Assessing Fears of Negative Consequences in Children with Symptoms of Avoidant Restrictive Food Intake Disorder

Abstract

Objective:

Fear of aversive consequences (FOAC), such as choking or vomiting, is an important associated feature of Avoidant/Restrictive Food Intake Disorder (ARFID). However, the manifestation of FOAC in young children is poorly understood. This study aimed to describe the fears of children with ARFID symptoms and examine the concordance between parent and child ratings of fear.

Method:

Child-reported FOAC was assessed using an interview designed for children between 6 and 10 years old, the Gustatory Avoidance and Gastrointestinal Stress Symptoms Interview. Parents were administered a semi-structured diagnostic interview regarding their child’s symptoms, the Pica, ARFID, and Rumination Interview.

Results:

Among 68 children with ARFID diagnoses or symptoms (41.2% female, 85.3% White, mean age = 8.2 years, SD = 1.1 years; range 5.2 – 9.9 years), 91.2% of children endorsed at least one fear relative to 26.5% of parents. Among parent-child dyads, 36.8% disagreed about the child’s fear of stomach pain (κ = .12) and 48.5% disagreed about the child’s fear of vomiting, (κ = .08), both indicating low inter-rater reliability. On average, children endorsed 4.3 (SD = 2.3) fears out of 9 options. The most frequently endorsed fears were that food will “taste bad,” (n = 43, 63.2%); “make you gag” (n = 37, 54.4%), and “look disgusting” (n = 36, 52.9%).

Discussion:

Findings highlight ways in which fear may manifest in children with ARFID that are not easily discernable by adults. Greater precision in depicting childhood fears may facilitate the earlier detection of problematic eating behaviors.

Introduction

Avoidant/Restrictive Food Intake Disorder (ARFID) was a critical addition to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (American Psychiatric Association, 2013) to aid in the diagnosis of children, adolescents, and adults with impairing food avoidance and/or restriction not captured by existing diagnostic categories. ARFID is an eating disorder characterized by at least one of the following: significant weight loss or failure to achieve sufficient growth or weight milestones, nutritional deficiencies, dependence on supplementation to achieve adequate nutritional quantity or dietary quality, or significant impairment in psychosocial functioning (American Psychiatric Association, 2022). Unlike those with anorexia nervosa (AN), avoidance or restriction of eating in individuals with ARFID is not driven by preoccupation with weight or body image (Zimmerman & Fisher, 2017).

Emerging data suggest that individuals with ARFID may have an earlier age of onset relative to other disorders characterized by food restriction and avoidance, such as AN. For example, the average age of presentation of ARFID ranged from 2.9 – 17.8 years across five eating disorder programs or treatment trials (Becker et al., 2019; Breiner et al., 2021; Lowe et al., 2019; Nicely et al., 2014; Watts et al., 2023). In comparison, the modal age of onset for AN was reported to be between ages 14 and 18 years across inpatient and outpatient eating disorder treatment samples (Becker et al., 2019; Jagielska & Kacperska, 2017; Norris et al., 2014). Indeed, the onset of ARFID may be younger than reported as patterns of impairing food avoidance may emerge before age 5 with some studies reporting a modal onset of food selectivity between the ages of 2 and 3 years old (e.g., Zickgraf et al., 2019). These data indicating early onset combined with those suggesting that ARFID may persist into adulthood if left untreated (Dumont et al., 2019), highlight the importance of early identification and intervention of problematic eating in young children.

Defining the bounds of healthy growth in young children is complicated. Separating children who are growing optimally at a low weight from those who are weight-suppressed due to psychological factors, such as reduced intake related to fear, is limited by a child’s capacity to articulate or demonstrate fears. The first few years of life mark a critical developmental window for a child to learn to identify and label sensations from the body that communicate emotional, motivational, and physical needs (Hietanen et al., 2016; Zucker et al., 2017). Whereas older children (ages 10–12) and adolescents may utilize abstraction or verbal expression to articulate their feelings, younger children generally do not have the same language capacity (Bauer, 1976; Grills and Ollendick, 2002; Grosse et al., 2021; Nook et al., 2020). As one example, a case study (Zucker et al. 2019) noted that a four- year-old child with ARFID displayed fears of certain bodily sensations, but had difficulty acknowledging or articulating them. The content of children’s fears may be mediated, in part, by their level of cognitive development (Muris & Field, 2011), with younger children (ages 4–6) reporting more global or imaginary fears (e.g., monsters, ghosts) while older children (ages 10–12) tend to report more realistic and specific fears (e.g., bodily harm, physical danger, or social evaluation; Bauer, 1976; Westenberg et al., 2004). One’s capacity to recall experiences may also contribute to age-related differences, given evidence that younger children (ages 7–9) are not as reliable as older children (ages 10–12) or adults in their recall or recognition of past events (Koriat et al., 2001). Older children also tend to understand more complex or mixed emotional experiences (Burkitt et al., 2018; Eastabrook et al., 2014) and recognize more internalizing symptoms as they develop (Thompson Jr et al., 1993). In sum, there is substantial evidence that there are developmental constraints on a child’s ability understand or communicate their fears, limits that may compromise adults’ abilities to fully comprehend the nature and experience of fears of young children with ARFID.

Data among adolescent and adult populations may provide insight into FOAC among children with ARFID. Ninety-two percent of adults with ARFID symptoms presenting to a neurogastroenterology clinic cited a fear of gastrointestinal symptoms as a reason for their avoidant/restrictive eating (Murray et al., 2020). In a younger sample of individuals hospitalized for eating disorder treatment (mean = 13.7 years, SD = 2.4 years), 43% of participants with ARFID symptoms endorsed FOAC including fears of nausea, pain, or choking (Norris et al., 2018). Notably, it appears that ARFID with the FOAC predominant presentation is more commonly seen in inpatient settings due to the associated precipitous weight loss and subsequent health consequences related to acute food refusal or restriction resulting from a terrifying eating-related incident (Katzman et al., 2022).

Findings from other areas of research show that children have unique information to provide that is crucial for the design and improvement of interventions. For example, Meltzer et al. (2013) reported that 40% of parents were not aware of child-reported sleep problems such as latency falling asleep, poor sleep quality, and night wakings – problems that occurred after children have been in bed. Children and adolescents enrolled in trials for cancer treatment reported experiencing more adverse events than clinicians detected and fewer adverse events than their parents reported (Freyer et al., 2022). In fact, discrepancies across adult and child reports of pediatric symptoms is common across domains of pain, mental health symptoms, and even observable behaviors such as sleep. In general, parents have been found to have more difficultly reporting on internal or internalizing symptoms (Grills & Ollendick, 2002), most likely because it is challenging to accurately detect the hidden internal experiences of another person. Parents have been reported to under-report pain (Chambers et al., 1998) – as when parents reported less pain in their 7–12 year old children than the children reported following surgery. Parents have also been shown to under-report emotional experiences such as anxiety, worry, fear, or depression in children ranging from 6–18 years old (Hansen & Jordan, 2020; Hughes & Gullone, 2010; Kashani et al., 1985; Kolko & Kazdin, 1993; Lagattuta et al., 2012; Sourander et al., 1999). There have been numerous hypotheses to explain discrepant reports: children may misunderstand what is being asked due to developmental constraints or may fail to disclose due to factors such as social desirability (Grills & Ollendick, 2002). Young children may have experiences that they do not even consider sharing due to a lack of awareness of the base rates for typical events, may have novel experiences that adults do not think to ask about, or may be communicating things in ways that adults fail to understand (e.g., as when an anxious child presents as argumentative). This research highlights the importance of developing ways to collect data on childhood experiences and to derive optimal methods to integrate these experiences with adult observations (De Los Reyes & Kazdin, 2004).

Developmentally tailored assessments may facilitate the characterization of FOAC in children with ARFID. For example, the Nine-Item Avoidant/Restrictive Food Intake Disorder Screen (NIAS) was designed as a self-report screening tool with items that assess eating restriction related to fear and was initially validated with adults ages 18–65 (Zickgraf & Ellis, 2018). The NIAS was subsequently validated for ages 10–76 (Murray et al., 2021). The Eating Disturbances in Youth-Questionnaire is a self-report assessment tool for screening eating disturbances in childhood with items that assess FOAC, which was tested in children 8–13 year olds (Kurz et al., 2015). The Fear of Food Questionnaire, a self-report questionnaire measuring fear and avoidance of food and eating associated with FOAC, was validated for adults ages 18 and above (Zickgraf et al., 2022). The Visceral Sensitivity Index is a self-report measure that specifically looks at gastrointestinal symptom anxiety that was initially validated for ages 19–79 (Labus et al., 2004) with a later version developed and tested for children, ages 8 to 12 (Lalouni et al., 2022). An interview of FOAC, the Pica, ARFID, and Rumination Disorder Interview (PARDI) assesses all relevant diagnostic criteria and symptom severity for ARFID, including measuring FOAC. The PARDI was initially validated for 10–22 year olds (Bryant-Waugh et al., 2019) and later for 9–23 year olds (Cooper-Vince et al., 2022). For children younger than 8 years old, different assessment strategies may be needed to elicit child experiences.

The aims of the present study are to 1) describe the child-reported experiences of FOAC using an interview designed specifically for young children; and 2) to examine the concordance of child fear ratings and parent-report of these fears. We hypothesized that the children in our sample would endorse a high frequency of fears of aversive consequences and that there would be poor correspondence between parent and child reports of child FOAC.

Method

Participants

Participants completed a two-stage screening. Children were required to be between 60 to 119 months old at the time of study screening, and children and parents were required to be fluent in English. In addition, the parent needed to endorse one or more of the following clinical features in the child: 3) a minimum score of 29 or above on the Child Food Neophobia Scale (the optimal sensitivity and specificity threshold scores were reported by Dovey et al. (2016) when employing this measure as a screening tool for ARFID); 4) underweight per parent report; 5) received a prior diagnosis of ARFID, feeding disorder, or failure to thrive; 6) a feeding tube due, in part, to an eating disorder; or 7) taking nutritional supplements to gain or maintain weight or correct nutritional deficiencies. Participants were excluded from the study if the child had an intellectual disability in the moderate, severe, or profound range based on their medical history and parent report. Eligible participants completed the remaining study procedures including a diagnostic interview to verify diagnosis (a diagnosis was not required for study participation). Children were permitted to have medical, psychiatric, and neurodevelopmental comorbidities provided that difficulties with eating were judged to be beyond those expected from comorbid diagnoses. Comorbidities in children with ARFID are common (Watts et al., 2023); thus, we deliberately included broad inclusion criteria. Given the limited work charactering ARFID in young children, it is imperative clarify the diagnostic boundaries among neurodevelopmental disorders, medical conditions, extreme childhood selective eating, and ARFID.

Participants were recruited through a variety of channels, including posts on parent social media groups dedicated to eating disorders or selective eating, participation at community festivals with local Hispanic and Black community organizations, and newsletter posts to listservs of parents of and providers for individuals with eating disorders.

Instruments

Screening

Child Food Neophobia Scale.

For eligibility screening, we used the six-item Child Food Neophobia Scale (CFNS; Pliner, 1994), a measure assessing the degree of food neophobia, or avoidance of new foods. Each item is answered on a 1 (Strongly Disagree) to 7 (Strongly Agree) Likert scale and the scale has a range of 10–70. The CFNS was previously examined as a screening tool for the possible presence of ARFID in children between the ages of 24 and 84 months. A score of ≥ 29 was recommended as the cutoff score and was employed in this study. This value was found to have sensitivity of 68% to detect ARFID cases and a specificity of 93% (Dovey et al., 2016). Cronbach’s alpha for the CRNS was .91 in our sample.

Characterization of ARFID Phenomenology and Diagnosis

Both the parent and child interviews contained questions about FOAC in ARFID. The Pica, ARFID, and Rumination Disorder Interview (PARDI; Bryant-Waugh et al., 2019) was administered for the parent interview, and the Gastrointestinal and Gustatory Avoidance Stress Scale Interview (GAGSS) was administered for the child interview (Gianneschi et al., 2024). Both the PARDI and GAGSS interviews were recorded.

Parent Interview.

Parents participated in the PARDI Interview (Versions 1.3 and 2.1; Bryant-Waugh et al., 2019), a semi-structured clinical caregiver interview used to assess the presence and severity of pica, ARFID, and rumination disorder in children. In addition to determining diagnosis, sections of the PARDI assess possible motivations for food avoidance such as FOAC, sensory sensitivity, and lack of interest in food. Examples of FOAC items include:

1. To your knowledge, over the past 4 weeks, has your child been concerned that eating will make him or her vomit (i.e., involuntarily) or cause diarrhea to the extent that it has restricted the amount or the type of food he/she eats?

2. To your knowledge, over the past 4 weeks, has your child been worried that eating might cause him or her pain (e.g., stomach pain) to the extent that it has restricted the amount or type of food they eat?

The interviewer rated items on a Likert scale from 0 = “never” to 6 = “always” with a rating of 2 having the anchor of “sometimes” and 4 with the anchor of “often.” Scores below a 4 indicated that the child may have a certain fear, but the interviewer judged that the fear was not meaningfully contributing to food avoidance. The PARDI algorithm (i.e. a combination of scores that make up the criteria of ARFID) was utilized in this study to determine ARFID diagnosis. Individuals who met one of the four components of criterion A (significant weight loss or failure to grow/gain weight, significant nutrition deficiency, dependence on enteral feeding or nutritional supplements, or psychosocial impairment) and also met criterion B (disturbance is not better explained by a lack of available food or cultural practice), C (diagnosis is not exclusively occurring with anorexia nervosa, bulimia nervosa, or other body weight or shape concerns), and D (the eating disturbance is not otherwise explained by another medical condition or psychiatric disorder) were diagnosed with ARFID. The PARDI was administered by individuals with a minimum of a master’s degree in clinical psychology.

To ensure the reliability and validity of ARFID diagnoses, interviewers met weekly with the principal investigator (Zucker) to go over interview questions that were difficult to rate. The investigative team initially had monthly meetings with one of the authors of the PARDI (Bryant-Waugh) to ensure appropriate implementation of the diagnostic interview. The PARDI was revised during the course of the study to better reflect the role of psychosocial impairment as an independent contributor to ARFID diagnosis as stipulated in the DSM-5-TR (American Psychiatric Association, 2022). Thus, all prior interviews in which a child did not receive an ARFID diagnosis were reviewed to ensure a diagnosis was not overlooked. Overall, 11.8% of diagnoses were changed to ARFID from no diagnosis given the change in diagnostic criteria. An additional 20% of randomly selected interviewed were reviewed, with no diagnoses changed. Our preliminary data support that the PARDI is a valid instrument for discriminating the clinical presence of ARFID in children (Cooper-Vince et al., 2022).

Body Mass Index.

Parents were asked to measure their child’s body weight and height before the PARDI interview to help the study team determine the body mass index, height, and weight percentiles. Families were mailed a scale and a tape measure and were provided instructional videos on obtaining the child’s height and weight so the measurements are standardized across individuals (see https://osf.io/bruh9/ for measurement materials). The Baylor College of Medicine Age-based Pediatric Growth Reference Charts website was used to determine BMI, the Center for Disease Control Height for Age Percentiles, and the Center for Disease Control Weight for Age Percentiles calculators were used to determine height and weight percentiles for each gender (WedMD, 2024, August 7; WedMD, 2024, August 8; WedMD, 2024, August 9; WedMD, 2024, August 10; Shypailo, 2020).

Child Interview

Children aged 84 to 119 months old were administered the Gastrointestinal and Gustatory Avoidance Stress Scale Interview (GAGSS; Gianneschi et al., 2024), an interview that was designed to understand the experience of children with ARFID as a clinical tool to gather data about the phenomenology of ARFID in young children for hypothesis generation and future intervention design. To ensure the appropriate use of responses obtained from the GAGSS interview, we constructed a validity argument in accordance with the recommendations of Kane (2013) and the Standards for Educational and Psychological Testing (American Educational Research Association et al., 2014). Kane (2013) advised that validity derives from the reasonableness of claims made from test scores (in this case, interview responses), in accordance with the intended use of those responses (see also, American Educational Research Association et al., 2014). To that end, our intended use of interview responses from children with ARFID or ARFID symptoms was to guide the development of treatment strategies. See Supplemental Methods for the full validity argument that guided the current use of responses from the GAGSS and suggested steps for further validation. To match the child’s level of development, interest, and ability to answer, we designed and described concrete scenarios that the child was likely to have encountered previously. We then probed how they would manage each scenario. Initial items for the GAGSS resulted from a focus group with eating disorder treatment providers who elicited common scenarios that children with ARFID had to manage and ways their patients managed those scenarios. The interview was then iteratively piloted with children and items modified resulting in ten iterations of the current interview as questions and scoring were modified to be more interpretable to children. Cognitive interviewing was conducted throughout all interview administrations. That is, in the limited sense that if children were struggling with an item, the interviewer probed whether the item made sense, the question was phrased more simply, and the incident was noted for further revisions to the interview. The interview was also piloted in two children below 84 months to examine the feasibility of assessing fears at younger ages, although this work is at a very preliminary stage. The current interview (Version 10.0) is available at https://osf.io/bruh9/.

The interviewer described different scenarios to the child and asked the child what they might do in each situation (e.g., going to a birthday party where they are serving a food the child has never had before). For each scenario, the child was allowed time to freely generate a response. After that open-ended question, the interviewer provided prompts of possible responses and the children were asked to agree, maybe agree, or disagree with those statements (Gianneschi et al., 2024).

To assess food- and eating-related fears, the interviewer asked the child whether he or she found it difficult to try a food that he or she has never tried before. Next, the interviewer asked the child about the difficulty of trying new foods from different categories (e.g., “Is it easy to try a new kind of chip?”). The interviewer then asked the child, “When you are thinking about trying a new food, do you do anything before you eat it to make sure the food is OK? If yes, what kinds of things do you do?” Finally, the interviewer asked the child, “Do you have any ideas of why it is hard to eat things?” After the free-response period, the interviewer then provided a list of potential aversive consequences such as “Afraid it will taste bad,” Figure 2. The interviewer coded the child’s responses as no, sometimes, or yes/most of the time.

Figure 2. Breakdown of Fears Endorsed by Individuals with ARFID or High Food Neophobia.

Note. Avoidant Restrictive Food Intake Disorder (ARFID) diagnoses were derived from The Pica, ARFID, and Rumination Disorder Interview (PARDI). Sub-Threshold Diagnostic Criteria were those participants who received a positive screen (i.e., Child Neophobia Score ≥ 29, prior diagnosis, underweight, and/or dependence on supplement or tube feeding to maintain or promote healthy growth) but did not meet the diagnostic threshold based on diagnostic interview. The length of the bar reflects the total number of participants that endorsed each fear. We further divided the bars by the proportion of children who met diagnostic criteria for ARFID that endorsed a fear and the children with high food neophobia who did not meet clinical threshold. In each group, the fractions reflect the proportion of that group. For example, 62.8% of children with ARFID were afraid the food will taste bad, and 64% of the children with high food neophobia were afraid the food will taste bad.

Child Behavior Checklist

Parents completed the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001) to assess behavioral, social, and emotional problems. This measure assesses numerous mental health symptoms including anxiety, depression, somatic complaints, and attentional difficulties, as well as providing a total score representing global problems. We utilized the parent-report version of the CBCL for school-age (6–18) children and the preschool version (CBCL/1.5–5) for parents of younger children (Achenbach & Rescorla, 2000; Achenbach & Rescorla, 2001). Parents reported their agreement with how much each item describes their child now or in the past 6 months on a three-point scale: 0 [Not True (as far as you know)], 1 (Somewhat or Sometimes True), and 2 (Very True or Often True). CBCL subscales have truncated T-scores with a mean of 50 and a standard deviation of 10, with different norms for males and females. T score transformation eliminates the bottom of the score distribution and scores that are at the mean or below are assigned a T-score of 50 (Thurber & Sheehan, 2012). The CBCL yields DSM-5 oriented subscales that help to identify broad emotional and behavioral problems that correspond to the DSM-5 diagnoses, as well as syndrome scales that are grouped into internalizing and externalizing problems. Cut-off scores provide ranges for normal, subclinical, and clinical levels of item endorsement. Cronbach’s alphas ranged from .50 (for Depressive Problems) to .87 (for Conduct Problems) across DSM-5 subscales in our sample.

Procedures

All study procedures were approved by the Institutional Review Board of Duke University. Parent and child interviews were part of a larger assessment battery that included self- and parent-report measures administered online and an in-home food approach task reported elsewhere (Kim et al., 2024). Study data were collected and managed using Microsoft Excel. Audio and video recordings were collected via an institutionally supported private Zoom^® virtual meeting platform. Participants scheduled a meeting with the study team to review and complete informed consent documents, first for screening and then, if the screen was positive, a second consent form was completed for further study participation.

Statistical Analysis

Data were summarized using descriptive statistics and evaluated for missing data and outlying values. Participants with incomplete or missing data were excluded from analysis (12.3% of the original total sample of 81 participants). Kappa statistics were used as index of parent-child agreement and were interpreted using Cohen’s suggested levels of agreement as follows: ≤ 0 = no agreement; 0.01 – 0.20 = none to slight agreement; 0.21 – 0.40 = fair agreement; 0.41 – 0.60 = moderate agreement; 0.61 – 0.80 = substantial agreement; 0.81 – 1.00 = almost perfect or perfect agreement (McHugh, 2012). Items pertaining to fears of abdominal pain and vomiting were most directly comparable across parent and child interviews and were used as the basis for comparison. To assess the concordance of child and parent reports of FOAC, the PARDI interview was recoded from its original 0–6 scale to a 0–2 scale (as used on the GAGSS interview) so that the data could be directly compared. PARDI interview data were recoded as follows:

• 0 “never” on the PARDI = 0 “no” on the GAGSS;

• 1, 2 “sometimes,” or 3 on the PARDI = 1 “sometimes” on the GAGSS;

• 4 “often,” 5, or 6 “always” on the PARDI = 2 “yes” on the GAGSS.

Analyses were run using IBM SPSS Statistics^® (Version 28) as well as Microsoft Office Excel (2021).

Results

Participants

Figure 1 presents the recruitment and retention of study participants. A total of 81 GAGSS interviews were conducted. Of these 81 children, two were excluded due to failure to complete the GAGSS interview. Ten children were excluded due to missing data (see Figure 1). One parent did not complete the FOAC section of the PARDI interview and was excluded from analysis, leaving a final sample of 68 parent-child pairs who completed both the GAGSS and the PARDI fear-related questions. See Table 1 for a demographic description of the sample. Overall, the sample was 58.5% cisgender male and 41.2% cisgender female with an average age of 8.18 (SD = 1.06) years, with ages ranging from 5.2 to 10.0 years. Sixty-three percent of the sample met criteria for ARFID with 36.8% screening positively for possible ARFID (average score of 40.4 (STD = 2.06) on the CFNS relative to the cut-off score of 29), but failing to meet the threshold for impairment required for diagnosis.

Figure 1. CONSORT Flowchart of Participants.

Note. *Participants were administered an older version of the GAGGS interview, which had a reduced version of the FOAC Section

** Participants were administered the full version of the FOAC section in the current GAGGS interview. Two children did not answer one question, one child did not answer 3 questions, and one child was not administered the FOAC section.

*** This participant completed the assessments in a different order and did not complete a PARDI Interview, so they are not included as part of “n=179” resulting in a discrepancy on 1 in this consort flow

Table 1.

Participant Characteristics and Descriptive Statistics

Participant Characteristics			n	%
Sex	Male		40	58.8%
	Female		28	41.2%
Race	White		58	85.3%
	Black/African American		3	4.4%
	More than 1 Race		3	4.4%
	Other		2	2.9%
	Asian		1	1.5%
	Prefer Not to Say		1	1.5%
Ethnicity	Not Hispanic/Latino		63	92.6%
	Hispanic/Latino		4	5.9%
	Prefer Not to Say		1	1.4%
Age in Years	M	SD	Minimum	Maximum
	8.18	1.06	5.17	9.96
ARFID Diagnoses from the PARDI Interview*			n	%
Participants Meeting ARFID Diagnostic Criteria			43	63.2%
Participants Meeting Sub-Threshold ARFID Criteria**			25	36.8%
BMI Centiles of Participants#			n	%
BMI Centile < 0.5%			3	4.4%
0.5% ≥ BMI Centile < 5%			6	8.8%
5% ≥ BMI Centile < 85%			48	70.6%
85% ≥ BMI Centile > 95%			5	7.4%
BMI Centile ≥ 95%			6	8.8%

Note:

^*Avoidant Restrictive Food Intake Disorder (ARFID) diagnoses were derived from The Pica, ARFID, and Rumination Disorder Interview (PARDI).

^**Sub-Threshold Diagnostic Criteria were those participants who received a positive screen (i.e., Child Neophobia Score ≥ 29, prior diagnosis, underweight, and/or dependence on supplement or tube feeding to maintain or promote healthy growth) but did not meet the diagnostic threshold based on diagnostic interview.

^#Body Mass Index (BMI) Centiles were calculated from home-based measurements using scales and instructional videos mailed to participants.

Table 2 summarizes the range of normal, subclinical, and clinical percentages across the syndrome and DSM oriented scales on the CBCL (Achenbach & Rescorla, 2000; Achenbach & Rescorla, 2001). The most frequent clinical problem was anxiety, with 24.4% of the sample (9 participants) meeting the clinical threshold for anxiety problems on the DSM-5 scale of the CBCL; 35.3% of the sample had at least one clinical scale elevation on the CBCL. See Supplemental Table 1 for additional information about comorbid psychopathology obtained via parent interview and Supplemental Table 2 for information about the child’s executive functioning per parent report.

Table 2.

Participant CBCL DSM-5 Oriented Scales Assessment Scores

	Depressive Problems	Anxiety Problems	Somatic Problems*	ADHD Problems	Oppositional Defiant Problems	Conduct Problems*
	n = 65	n = 66	n = 64	n = 65	n = 66	n = 65
Normal	38 (58.5%)	35 (53.0%)	49 (76.6%)	53 (81.5%)	57 (86.4%)	56 (86.2%)
Subclinical	16 (24.6%)	15 (22.7%)	6 (9.4%)	9 (13.8%)	4 (6.1%)	5 (7.7%)
Clinical	11 (16.9%)	16 (24.2%)	9 (14.1)	3 (4.6%)	5 (7.6%)	4 (6.2%)

Note:

^*Included only in the CBCL Age 6 and older. Classifications are based on normative score cut-offs provided in the CBCL manual (reference).

Description of Child Fears

Based on the GAGSS interview, child participants endorsed an average of 4.3 fears (SD = 2.7) out of a possible 9. The most frequently endorsed fear was “afraid [the food] will taste bad,” with 63.2% of children (n=43) indicating that it was a reason behind their food avoidance or restriction (Figure 2). The second and third most endorsed fears were feeling “afraid [the food] will make you gag” with 54.4 % child endorsements (n=37) and “[thinking that] the food looks disgusting” with 52.9% child endorsements (n=36). See Figure 2 for the complete breakdown of fears from most to least endorsed by participants. We also provided these frequencies based on whether the child had an ARFID diagnosis or elevated food neophobia alone without impairment (i.e., not meeting a clinical threshold for diagnosis).

Comparison of Parent and Child Fear Ratings

Of the 67 parents who completed the FOAC section of the PARDI, 26.5% of parents (n=18) endorsed clinically significant FOAC for their child. Contrary to this finding, 91.2% (n=62) of the children in our sample with ARFID symptoms endorsed at least one fear of aversive consequences in relation to their eating. As for the specific fear questions about stomach pain and vomiting, children endorsed these fears more commonly than their parents did. Twenty-seven percent of children endorsed “yes” for the fear of stomach pain (n=18) compared to only 1.5% of parents (n=1) recognizing such. As for fear of vomiting, 39.7% of children endorsed “yes” for this fear (n=27), while 10.3% of their parents did not (n=7).

In exploratory analyses, a Kappa coefficient of agreement showed none to slight reliability between parent and child ratings of child fears of stomach pain (κ = .12) and vomiting (κ = .08) when judged relative to Cohen’s benchmarks for reliability (McHugh, 2012), see Table 3. Of the 25 parent-child dyads who disagreed on the child’s fear of stomach pain, in 96.0% (n=24) of the cases, the child reported a higher frequency of fear of stomach pain than did the parent. Further, in 48.0% of these 25 cases (n=12), the child reported clinically significant levels of fear while the parents reported no fear of stomach pain. Regarding the child’s fear of vomiting, 33 parent-child pairs disagreed on the frequency of the child’s fear. In 87.9% of these 33 dyads (n=29), the child reported a higher frequency of a fear of vomiting than the parent. Additionally, in 57.8% of these 33 cases (n=19), parents did not report any fear of vomiting while the child reported a clinically significant fear of vomiting.

Table 3.

Parent-Child Fear Rating Congruence from PARDI and GAGGS Interviews

Fears	Congruent Ratings	Incongruent Ratings	Kappa
	% (n)	% (n)	(0.00 – 1.00)
Stomach Pain	63.2% (43)	36.8% (25)	κ = 0.12 (SE = 0.06)
Vomiting	51.5% (35)	48.5% (33)	κ = 0.08 (SE = 0.08)

Note: Cohen’s suggested levels of agreement as follows: ≤ 0 = no agreement; 0.01 – 0.20 = none to slight agreement; 0.21 – 0.40 = fair agreement; 0.41 – 0.60 = moderate agreement; 0.61 – 0.80 = substantial agreement; 0.81 – 1.00 = almost perfect or perfect agreement

Discussion

The current study utilized parent and child reports of fears of aversive consequences related to food consumption (FOAC) to characterize the experiences of children with ARFID and examine the agreement between child and parent reports. Findings indicated that children may experience a greater number and intensity of feared consequences from food and eating than their parents realize. Nearly the entire sample of children (91.2%) endorsed at least one fear in relation to food or eating relative to 26.5% of parents who endorsed clinically significant fear as a driver of food avoidance in their children. In parent-child dyads that disagreed on child fear ratings, children indicated a higher frequency of their fears in the majority of cases (e.g., 96.0% for discrepant reports of stomach pain), which bolsters previous findings that suggest parents may underestimate their child’s internal pain or distress (Chambers et al., 1998; Lagattuta et al., 2012; Pinheiro et al., 2018).

One explanation for the discrepancies is that the way anxiety may manifest in children, such as behavioral rigidity or refusal to engage in a food or activity, may evoke parent hypotheses of noncompliance or oppositional behavior when, in fact, such avoidance is rooted in fear. These results are consistent with findings from a preschool cohort study of young children (ages 24 to 71 months) categorized as selective eaters via parent diagnostic interview (Zucker et al., 2015). Children who were severe selective eaters had approximately twice the likelihood of social anxiety or depressive disorders than typically developing children but not a greater likelihood of conduct disorder or oppositional defiant disorder. Helping parents explore the role of emotional experience in their child’s food avoidance may help to facilitate better parent-child collaboration in food approach.

The nature of children’s fear also appears to differ from that of older children with ARFID. The most commonly reported fears endorsed by children in a scenario-based interview designed for children, the GAGSS interview, were “afraid [the food] will taste bad” and “afraid [the food] will make [me] gag,” with over half of participants endorsing each of these fears. Commonly referenced FOACs in older child, adolescent, and the adult ARFID literature include gastrointestinal distress (e.g., nausea or vomiting) or choking (Bryant-Waugh et al., 2019; Fink et al., 2022; Murray et al., 2020). The children in our sample endorsed these fears with much less frequency – vomiting and stomach pain were the least frequently endorsed fears. Children’s fears typically change as they age, with younger children expressing more fears of the imaginary and older children expressing more realistic fears (Bauer, 1976), like fear of negative social evaluation (Westenberg et al., 2007). However, another factor may be the timing of fears in relation to eating and the child’s capacity to predict uncertain future events. The fears endorsed most frequently by the children in this study occur immediately: the taste of a food and gagging from a food are immediate experiences from food consumption. In contrast, abdominal pain or nausea in response to food, fears most commonly reported in adolescents or older children, may occur up to hours after a meal. These findings are consistent with studies demonstrating that the capacity to predict future states improves with age. A study by Atance and Meltzoff (2005) found that children at age 5 were able to explain and make predictions of future states much better than 3- or 4-year-olds. The 5-year-olds in the study were also able to make more references to future states than the 3-year-olds, suggesting that as children age, their capacity to explain a future feeling or state may increase (Atance & Meltzoff, 2005). A 2016 review examining children’s capacity to report past and future pain states concluded that recognition of the self and greater representational theory of mind, often developed at ages 5 years and older, impacts children’s ability to anticipate future pain (Jaaniste et al., 2016). These results support previous data in older children with ARFID, suggesting that FOAC may be a key motivation behind food avoidance/restriction in children who exhibit ARFID symptoms (Katzman et al., 2019; Murray et al., 2020).

The “fears” most frequently identified in our sample were those that can also be considered somatosensory or anticipatory disgust responses. Disgust is a primary emotion that is considered one of the first emotions to emerge in a child’s development (Rozin & Fallon, 1987). From a functionalist perspective of emotional experience, disgust is hypothesized to help an individual avoid exposure to pathogens that increase vulnerability for future disease/illness (e.g.,Olatunji & Cisler, 2009), and often is accompanied by nausea (e.g. Rozin et al., 2009), associated negative affect (e.g. Angyal 1941), and a visceral reaction (e.g. Verstaen et al., 2016). Distinguishing between fear and disgust is challenging due to the similarities in the outward behavioral response to these emotions (e.g., avoidance). These emotions are hypothesized to differ in function such that disgust functions to avoid disease or contamination and fear functions to protect against imminent danger. Fear can also co-occur with disgust (see Stevenson et al. (2019)’s model of disgust). Research has suggested that fear and disgust react differently to intervention; disgust is typically more resistant to extinction such that more presentations of a disgusting item do not meaningfully impact the degree of disgust experienced (Mason & Richardson, 2012), whereas the most validated treatment for fear is repeated exposure (e.g., anxiety; Kaczkurin & Foa, 2015). Thus, future research aimed at deciphering the fears of young children with ARFID and identifying those that illicit disgust, fear, or both will help tailor interventions for clinically significant food avoidance.

Consistent with that intent, this study was also motivated by our clinical experiences. Not only did clinicians, parents, and children have difficulty identifying or articulating the fears of young children, but there was greater collaboration and cooperation when these fears were identified. Child behaviors are often used as a proxy for self-reports of fear. For example, in Zucker et al. (2010), a workgroup recommended changes to the diagnostic criteria for anorexia nervosa so that behavioral indices could serves as substitutes for reported fears of weight gain to aid in the earlier identification of anorexia nervosa in younger individuals. It is thus interesting to think about the behaviors of fear responding – fighting, fleeing, or freezing. When a child is being insistent and demanding, rather than trembling and shaking, an adult’s initial hypothesis may not be fear – but striving for autonomy, oppositional, etc. Thus, finding ways to learn about childhood motivations may help us define treatment targets and help us adults understand childhood experiences. It may also highlight novel fears. In ARFID, we largely focus on gastrointestinal issues, vomiting, or choking. But a child recently reported that he was afraid his teeth were not strong enough and would break if he ate the food, highlighting novel perceptions of children and the need to further develop our interview to ensure we are not missing relevant childhood fears. Thus, probing discrepancies can highlight areas parents notice that the child may not be aware of or have insight into, can elicit novel insights by the child, and can emphasize areas in need of better adult and child communication.

The present study has several strengths, including the use of an interview that was designed specifically for younger children. At face value, the content of the fears section of the GAGSS interview overlapped with the content of the fears section in the validated PARDI interview, allowing for comparisons between parent and child fear ratings. Additionally, our sample was composed of children under 10-years-old, an under-studied population as it relates to ARFID research. Finally, this study is novel in its analysis of parent and child agreement regarding FOAC in an ARFID population.

Although this study presents novel findings regarding parent and child concordance of anxiety ratings as they relate to ARFID, we faced several limitations. The GAGSS interview is still in its preliminary stages of development and is currently not designed as a diagnostic tool. Depending on the intended future use of this interview, and the important clinical decisions that may be made based on interview data, particular validation strategies would be needed and the integration of parent and child data to inform those decisions would be essential. Notably, De Los Reyes and Kazdin (2004) evaluated different methods for measuring child (ages 6 −16) and parent discrepancies and concluded that given the different information provided by each method, no one informant should be seen as the sole standard for important clinical decisions. In its current intended use as a tool to better understand the experience of ARFID and to develop tailored intervention strategies, additional validation strategies are also needed. Possible strategies for validation include the use of cognitive interviews with children to elucidate the strategies children use to answer interview questions. Comparisons of personalized treatments guided by interview responses relative to standard treatments would provide a further index of criterion validity. A parent version of the interview would help us to understand the situations in which children had more or less capacity to cope than parents understand. Work exploring the comprehensiveness of items in assessing the full domain space of fearful situations children encounter will be needed to establish content validity.

It also should be noted that the semi-structured GAGSS interview did not assess the children’s fears of choking. This was intentional as we did not want to suggest a fear that the child did not currently endorse, but future research should probe the wisdom of that decision. Further, our sample size was modest with 68 participants, some of whom did not meet full ARFID diagnostic criteria. A larger sample size is needed to further evaluate the generalizability of these findings. We also note that discrepancies between parent and child report may be due to differences in how fears were assessed – with parents given more nuanced options and the interviewer making judgements of yes, no, or maybe to child responses as to whether a given fear contributed to their food avoidance. Although we attempted to address these measurement differences by re-coding responses, measurement differences should be considered when evaluating our findings.

The diagnosis of ARFID is relatively new and information about the presentation of ARFID in young children is emerging. Greater precision in delineating eating challenges that can be ascribed to various neurodevelopmental, psychiatric, and medical disorders and treated within the bounds of existing interventions or warranting an ARFID diagnosis with tailored invention will help streamline our future treatment efforts. These considerations motivated the use of broad inclusion criteria, but differences among children and parent-child correspondence may vary according to the presence of food neophobia or comorbidity. Finally we note that the current sample is predominately white and non-Hispanic/Latino, which may limit generalizability. Future research with larger sample sizes that better reflect current population demographics is needed to address these issues.

Conclusion

We have much to learn from the perspectives of young children, insights that can help us better understand ARFID and design more effective interventions. Children—including those with ARFID—may express fears in ways that are not easily perceptible by adults (Katzman et al., 2019). The present study indicated that children and their parents exhibit poor concordance when it comes to reporting the experience and frequency of fears, with children reporting more fears than parents detected. The characterization of fear expression in young children with ARFID may enhance our ability to understand, detect, and treat ARFID in its earliest manifestations.

Public Significance Statement.

We have much to learn from the perspectives of young children, insights that can help us better understand ARFID and design more effective interventions. Children—including those with ARFID—may express fears in ways that are not easily perceptible by adults (Katzman et al., 2019). The present study indicated that children and their parents exhibit poor concordance when it comes to reporting the experience and frequency of fears, with children reporting more fears than parents detected. The characterization of fear expression in young children with ARFID may aid our ability to understand, detect, and treat ARFID in its earliest manifestations.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. The GAGSS interview can be accessed at https://osf.io/bruh9/.