37]), anxiety symptoms (b = −0.71 [−0.95, −0.48]), and depression symptoms (b = −0.86 [−1.07, −0.64]).

Discussion

Youth and adult patients demonstrated reliable symptom improvements over the course of treatment across all measures, offering preliminary support for the effectiveness of FBT‐ARFID and CBT‐AR delivered virtually by a multidisciplinary care team.

1.

Summary.

• This is the first large‐scale naturalistic study of ARFID patients receiving evidence‐based treatments from a multidisciplinary team via telehealth.

• Meaningful symptom and mood improvements were seen across a number of validated measures in both youth and adult patients.

• This study offers preliminary support to acceptability and effectiveness of telehealth delivery of leading ARFID treatments.

2. Introduction

Avoidant/restrictive food intake disorder (ARFID), an eating disorder (ED) introduced in 2013 in the Fifth Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM‐5), is characterized by a disturbance in eating patterns resulting in intake that is restricted, avoidant and/or highly selective. Studies of non‐clinical samples report wide‐ranging prevalence estimates ranging from 0.3% to 15.5% in youth (Sanchez‐Cerezo et al. ²⁰²³) and 0.3% to 4.8% in adults (Hay et al. 2017; D'Adamo et al. ²⁰²³;). Unlike other EDs, diagnostic criteria exclude weight and shape concerns as a core feature. Symptoms of ARFID may be categorized into one or more of the following profiles: (1) Sensory sensitivity or avoidance, where foods are avoided due to their sensory properties; (2) lack of interest, where one demonstrates a persistent poor appetite and limited intake; or (3) fear of aversive consequences from eating, such as choking or vomiting. Given the relative newness of ARFID compared to other EDs, research on treatment outcomes has been limited.

To date, ARFID treatment studies have primarily consisted of small feasibility trials. The two leading, manualized treatments for ARFID include cognitive behavioral therapy for ARFID (CBT‐AR) and family‐based treatment for ARFID (FBT‐ARFID). Both are drawn from current evidence‐based ED treatments and anxiety treatments, as well as from established feeding therapies. Thomas et al. (2020, 2021) tested the feasibility and preliminary efficacy of CBT‐AR in small samples of youth (N = 20) and adult (N = 15) patients with ARFID. Patients received 20–30 sessions of CBT‐AR (1–2 times weekly) over approximately 9 months. Both ARFID symptoms and weight (for underweight patients) demonstrated significant improvement over the treatment period (Thomas et al. 2020; Thomas et al. ²⁰²¹). In a case series describing adaptations of FBT‐ARFID for three youth patients (ages 8, 9, and 11) with ARFID, all patients made improvements in weight and ARFID symptoms, though none of the three achieved remission by the end of the study period (Lock et al. ^2019a). Another small study (N = 28) assessing the feasibility of conducting a larger randomized trial comparing FBT‐ARFID to usual care demonstrated both acceptability for families and significant moderate to large improvements in weight and clinical symptoms for patients receiving FBT‐ARFID (Lock, Sadeh‐Sharvit, and L'Insalata ^2019b). This preliminary evidence shows promise for the positive outcomes of CBT‐AR for adults and FBT‐ARFID for youth, and larger trials for both are underway (Lock, Sadeh‐Sharvit, and L'Insalata ^2019b).

Additional evidence supports that evidence‐based therapeutic ARFID treatments can be enhanced with a multidisciplinary care team that includes additional healthcare providers such as psychiatrists, dietitians, and occupational therapists (Fonseca et al. 2024). Moreover, to expand the reach of evidence‐based treatments, we need to consider innovative ways to deliver ARFID treatment more broadly. Emerging evidence supports that effective ED treatment can be delivered virtually through telemedicine, and rival outcomes from in‐person treatment (Anderson et al. 2017; Sproch and Anderson ²⁰¹⁹; Matheson, Bohon, and Lock ²⁰²⁰; Hellner et al. ²⁰²¹; Steinberg et al. ²⁰²³). However, these studies did not evaluate outcomes from ARFID treatments delivered virtually.

This study helps fill these research gaps by (1) utilizing a large cohort of youth and adults to evaluate clinical outcomes from the aforementioned evidence‐based treatments, and (2) describing outcomes from a treatment approach that employs a multidisciplinary care team approach and delivers treatment virtually.

3. Methods

3.1. Study Design and Participants

This study is a pre‐post single cohort design evaluating the effectiveness of virtual treatment among patients with ARFID. The sample (N = 783 patients) included patients diagnosed with ARFID who received treatment between August 14, 2023, and July 31, 2024. ARFID diagnosis was determined by the therapist on the treatment team during a comprehensive intake appointment. This included 532 youth (mean age = 11.8 years; SD = 3.5; range 3–17 years) and 251 adults (mean age = 24.9 years; SD = 7.7; range 18–60 years).

3.2. Treatment Approach

This single‐site study took place at Equip Health, which provides nationwide virtual ED treatment to patients of all ages. Treatment is voluntary, and patients are not required to remain in treatment for any length of time. Patients, or caregivers of minors, provide informed consent upon admission for treatment data to be evaluated and published for research purposes. This study was reviewed by the Western Institutional Review Board and was deemed exempt from IRB oversight.

Both FBT‐ARFID and CBT‐AR are utilized as a treatment for ARFID. Treatment selection for each patient was based on age, parental involvement, patient motivation, and clinician or family preference. CBT‐AR was the primary treatment used for adults with ARFID, while both FBT‐ARFID and CBT‐AR were used for children and teens with ARFID. As this study was naturalistic and did not involve random assignment of treatment modalities and given that patients occasionally switched treatment modalities (e.g., started with FBT‐ARFID and switched to CBT‐AR), treatment effectiveness was examined by age cohort and not by treatment modality. Patients and families start treatment by having appointments with providers weekly, and this is tapered throughout treatment as needed based on patient and family needs. More detail on the treatment approach can be found elsewhere (Steinberg et al. 2023). Providers receive extensive training on both treatment modalities and receive regular supervision.

3.2.1. Multidisciplinary Family‐Based Treatment for ARFID (FBT‐ARFID)

FBT‐ARFID is a treatment developed with children and early adolescents and has been tested in ages 5–12 (Lock, Sadeh‐Sharvit, and L'Insalata ^2019b). Unlike typical outpatient treatment using FBT‐ARFID, each patient at Equip includes a multidisciplinary team consisting of a therapist, a registered dietitian (RD), a medical provider, and a peer and/or family mentor (Hellner et al. 2021). The peer mentor is an individual who has recovered from an ED and serves as a support to the patient. The family mentor has previously supported a loved one in recovery from an ED and provides support to caregivers. FBT‐ARFID consists of three phases and focuses primarily on weight restoration (if needed) and dietary expansion (Lock, 2021). Treatment begins with empowering caregivers to take charge of managing ARFID behaviors and gradually returns age‐appropriate control to the patient while working to generalize eating across multiple settings.

3.2.2. Multidisciplinary Cognitive‐Behavioral Therapy‐ARFID (CBT‐AR)

CBT‐AR is a modular treatment with individual and family‐supported variations and was developed for individuals ages 10 and up. CBT‐AR progresses through four stages: (1) psychoeducation and early change; (2) treatment planning; (3) treatment of maintaining mechanisms; and (4) relapse prevention (Thomas et al. 2020). Stage 3 consists of 3 modules with specific approaches for addressing sensory sensitivity, fear of aversive consequences, and lack of interest. Similar to FBT‐ARFID, Equip enhances manualized CBT‐AR by assigning each patient a multidisciplinary team, composed of a therapist, RD, medical provider, and peer mentor.

3.3. Measures

Self‐report measures were completed by patients and/or caregivers as part of standard assessment protocols. Measures were completed electronically at treatment admission and monthly for the following two months, then every three months thereafter. For patients under the age of thirteen, the measures are typically completed by caregivers. Patients or caregivers received a notification to complete surveys via their electronic health record at the designated intervals.

3.3.1. Anthropometrics and Target Weight Assessment

At the start of treatment, patients or caregivers of youth patients receive instructions on proper procedures for obtaining weight. Equip recommends that weights are taken in the morning, after voiding, prior to any oral intake and patients are asked to wear light clothing. Weights are taken twice weekly and submitted via text message to their HIPAA‐compliant electronic health record. Height is self‐reported upon admission and is also verified against the most recent medical records and growth chart data. For youth patients, height is measured throughout the treatment course every few months by a caregiver or in‐person medical provider. For patients requiring weight restoration, the RD determines a minimum target weight. Assessment of target weight for both youth and adult patients includes an evaluation of current dietary intake and ED behaviors, current and historical physical activity patterns, and all available medical data. The evaluation for youth patients also relies heavily on weight and stature data from each patient's CDC age‐adjusted BMI growth chart. Deviations from the patient's growth trajectory at the onset of the ED are noted, with the objective generally being to restore the patient to their premorbid growth trajectory (Steinberg et al. 2023).

3.3.2. Pica, ARFID and Rumination Disorder Interview ARFID Questionnaire (PARDI‐AR‐Q)

The PARDI‐AR‐Q is a 32‐item measure that screens for the presence and severity of ARFID symptoms and key features. In this treatment setting, a shortened version of the PARDI‐AR‐Q was used, which includes the nine items comprising the ARFID symptom presentation subscales, with scores ranging from 0 to 6 designed to assess the presence and severity of each of the three ARFID profiles (Table 1). The severity subscale was not evaluated in this study, as the severity items were initially excluded to minimize patient and family burden. The sensory sensitivity subscale measures avoidance of foods due to their sensory properties; the lack of interest subscale measures symptoms of poor appetite and/or limited intake; the third subscale measures fear of aversive consequences from eating, such as choking or vomiting. Higher scores on each of the subscales indicate greater higher levels of ARFID symptoms (Bryant‐Waugh et al. ²⁰²²). In evaluating symptom change over treatment, each subscale was examined both across all patients with ARFID, and across all patients who reported > 0 for that subscale, as minimal change would be expected for patients who aren't experiencing the symptom in question (Table 2). The PARDI‐AR‐Q has been validated in patients over age 13 (Bryant‐Waugh et al. ²⁰²²). For patients 13 and under, there is a parent version available for caregivers to complete. In the current study, we only used the patient version but caregivers were instructed to help patients fill it out if they were unable to complete it themselves. The Cronbach alpha was 0.82 (0.76, 0.89) for patients in the current study.

TABLE 1.

Items from the PARDI‐AR‐Q and FNS used to evaluate changes in ARFID symptoms.

PARDI‐AR‐Q

Sensory subscale items Over the past month, have you been particularly sensitive to variation in taste (e.g., noticing slight differences in the taste of foods), which has put you off from eating any foods or trying any new foods? Responses range from 0 (=no negative effect/no particular sensitivity to taste) to 6 (=extremely negative effect, for example, leading to refusal to eat many foods, sticking only to a limited number of preferred foods, or extreme caution when eating). Over the past month have you been particularly sensitive to the texture or consistency of food, which has put you off eating any foods or trying any new foods (e.g., do you stick to foods of a certain texture only or have you had difficulty eating foods that have different textures mixed together such as pasta with sauce or sandwiches)? Responses range from 0 (=no negative effect/no particular sensitivity) to 6 (=extremely negative effect, for example, leading to refusal to eat many foods, sticking only to a limited number of preferred foods, or extreme caution when eating) Over the past month, have you been particularly sensitive to the appearance of food, which has put you off eating any foods or trying any new foods (e.g., if food does not look “right”, such as burnt ends of chips/fries, broken biscuits/cookies, or being the “wrong” color)? Responses range from 0 (=no negative effect/no particular sensitivity) to 6 (=extremely negative effect, for example, leading to refusal to eat many foods, sticking only to a limited number of preferred foods, or extreme caution when eating) Lack of interest subscale items Over the past month, how often have you forgotten to eat or found it difficult to make time to eat? Responses range from 0 (=never) to 6 (=always). Over the past month, how often have you lacked enjoyment in food or eating (even if only certain foods)? Responses range from 0 (=never) to 6 (=always). Over the past month, how often have you felt full before your meal is finished, or stopped eating sooner than others because you had had enough? Responses range from 0 (=never) to 6 (=always). Fear of aversive consequences subscale items Over the past month have you been avoiding or restricting the amount or type of food you eat, because you were afraid that something bad might happen, like being sick, choking, having an allergic reaction, or being in pain? Responses range from 0 (=never) to 6 (=always) Over the past month have you avoided eating situations because you were worried something bad might happen, like being sick, choking, having an allergic reaction, or being in pain while eating (e.g., because you might be served something you usually avoid for these reasons, or because you have had a bad experience in the past)? Responses range from 0 (= never) to 6 (=always). Over the past month have you had any physical feelings of panic or anxiety (examples might include a racing heart, sweaty palms, feeling sick) when you have seen something that has made you think something bad might happen, like being sick, choking, having an allergic reaction, or being in pain while eating? Responses range from 0 (=never) to 6 (=always). Shortened Food Neophobia Scale (FNS) I am constantly sampling new and different foods (reverse scored). Responses range from 1 (=disagree strongly) to 7 (=agree strongly). I don't trust new foods. Responses range from 1 (=disagree strongly) to 7 (=agree strongly). If I don't know what is in a food, I won't try it. Responses range from 1 (=disagree strongly) to 7 (=agree strongly). I am afraid to eat things I have never had before. Responses range from 1 (=disagree strongly) to 7 (=agree strongly). I am very particular about the foods I will eat. Responses range from 1 (=disagree strongly) to 7 (=agree strongly). I will eat almost anything (reserve scored). Responses range from 1 (=disagree strongly) to 7 (=agree strongly)

TABLE 2.

Demographic and clinical characteristics.

Demographic	Youth (under 18)	Adults (18 and over)
Patients (N)	532	251
Age (M, SD)	11.8 (3.5)	24.9 (7.7)
Gender (N, %)
Cisgender girls/women	252 (47%)	161 (64%)
Cisgender boys/men	255 (48%)	55 (22%)
Transgender women	2 (0.4%)	3 (1%)
Transgender men	7 (1%)	10 (4%)
Nonbinary	16 (3%)	22 (9%)
Ethnicity (N, %)
White	380 (71%)	192 (77%)
Hispanic	31 (6%)	14 (6%)
Asian	24 (5%)	8 (3%)
Black	11 (2%)	8 (3%)
Other, multiracial/biracial	67 (13%)	25 (10%)
Did not report	19 (4%)	4 (2%)
Weight restoration (N, %)
Require weight restoration	295 (56%)	118 (47%)
Did not require weight restoration	237 (44%)	133 (53%)
BMI at admission (median [25%, 75%])	—	19.1 [17.5, 23.9]
Percentile BMI at admission (N, %)		—
25th percentile or lower	204 (43%)	—
25th–50th percentile	104 (22%)	—
50th–75th percentile	82 (17%)	—
75th percentile or higher	81 (17%)	—

3.3.3. Shortened Food Neophobia Scale (FNS)

The original FNS comprises 10 Likert‐scale items with scores ranging from 1 to 7 and assesses fear of new or unfamiliar foods (Pliner and Hobden 1992). Food neophobia is highly correlated with picky or selective eating and may be an important mechanism of change involved in the treatment of the sensory‐sensitivity presentation of ARFID (Dovey et al. 2008; Burton Murray et al. ²⁰²²). Due to concerns about cultural bias with some of the items, the FNS was adapted for use in this treatment setting to include only six of the original 10 questions (Table 1). In this shortened version, total scores range from 6 to 42, with higher scores indicating more severe food neophobia. The FNS was originally developed for adolescents and adults, but a child version administered to parents has been developed (Firme et al. 2023). In the current study, we only used the patient version but caregivers were instructed to help patients fill it out if they were unable to complete it themselves. The Cronbach alpha was 0.89 (0.85, 0.93) for patients in the current study.

3.3.4. Depression and Anxiety Measures

The Generalized Anxiety Disorder Scale‐7 (GAD‐7) is a seven‐item tool used to assess the severity of anxiety symptoms. Items are scored from 0 to 3, with higher total scores indicating increased severity of anxiety (Spitzer et al. 2006). Scores greater than 10 indicate the presence of clinically significant anxiety. The GAD‐7 has been validated in ages 12 and up (Mossman et al. 2017), however, in the current study caregivers were instructed to help patients fill it out if they were unable to complete it themselves. The Cronbach alpha was 0.90 (0.87, 0.92) for patients in the current study.

The Patient Health Questionnaire‐8 (PHQ‐8) is an eight‐item Likert‐scale assessment of symptoms of depression that one has found bothersome ‘over the last two weeks’ with score options ranging from 0 to 3, with higher total scores indicating greater severity of depression Scores greater than or equal to 10 indicate clinically significant depression (Kroenke, Spitzer, and Williams ²⁰⁰¹). The PHQ has been validated in ages 12 and up (Fonseca‐Pedrero et al. ²⁰²³), however, in the current study caregivers were instructed to help patients fill it out if they were unable to complete it themselves. The Cronbach alpha was 0.85 (0.80, 0.91) for patients in the current study.

3.4. Statistical Analysis

Descriptive statistics, including mean (standard deviation) and patient counts (percentage), were used to describe patient characteristics at admission. Because this is a naturalistic study and the time spent in treatment can be different for each individual patient, a survival analysis was initially conducted to describe this patient sample's expected time in treatment. This average median point in time was used as a treatment endpoint to evaluate outcomes. Outcomes at week 16 were also provided, as this generally corresponds to the end‐of‐treatment for clinical trials of traditional FBT (Lock et al. ^2019a; Lock, Sadeh‐Sharvit, and L'Insalata ^2019b).

Bayesian linear mixed‐effects models were used to evaluate outcomes throughout treatment. This accounted for varying lengths of treatment among patients, as well as individual differences in trajectories over time. Where applicable, the week 0 scores were reported, and the outcomes were at weeks 16 and 35. The results in the text are reported as the estimate [2.5%, 97.5% percentiles] for youth, adults, and the overall population. Bayesian estimation returns a posterior probability distribution for each parameter value. These lower and upper bounds represent the plausible range of the coefficient consistent with the data. They are different from confidence intervals but play a similar role. One can think of values outside this plausible range as unlikely, given the data.

In Wilkinson's notation, the model used to capture survey response over time was y ~ 1 + is_adult*log(treatment_week) + (1 + log(treatment_week) | patient), where is_adult was an indicator variable indicating whether a patient was a youth (under 18) or adult (18 and over). Conservative priors were used for the model parameters centered around the null hypothesis with a large enough standard deviation to capture non‐null parameter estimates. This approach requires the data to overcome a small prior favoring the null hypothesis. Available data for each patient was included in the analyses. This allows for robust estimation of both the within‐person (random) and fixed effects. Complete‐case analysis was used here for simplicity, but through sensitivity analysis, it was confirmed that two methods of imputation (multiple imputation through chained equations (White, Royston, and Wood ²⁰¹¹), and Bayesian model imputation) do not change the results or conclusions.

All analyses were conducted using R version 4.4.0 (R Core Team ²⁰²²) and tidyverse version 2.0.0 (Wickham, Averick, and Bryan ²⁰¹⁹). Fitting was done using the brms package version 2.21.0 (Bürkner 2021), which is a wrapper around the Stan probabilistic programming language (Stan Development Team ²⁰²²). The survminer package version 0.4.9 was used for survival analysis. The target package version 1.7.0 was used for project management (Landau 2021).

4. Results

Demographic and clinical characteristics of the patients are shown in Table 2. Patients were predominantly White (73%) and under the age of 18 (68%). The mean age was 11.8 years (SD = 3.5) for youth patients and 24.9 years (SD = 7.7) for adults. Approximately half of the youth patients were cisgender girls and half cisgender boys, though the proportion of cisgender women in the adult group was substantially higher. Roughly half (53%) of all patients required weight restoration at the start of treatment. Sixty‐five percent of youth patients were under the 50th percentile for age‐adjusted BMI, and the median BMI across adults was 19.1 kg/m². The median time in treatment was 35 weeks for both age groups.

4.1. Weight Restoration

Both youth and adult patients on weight restoration gained weight throughout treatment (b = 0.03 [0.02, 0.03]). Youth patients on weight restoration started treatment around 85% [84%, 86%] of their target weight, and increased to 94% [93%, 96%] by week 35. Adult patients on weight restoration started at 85% [84%, 87%] and reached 92% [90%, 94%] by week 35.

4.2. ARFID Symptoms

Mean scores for each subscale at admission, 16 weeks, and 35 weeks are shown in Tables 3 and 4. Youth and adult patient scores improved on all PARDI‐AR‐Q subscales over time: (sensory sensitivity: b = −0.25 [−0.33, −0.16]; lack of interest: b = −0.08 [−0.16, −0.00]; fear of aversive consequences: b = −0.12 [−0.19, −0.04]). Similar trends are seen when restricted to only those patients with scores > 0 at baseline at each of the subscales. Scores on the abbreviated FNS also improved reliably over time in both youth and adults (b = −0.64 [−0.89, −0.37]).

TABLE 3.

Survey score estimates for youth patients (under 18 years old) at admission, week 16, and week 35.

	Youth
Survey	Score at week 0 Mean [CI]	Score at week 16 Mean [CI]	Score at week 35 Mean [CI]
Percent of Target Weight (%) Patients on weight restoration	85% [84%, 86%]	92% [91%, 93%]	94% [93%, 96%]
PARDI‐AR‐Q: Sensory based avoidance subscale
All youth patients	3.59 [3.37, 3.80]	2.89 [2.66, 3.12]	2.70 [2.44, 2.98]
Patients who score $>$ 0 on this subscale	3.76 [3.53, 3.99]	2.98 [2.73, 3.24]	2.78 [2.50, 3.10]
PARDI‐AR‐Q: Lack of Interest Subscale
All youth patients	2.92 [2.74, 3.11]	2.69 [2.48, 2.90]	2.63 [2.38, 2.88]
Patients who score > 0 on this subscale	3.08 [2.89, 3.28]	2.89 [2.65, 3.14]	2.84 [2.56, 3.14]
PARDI‐AR‐Q: Fear of Aversive Consequences Subscale
All youth patients	1.76 [1.56, 1.98]	1.43 [1.22, 1.64]	1.34 [1.09, 1.59]
Patients who score > 0 on this subscale	2.51 [2.26, 2.77]	1.84 [1.54, 2.14]	1.68 [1.34, 2.03]
Modified Food Neophobia Scale (FNS)	35.6 [34.8, 36.5]	33.8 [32.9, 34.8]	33.4 [32.3, 34.4]
Anxiety Scores (GAD‐7)
All youth patients	8.36 [7.70, 8.98]	6.33 [5.68, 6.98]	5.79 [5.03, 6.55]
Patients with scores $\ge$ 10 at admission	14.9 [14.1, 15.6]	9.94 [8.58, 11.31]	8.54 [6.83, 10.3]
Depression Scores (PHQ‐8)
All youth patients	7.61 [6.97, 8.25]	5.20 [4.53, 5.85]	4.55 [3.80, 5.30]
Patients with scores $\ge$ 10 at admission	14.9 [14.0, 15.8]	8.79 [7.30, 10.4]	7.35 [5.60, 9.20]

TABLE 4.

Survey score estimates for adult patients (18 years and older) at admission, week 16, and week 35.

	Adults
Survey	Score at week 0 Mean [CI]	Score at week 16 Mean [CI]	Score at week 35 Mean [CI]
Percent of Target Weight (%) Patients on weight restoration	85% [84%, 87%]	91% [89%, 92%]	92% [90%, 94%]
PARDI‐AR‐Q: Sensory based avoidance subscale
All adult patients	3.64 [3.38, 3.89]	2.85 [2.56, 3.11]	2.64 [2.31, 2.96]
Patients who score $>$ 0 on this subscale	3.86 [3.60, 4.12]	2.97 [2.66, 3.27]	2.74 [2.38, 3.09]
PARDI‐AR‐Q: Lack of Interest Subscale
All adult patients	3.67 [3.45, 3.87]	3.03 [2.78, 3.28]	2.86 [2.56, 3.16]
Patients who score > 0 on this subscale	3.69 [3.47, 3.90]	3.06 [2.78, 3.32]	2.90 [2.58, 3.22]
PARDI‐AR‐Q: Fear of Aversive Consequences Subscale
All adult patients	2.26 [2.02, 2.51]	1.79 [1.53, 2.04]	1.66 [1.37, 1.96]
Patients who score > 0 on this subscale	2.90 [2.62, 3.17]	2.29 [1.97, 2.60]	2.14 [1.79, 2.50]
Modified Food Neophobia Scale (FNS)	35.1 [34.1, 36.0]	32.5 [31.4, 33.6]	31.9 [30.6, 33.1]
Anxiety Scores (GAD‐7)
All adult patients	10.2 [9.47, 10.9]	8.85 [8.05, 9.63]	8.49 [7.55, 9.41]
Patients with scores $\ge$ 10 at admission	15.6 [14.9, 16.3]	12.5 [11.2, 13.9]	11.7 [9.96, 13.4]
Depression Scores (PHQ‐8)
All adult patients	10.5 [9.75, 11.3]	9.19 [8.40, 9.98]	8.84 [7.92, 9.73]
Patients with scores $\ge$ 10 at admission	15.7 [14.9, 16.4]	13.3 [12.0, 14.6]	12.8 [11.2, 14.3]

4.3. Anxiety and Depression Symptoms

Mean scores and changes over time can be found in Table 3 for youth and Table 4 for adults. Both GAD‐7 scores (b = −0.71 [−0.95, −0.48]) and PHQ‐8 scores (b = −0.86 [−1.07, −0.64]) improved reliably over time in both youth and adults. When restricted to only those patients who started treatment with scores $\ge$ 10 (scores indicating the presence of clinically significant symptoms), larger improvements are found in GAD‐7 (b = −1.78 [−2.29, −1.25]) and PHQ‐8 (b = −2.15 [−2.67, −1.64]) scores over time.

5. Discussion

In the largest known sample of treatment‐seeking youth and adults with ARFID, findings indicate that virtual, evidence‐based treatment led to meaningful reductions in ARFID symptoms, as well as improvements in anxiety and depression. Both youth and adults showed similar, significant improvements in ARFID symptoms, mood symptoms, and weight status throughout the treatment course. Findings corroborate previous research evaluating outcomes of FBT‐ARFID and CBT‐AR treatment (Thomas et al. 2020; Thomas et al. ²⁰²¹; Lock et al. ^2019b). However, higher baseline PARDI‐AR‐Q subscale scores and greater reductions in scores over time were observed in both youth and adults relative to those reported in CBT‐AR feasibility studies (Thomas et al. 2020; Thomas et al. ²⁰²¹). The effects on ARFID symptoms observed in this sample may be attributed to longer time in treatment and/or ongoing specialized multidisciplinary team support throughout their time in the program. However, FNS score improvements were more modest in this sample, particularly among the youth population. While Thomas and colleagues noted reductions of 17% (youth) and 12% (adults), here, a 6% improvement in youth and 9% improvement was observed in adults. The large difference in sample size relative to existing ARFID treatment outcome research likely also explains some of the differences. Similar effects were found for anxiety and depression reductions as have been found in previous ARFID treatment studies; (Thomas et al. 2020; Thomas et al. ²⁰²¹) both youth and adults saw reductions of approximately 20%–30% in anxiety or depression symptoms.

Treatment was delivered virtually and by a multidisciplinary team. Results show positive outcomes that align with prior studies on ED treatment delivered via telemedicine. (Anderson et al. 2017; Sproch and Anderson ²⁰¹⁹; Matheson, Bohon, and Lock ²⁰²⁰; Hellner et al. ²⁰²¹; Steinberg et al. ²⁰²³). This study is the first that we are aware of to deliver ARFID treatment virtually. To truly evaluate the effects of treatment delivered virtually, future research should evaluate more rigorously how virtual treatment compares to in‐person treatment. Moreover, the use of a multidisciplinary care treatment approach and team configuration can vary across treatments, making it challenging to compare across studies; however, many of the clinical outcomes found in the current study appear to be relatively comparable to multidisciplinary treatment approaches in the literature (Sharp et al. 2016; Ornstein et al. ²⁰¹⁷; Shimshoni, Silverman, and Lebowitz ²⁰²⁰; MacDonald, Liebman, and Trottier ²⁰²⁴).

Notably, given the large sample size, we can evaluate the personal characteristics of patients and compare findings to what is typically seen in the ARFID literature. Approximately half of the youth patients in the current study were cisgender boys and half cisgender girls, aligning with existing literature demonstrating equivalence across boys and girls (Sanchez‐Cerezo et al. ²⁰²³). Gender distribution in adults in this sample also fit with existing literature, as a greater proportion of women (64%) was observed. n In a large community sample of adults (N = 47,705), 76% of those screening positive for ARFID were women (D'Adamo et al. ²⁰²³) and 62% of a clinical sample of adults with ARFID were women (MacDonald, Liebman, and Trottier ²⁰²⁴).

Moreover, the weight status at the start of treatment indicated that about ⅓ of youth patients were above the 50th percentile, the median BMI for adults was in the “normal” range, and only about half required weight restoration throughout treatment. This suggests that weight status may not be the best to reliably evaluate ARFID severity as individuals with ARFID are less likely to demonstrate acute weight loss relative to other ED diagnoses (Yule et al. 2021; Duncombe Lowe et al. ²⁰¹⁹; Downey, Richards, and Tanner ²⁰²³). Findings echo those of D'Adamo et al. (2023) demonstrating an average BMI in the normal range in a large community sample of adults who screened positive for ARFID, and Van Buuren et al. (2023) showing only 11% of youth with ARFID in the community sample as underweight. This underscores the importance of using validated tools to capture change in ARFID symptoms given that weight gain, for many, is not a target of treatment.

This study has a number of important strengths. First, the large sample size allowed for a meaningful evaluation of symptoms and treatment responses across both youth and adult patients. This was also the first large‐scale naturalistic study of ARFID patients receiving treatment via telehealth. Second, care was delivered by a multidisciplinary team, in congruence with American Psychiatric Association (APA) guidelines recommending a specialized multidisciplinary care team across all ED diagnoses to expertly address psychological, medical, and nutritional concerns (American Psychiatric Association ²⁰²³). Third, a number of validated measures were used to capture symptom change, which allows for a more holistic view of the impact of ARFID treatment. Finally, the entire sample included patients with an ARFID diagnosis confirmed by clinical interview by the therapist, in contrast to methods commonly used to approximate a diagnosis, such as retrospective chart review (Bryant‐Waugh et al. ²⁰²²).

This study also had some limitations that are worth noting. First, this was a non‐randomized design, though randomization of patients to different interventions would have been impractical as patients were assigned an appropriate modality based on age and clinical need. Second, this was not a comparison of CBT‐AR and FBT‐ARFID, but a description of the patient's clinical presentation and treatment response. The use of these treatment modalities is mentioned because they are clinically appropriate for adults and youth, respectively. Third, patients' ARFID profiles were not categorized for analyses based on which ARFID presentations were most prominent at the start of treatment, which may have diluted the effects on the various PARDI‐AR‐Q subscale scores. However, emerging evidence suggests that symptoms typically occur on a continuum, as profiles are not mutually exclusive (Thomas et al. 2017; Zickgraf et al. ²⁰¹⁹; Duncombe Lowe et al. ²⁰¹⁹). Fourth, patients are not required to end treatment with outside providers therefore, it is possible this could impact some of the effects related to depression and anxiety. However, providers continuously monitor and address mood and anxiety‐related symptoms as part of standard treatment so it is reasonable to assume that these symptoms were being addressed solely by treatment providers for nearly all patients in this study. Finally, this study did not evaluate the severity of the impact of struggling with ARFID using the PARDI‐AR‐Q severity subscale. The severity subscale questions were later added to treatment for evaluation purposes, but data were not available at the time of this study. Future analysis of ARFID symptom change should include this construct.

6. Conclusions

Both youth and adult patients receiving ARFID treatment demonstrated significant improvements across weight, ARFID symptoms, and mood symptoms, building upon previous research with a large cohort of patients receiving evidence‐based ARFID treatment delivered virtually. Although findings must be interpreted with regard to limitations, they add to the research literature on how to improve outcomes in patients with ARFID and highlight areas of interest for future research.

Author Contributions

Megan Hellner: conceptualization, methodology, writing – original draft, writing – review and editing. Kelly Cai: conceptualization, data curation, formal analysis, methodology, writing – original draft, writing – review and editing. Jessica H. Baker: conceptualization, methodology, writing – review and editing. Jessie Menzel: conceptualization, methodology, writing – review and editing. Dave Freestone: conceptualization, data curation, formal analysis, methodology, writing – review and editing. Dori M. Steinberg: conceptualization, methodology, writing – review and editing.

Conflicts of Interest

All authors hold equity in Equip Health.

Data Availability Statement

The data that support the findings of this study are derived from patient medical record data. The data are not publicly available and individual data cannot be shared due to privacy restrictions.