Food selectivity, nutritional inadequacies, and mealtime behavioral problems in children with autism spectrum disorder compared to neurotypical children.

OBJECTIVE
To evaluate body composition, nutritional status through food selectivity and degree of inadequate intake, and mealtime behavior in children with autism spectrum disorder (ASD) compared to neurotypical children.


METHOD
A cross-sectional case-control study was carried out in 144 children (N = 55 with ASD; N = 91 with neurotypical children) between 6 and 18 years of age. Body composition, nutritional intake, food consumption frequency (FFQ), and mealtime behavior were evaluated.


RESULTS
Results showed a greater presence of children with a low weight (18.4% ASD vs. 3.20% comparison group) and obesity (16.3% ASD vs. 8.6% comparison group) in the ASD group for body mass index (BMI) categories (p = .003; number needed to take [NNT] = 8.07). The presence of obesity in ASD children compared to the comparison group was even higher when considering the fat component (47.5% ASD vs. 19.4% comparison group, p = .002; NNT = 10.3). ASD children had greater intake inadequacy (50% ASD vs. 22% comparison group, p = .014; NNT = 3.58), high food selectivity by FFQ (60.6% ASD vs. 37.9% comparison group, p < .037; NNT = 4.41), and more eating problems (food rejection, limited variety, disruptive behavior), compared to neurotypical children (p = .001).


CONCLUSION
Children with ASD showed an unbalanced body composition toward both underweight and obesity, a greater degree of inadequate intake, high food selectivity as indicated by their consumption frequency, and more disturbed eating behavior than children with neurotypical development. We suggest monitoring nutritional inadequacies and implementing nutritional strategies to expand the variety of foods children with ASD consume.


| INTRODUCTION
Autism spectrum disorders (ASD) is a behaviorally defined complex neurodevelopmental syndrome characterized by impairments in social communication and restricted and repetitive behaviors, interests and activities, as well as abnormalities in sensory reactivity (American Psychiatric Association, 2013) that are frequently associated with different comorbidities such as disruptive behavior, gastrointestinal symptoms, and eating problems (Peverill et al., 2019). These atypical eating patterns are often governed by food rejection/denial or a preference for certain kinds of food (Bandini et al., 2017;Wentz, Björk, & Dahlgren, 2019), and may lie in a sensory-associated physiological alterations probably caused by their behavioral problems (Ashley, Steinfeld, Young, & Ozonoff, 2020;Peverill et al., 2019). ASD children's atypical eating patterns include mealtime rituals (Hubbard, Anderson, Curtin, Must, & Bandini, 2014), food selectivity (Bandini et al., 2017;Sharp et al., 2018), and disruptive behavior at mealtimes (Dovey et al., 2019;Leader, Tuohy, Chen, Mannion, & Gilroy, 2020;Murphy, Zlomke, VanOrmer, & Swingle, 2020). Food selectivity is one of the most common eating problems in children with ASD and is an important cause of concern because of its negative impact upon nutritional adequacy (Kral et al., 2015;Sharp et al., 2013) and anthropometric parameters Sharp et al., 2018), as well as the family stress generated at mealtime (Thullen & Bonsall, 2017). Severe forms of feeding concerns in ASD meet the diagnostic criteria for avoidant-restrictive food intake disorder (ARFID), a psychiatric diagnosis that recognizes ASD as a risk factor for nutritional concerns associated with food selectivity (Eddy & Thomas, 2019).
Previous review  and metaanalysis studies (Sharp et al., 2013) concluded a higher food selectivity and nutrient insufficiency in ASD children compared to control children. Nevertheless, they point out a lack of complete nutritional assessment in ASD child. Nowadays, there is still a need of analytical studies that cover all aspects involved in nutritional intake in children with ASD, including food intake frequency by groups, as well as the number of deficiencies and the dietary preferences in ASD that may explain conflicting results among past reports. To date, a few studies  have assessed in detail nutrient intake in children with ASD, based on 72-h food diaries describing insufficient intake of certain nutrients such as vitamins A, C, B6, B9 (folate), B12, D, E, and K, or minerals such as phosphorus, zinc, calcium (Ca), or iron (Fe). Very few studies (Marí-Bauset, Llopis-González, Zazpe, Marí-Sanchis, & Morales Suárez-Varela, 2017) have additionally analyzed food consumption habits through the application of the food frequency questionnaire (FFQ) to identify unbalanced diets due to excess or absence in terms of the consumption of certain food groups. It has been suggested that children with ASD could be at risk of suffering nutritional inadequacies  resulting in negative middle-to long-term consequences for growth and development (Bandini et al., 2017;Graf-Myles et al., 2013). Moreover, the relative contribution of food selectivity and dietary disturbance to the increased risk of overweight and obesity that has been observed in ASD is still a prominent area of research, based on dietary preferences for processed foods, snacks, and sweets and the frequent rejection of fruits and vegetables (Sharp et al., 2013). Since the literature on the assessment of ASD children's nutritional status is limited, to our knowledge, there are few studies Marí-Bauset et al., 2017) that have performed a detailed comprehensive assessment of intake using the reference technique of 72-h food diaries, complemented by food consumption frequency.
The existence of differences in anthropometric measurements between children with ASD and neurotypical children has been described as possibly related to severe food selectivity and mealtime behavioral problems, as well as to special diets and/or gastrointestinal problems in children with ASD (Kamal Nor, Ghozali, & Ismail, 2019).
Although there are different methods by which to assess overweight and obesity in children and adolescents (Chan & Woo, 2010), body composition methods demonstrate high accuracy in the assessment of adiposity (Jensky-Squires et al., 2008). To date, very few studies have evaluated body composition in children with ASD, and the existing research has produced inconsistent or even contradictory results. On the one hand, Castro et al. (2017) concluded that a large percentage of children and adolescents with ASD had total overweight or obesity with truncal adiposity, and there was a significant percentage of underweight participants. In contrast, a recent study (Esteban-Figuerola, Morales-Hidalgo, Arija-Val, & Canals-Sans, 2021) revealed the absence of significant differences in bioelectrical impedance analysis in children and adolescents with ASD, compared to children with neurotypical development. Therefore, further studies are needed to analyze body composition imbalances in this concrete population, with a view to associating such alterations with atypical behavioral patterns and the intake inadequacy.
Based on the above, the present study was carried out to evaluate body composition, nutritional status through food selectivity and degree of inadequate intake, as well as mealtime behavior in children with ASD, compared to neurotypical children. We hypothesized that children with ASD would (a) present an unbalanced body composition, (b) have altered macro-and micronutrient food intake, (c) have food selectivity as indicated by their consumption frequency, and (d) exhibit more disruptive behavior than children with neurotypical development.

| Participants
The present cross-sectional, observational case-control study was carried out at Madrid Complutense University and the University of Granada (Spain) between January 2016 and December 2018. This was a multidisciplinary study involving specialists in dentistry, nutrition, physiology, and ASD. The study sample comprised 144 children (51 children with ASD and 93 neurotypical children) and convenience sampling was used to select participants attending special schools and primary schools. The neurotypical children were selected from the same inclusive public centers that have specific classrooms for children with ASD. ASD children aged between 6 and 18 years were diagnosed based on the criteria in the Diagnostic and Statistical Manual of Mental Disorders V (DSM-V), but disorder severity was not considered. The comparison group members were selected from a geographical setting and socioeconomic context similar to that of the children with ASD. The exclusion criteria for both groups were as follows: use of dietary supplements, receiving treatment or following a special diet (e.g., gluten-and casein-free), any medical condition (endocrine, metabolic diseases, etc.) that could influence food intake, use of drugs (e.g., stimulants, atypical antipsychotics, tricyclic antidepressants, steroids, and mood stabilizers) that could affect food intake; not attending study interview appointments, and not completing nutritional records properly.
After contacting the school administration, the center notified the parents and coordinated an informative meeting. After the initial session, those parents who decided to allow their children to participate were given an informed consent form to sign. At the first appointment, before the study began, the parents received an explanation of the tests that would be applied, as well as data confidentiality in compliance with Spanish legislation, and informed consent for children's participation was obtained. The study was approved by the Ethics Committee of Hospital Clínico Universitario San Carlos (Madrid, Spain). The children's parents or legal representatives were scheduled to attend an informative meeting during which the study's objectives, the relevance of the expected results, and the study's significance to the children's health were explained. For both groups, we excluded families who declined to participate and those who we were unable or failed to provide complete nutritional records.

| Body composition and anthropometric measurements
Height was measured using a stadiometer (precision 0.1 cm, Secca The impedance between each child's feet and hands, respectively, was measured while an alternating current (50 kHz, 90 mA) was passed through the entire body. The participants or their parents/legal representatives were informed in advance about the conditions that had to be observed prior to measurement: no vigorous exercise for at least 12 h before the test, no food or drink for at least 3 h before the test, and urination immediately before measurement. All anthropometric measurements were taken at the same time and recorded by the same researcher. The following measurements were obtained: weight, body mass index (BMI), and lean mass and fat mass expressed as a percentage of total body fat (fat mass index [FMI]).

| Seventy-two-h food diary
A 72-h recall questionnaire was used to record the participant's food and drink consumption over a 3-day period. This type of 3-day food diary is currently considered the gold standard among the methods for assessing diet (Barrett-Connor, 1991). The questionnaire was based on recall of food intake for 2 days of the week, in addition to one weekend day, in order to avoid weekly mean data registry bias attributable to special consumption habits on weekends. The food intake information was obtained from individual interviews with the children's parents or legal representatives, conducted by a registered dietitian. Experts on the research team held informative workshops before data collection and resolved any possible doubts through telephone support. The data collected on food and beverage consumption were finally converted to absolute values referring to energy consumption, macro and micronutrient intake, and percentage intake adequacy for each nutrient (recommended daily allowance [RDA]), using Nutriber software (version 1.1.5, Barcelona, Spain).
Nutritional data obtained were compared against the Spanish food composition tables (Mataix, 2011). Vitamin and mineral intake was also compared against the recommended dietary intake (DRI) for both the Spanish population and the European Union (Cuervo et al., 2009). Intake adequacy or inadequacy was determined by comparing each participant's real intake to the recommended intake for each nutrient, with two cut-off points: insufficient intake (less than 75% of the DRI) and excessive intake (above the DRI). In addition, to assess the children's specific degree of inadequate intake, we calculated the total number of macronutrients and micronutrients for which there was insufficient intake and established three cut-off points: mildly inadequate intake (insufficient intake of one to five nutrients), moderately inadequate intake (insufficient intake of 6 to 10 nutrients), and highly inadequate intake (insufficient intake of over 10 nutrients) (Marí-Bauset, Llopis-González, Zazpe-García, Marí-Sanchis, & Morales-Suárez-Varela, 2015).

| Food frequency questionnaire
The FFQ was used to determine the daily, monthly, and annual consumption frequency of each food groups, based on a list of over 200 foods. For each participant, we determined the percentage adequacy of the consumption for the different food groups in relation to the recommendations of the Spanish Society of Community Nutrition (SENC) (Aranceta Bartrina et al., 2016). The participants were classified into two categories: (a) an FFQ score below the recommended servings of a given food group, or (b) an FFQ score above the recommended servings based on the food pyramid proposed by the SENC for each food group (Aranceta Bartrina et al., 2016). Those food groups without a recommendation were expressed as servings consumed per week. Food selectivity was assessed using the information reported on the FFQ, defined as the degree of adequacy, and used to classify the participants into two categories according to the percentage of foods each child refused among those offered, that is, showing no intake frequency: (a) nonfood selectivity (an FFQ food score of less than 33% of the total foods); or (b) high food selectivity (an FFQ score of over 33% of the total foods) (Chistol et al., 2018;Curtin et al., 2015).

| Brief assessment of mealtime behavior in children
Mealtime behavior was evaluated using the BAMBIC (Hendy, Seiverling, Lukens, & Williams, 2013). The questionnaire was administered to participants' parents to assess three common food problems: (a) limited variety of foods, (b) rejection of foods, and (c) disruptive behavior at mealtimes. The questionnaire comprised a total of 10 items, with scores for each BAMBIC subscale based on a 5-point Likert scale. Each item's scores were initially multiplied by the factor loadings (Hendy et al., 2013) and, subsequently, the mean scores for each dimension were calculated. Regarding alpha coefficients, 0.79 was found for the limited variety factor, 0.69 for the food refusal factor, and 0.69 for the disruptive behavior factor. The parents or legal representatives were asked to score the children's mealtime eating behavior according to the frequency of incidence of each situation in the last 6 months on a scale ranging from 1 to 5, where 1 = never, 2 = rarely, 3 = occasionally, 4 = often, 5 = almost always.

| Statistical analysis
The data were analyzed using the SPSS version 25.0 statistical pack-   (Lenhard & Lenhard, 2017). Number needed to take was also calculated for categorical variables as the number of children with altered body compartment relative to the total number of children for each group (NTT: < 4: strong effect; 4 to 9: moderate effect; > 9 weak effect) (Kraemer & Kupfer, 2006). The body mass index (BMI) was classified into four categories in accordance to the CDC growth charts for the BMI by age (for boys or girls) to obtain the percentile category: underweight, less than the 5th percentile; healthy weight, 5th percentile to 85th percentile; overweight, 85th to 95th percentile; obese, equal to or greater than the 95th percentile (Nagy et al., 2014). The fat mass index was classified into three categories in accordance to the TANITA healthy body fat ranges for children by age (for boys or girls): healthy, overfat and obese (McCarthy et al. 2006 (Lenhard & Lenhard, 2017), being calculated from the Z of the applied nonparametric test (Fritz et al., 2012). Number needed to take was also calculated for categorical variables (NTT: < 4: strong effect; 4 to 9: moderate effect; > 9 weak effect) (Kraemer & Kupfer, 2006). The DRI percentages were calculated according to the Spanish recommendations (Cuervo et al., 2009). The macronutrient distribution was expressed as percentage of total energy consumed. Intake inadequacy was determined by comparing each participant's real intake to the recommended intake for each nutrient, with two cut-off points: an intake below the 75% of DRI was classified as "insufficient intake"; an intake above the 100% of DRI was classified as "intake excess." Abbreviations: ASD, autism spectrum disorder; DRI, dietary reference intake.
*p-value obtained in Chi-square test using the Bonferroni adjustment (p < .05).
T A B L E 3 Food consumption frequency and adequacy in children with autism spectrum disorder and neurotypical children  (Lenhard & Lenhard, 2017), being calculated from the Z of the applied nonparametric test (Fritz et al., 2012). Number needed to take was also calculated for categorical variables (NTT: < 4: strong effect; 4 to 9: moderate effect; > 9 weak effect) (Kraemer & Kupfer, 2006 The continuous variables data were expressed as the mean and SD. The categorical variables were expressed number and percentage of subjects. ANCOVA analysis was used to compare values for ASD and the comparison group. The model has been corrected by the age factor. Bonferroni-corrected significance levels for multiple comparisons were calculated for nutritional inadequacy (p = .05/4, being considered significant if p < .01) and the frequency of problematic child mealtime behaviors (p = .05/3, being considered significant if p < .02). The BAMBIC classified the scores into three categories: food rejection, limited food variety, and disruptive behavior (Hendy et al., 2013). Cohen's d coefficient for continuous variables were provided (.20: small effect; .50: intermediate effect; .80 and higher: large effect) (Lenhard & Lenhard, 2017). Number needed to take was also calculated for categorical variables (NTT: < 4: strong effect; 4 to 9: moderate effect; > 9 weak effect) (Kraemer & Kupfer, 2006). For the categorical variable degree of inadequate nutrient intake, low and moderate inadequacy computed as unfortunate outcome. The nutritional inadequacy was classified into three categories according to the range of nutrients with inadequate intake (Mari-Bauset et al., 2015): low, inadequate intake of 1 to 5 nutrients; moderate, inadequate intake of 6 to 10 nutrients; high, inadequate intake greater than 10 nutrients. Food selectivity was assessed using the information reported on the FFQ, defined as the degree of adequacy, and used to classify the participants into two categories according to the percentage of foods each child refused among those offered, that is, showing no intake frequency: nonselectivity, an FFQ food score of less than 33% of the total foods; high food selectivity, an FFQ score of over 33% of the total foods (Chistol et al., 2018;Curtin et al., 2015). a p-Value obtained in the Chi-square test using Bonferroni's analysis. χ 2 = 8.491, df = 3, p < .05. b p-Value obtained in the Chi-square test using Bonferroni's analysis. χ 2 = 4.351, df = 1, p < .05.
'failure' (e.g., BMI disturbance, nutritional inadequacy, or food selectivity) than if you had sampled neurotypical children" (Kraemer & Kupfer, 2006;Striegel-Moore et al., 2009). Table 1 reports the characteristics of the ASD and neurotypical children included in the study. Statistically significant age differences were observed; specifically, the children with ASD were younger than the comparison group. The group comparison of the mean z score for ageadjusted BMI was not statistically significant. However, the sample distribution analysis revealed significant between-group differences with a moderate effect for BMI categories; specifically, comparatively more children with ASD were in the low weight category (18.4% vs. 3.20% in the comparison group) and also in the obese category (16.3% vs. 8.6% in the comparison group). Significantly more children with ASD were also found to be in the obese category when stratifying by fat percentage (47.5%

| RESULTS
vs. 19.4% in the comparison group; small effect). Table 2 shows the results of the comparative analysis referred to median energy, macro-and micronutrient intake, and intake adequacy in the study groups. Despite the fact that the median energy was simi- vs. 22.0% in the comparison group) (small to moderate effect). Table 3 in turn shows the adequacy of mean food consumption frequency in the study population. After adjusting for multiple comparisons, no differences were observed between groups in the consumption of any of the analyzed food groups-though there was a trend toward statistical significance for the consumption of vegetables (p = 0.009) and sweets, snacks, and soft drinks (p = 0.006). Most of the children in both groups exceeded the recommended consumption of occasional foods such as sweets, snacks, and soft drinks, while they did not frequently consume recommended foods such as potatoes, rice, bread, wholemeal bread, pasta, fruit, or nuts. In addition, more children in the neurotypical comparison group showed a lower frequency of vegetable consumption than in the ASD group.

| DISCUSSION
The present study compared body composition, nutritional status, and mealtime eating behaviors in children with ASD versus neurotypical children. The main findings were a prevalence of underweight and obesity of 18.4 and 16.3%, respectively, among the children with ASD on considering BMI. The figure in turn increased to 47.5% for the latter fat mass category on considering the body fat component.
Furthermore, although no differences in median nutrient intake were observed between the groups, 50% of the children with ASD showed high inadequacy for nutrient intake due to inadequate intakes corresponding to more than 10 nutrients, high food selectivity, as well as more disruptive mealtime behavioral problems compared to neurotypical children.
In  et al. (2012) found 5-to 11-year-old children with ASD to have a greater incidence of low weight, while those between 2 and 5 years of age showed a greater incidence of overweight and obesity, compared to neurotypical children. In our study, there was a greater presence of low weight as well as a higher frequency of children with obesity in the ASD group when BMI was considered (moderate effect). This is in agreement with the study published by Castro et al. (2017), who used bioelectrical impedance as a reference technique for body composition. Based on the above, it could be concluded that children with ASD might be characterized by a deviation toward both low weight and obesity, with a similar distribution of both extremes.
Evaluating food intake is a key element in nutritionally assessing a population. Several studies have reported on nutritional inadequacies in children with ASD Sharp, Berry, Burrell, Scahill, & McElhanon, 2020). In this regard, Adams, Johansen, Powell, Quig, and Rubin (2011) affirmed that insufficient protein intake, metabolic imbalances or alterations in protein digestion in children with ASD may be associated to increased gastrointestinal problems. In fact, these authors reported that nutritional status depends not only on intake but also on digestion, absorption, metabolic processing and metabolic demand. Regarding dietary intake, children with ASD consume 16% less calories, a greater percentage of carbohydrates, 37% less protein, and 29% less fats (Neumeyer et al., 2018). Interestingly, from 34.9 to 39.5% of our children with ASD showed an altered consumption of calories-both insufficiency and excess-this being mainly due to an unbalanced intake of certain macronutrients such as fats and fiber, with a moderate effect and described by the homogeneous response despite the sample size in both groups. This fact could also be related to the abovementioned distribution of the children between extreme nutritional status categories (underweight and obese). In contrast to other studies, a high protein intake was observed in both groups-this being consistent with data previously reported in Spanish children (Serra-Majem, Ribas-Barba, Pérez-Rodrigo, & Bartrina, 2006). In relation to micronutrient intake, First, our study applied a more conservative statistical analysis, avoiding overestimations or false positives of the differences between the two groups that could appear in other studies (Esteban-Figuerola et al., 2019). Second, the children with ASD would be less likely to suffer from insufficient nutrient intake, mainly because of greater parental control over food consumption (Hyman & Johnson, 2012).
Although used in routine practice, 72-h food diaries have a number of limitations that may result in the underestimation of food intake. The data they afford therefore must be interpreted with caution (Burrows, Martin, & Collins, 2010). The FFQ is considered to be an essential tool in the qualitative assessment of nutrient intake, taking into account our Mediterranean diet model based on the food pyramid proposed by the SENC (Aranceta Bartrina et al., 2016 Atypical mealtime behavior and food selectivity are common in children with ASD (Sharp et al., 2013. In a previous study, we reported altered mealtime eating behavior in relation to severe food selectivity and its association to oral health in children with ASD Similarly, prior evidence (Chistol et al., 2018;Murphy et al., 2020;Peverill et al., 2019) would confirm such eating problems in these children, including food rejection, limited variety, unbalanced intake and food consumption frequency, and mealtime behavioral problems, where such food preferences would have been attributed to the influence of other factors such as sensory sensitivity and familial food preferences (Bandini et al., 2017). It has been highlighted that children with ASD exhibit poorer behavior during meals, as well as lower dietary quality, though food consumption questionnaires were not used to evaluate this aspect (Johnson et al., 2014). Other studies (Bandini et al., 2017;Zimmer et al., 2012) have also recorded a greater risk of nutritional insufficiency when children with ASD exhibit food selectivity. The assessment of intake alterations, the application of the FFQ, and the statistical adjustment for multiple comparisons are among the strengths of the present study, evidencing that although no differences between groups were observed in absolute values of intake and FFQ, a greater percentage of children with ASD presented highly inadequate intake (insufficient intake of 6-10 or more nutrients), as well as high food selectivity, as determined by the FFQ-though with a weak effect size. We therefore should be cautious when generalizing such differences. In line with our findings, Sharp et al. (2018) found that 78.5% of the children with ASD followed a diet with an inadequate intake of five or more nutrients. Additionally, these food selectivity and nutritional inadequacies were greater in the group of children with ASD, who were the children with the most altered mealtime behavior (moderate effect), explaining that these differences would stem from the greater alteration and homogeneous response of these behaviors in children with ASD.
From a holistic perspective, in order to provide an answer to the observed alterations in body composition in our children with ASD, a longitudinal study (Bandini et al., 2017)  The present study has some limitations that should be considered. We did not measure the degree of ASD, although the children with ASD who participated in the study were enrolled in specialized centers and had been previously officially diagnosed. Another limitation is the use of questionnaires, which, although validated,

| CONCLUSIONS
The present study found children with ASD to be more likely to have unsatisfactory BMI values at either the underweight or obese extremes of the range, a greater degree of inadequate intake, high food selectivity as indicated by their consumption frequency, and more disruptive mealtime behavior than neurotypical children. We suggest monitoring nutritional inadequacies and implementing nutritional strategies to expand the variety of foods children with ASD consume. The link between more disruptive eating behaviors and greater nutritional inadequacy might encourage parents and therapists in their endeavors to improve such behaviors even before body composition becomes affected.

ACKNOWLEDGMENTS
The present study was supported in part by grants from the Mutua