We found that children with ASDs displayed more food refusal and exhibited a more limited food repertoire than did typically developing children, although food refusal was present in both groups of children. The commonly held belief that dietary “pickiness” is outgrown with age was not supported by our cross-sectional findings. We found that among typical children, levels of food refusal and food repertoire were similar by age, whereas among children with ASDs, food refusal, but not food repertoire, was marginally lower among older children. The lack of statistical significance for the test of the interaction term and the cross-sectional nature of our study preclude any conclusions with regard to the association of food selectivity measures and age.
Although anecdotal reports suggest that some children with ASDs consume a certain food or foods nearly exclusively, we did not observe this phenomenon in our sample. Parents of only four of 53 children with ASDs (7.5%) reported that their child consumed a single food more than 4–5 times per day. This suggests that HFSFI does not occur as commonly in children with ASDs as might be assumed based on anecdotal reports. However, the operational definition of HSFSI chosen for this study may have masked some of the unique eating habits of children with ASDs. For example, if a child ate macaroni and cheese for breakfast, lunch, and dinner (three times daily) this behavior would not have met our definition of HFSFI, although most would consider this eating pattern to be unusual. However, we did not observe this in our 3-day food records.
The association between limited repertoire and nutrient inadequacy suggests that a very limited diet may put any child at risk for nutritional deficiencies.4, 30
We found that children who had a narrower food repertoire were more likely to have inadequate intake of more nutrients. We chose to use the EAR,31
which is the nutrient intake needed to meet the estimated requirement of about 50% of the population, where one was available. In the absence of an EAR, we used the AI. We saw associations between nutrient inadequacy and limited food repertoire but not with food refusal. This discrepancy may be explained by the fact that food refusal was determined from the FFQ and nutritional adequacy from the 3-day food record. The FFQ assesses usual intake over the previous 12 months; limited food repertoire and nutrient inadequacy were both based on the same dietary assessment tool, i.e., a food record which covered 3 days of intake.
Gluten-free/casein-free and lactose-free diets are followed by some children with ASDs. Because children who adhere to these diets restrict all dairy products, their calcium and vitamin D intakes would be expected to be low. In this scenario, these children’s food intakes would reflect parental restriction rather than the children’s food selectivity. However, the observation that the association between food selectivity and nutrient inadequacy did not differ by autism status, even after excluding children on special diets, suggests that our observations did not reflect parental restriction.
We acknowledge several limitations in our study. Our methodology for determining food refusal and HFSFI was based on a modified FFQ. Because a parent may not offer a food that he/she believes the child would refuse, we cannot determine whether not offering a food is influenced by the child’s presumed or historical refusal of that food. We observed a great deal of variability in the number of foods that were not offered to children, which was greater among children with ASDs (range of 0 to 90 foods not offered, compared with a range of 3 to 46 foods not offered to typically developing children). Among the children with ASDs, being on a special diet was associated with more foods not being offered but not with percent of foods refused of those offered. Thus, some parents of children with ASDs offered a limited number of foods; this would preclude the child refusing those foods. Additionally, we disaggregated all fruit and vegetable items but not other aggregated foods listed in the FFQ, based on our understanding that fruits and vegetables are commonly refused and our desire to be able to accurately estimate refusal of these foods. However, aggregated entrees (such as “meatballs or meatloaf” or “roast beef or ham sandwich”) and grains (such as “English muffins or bagels”) were left together, in order to minimize participant burden. This could introduce error in our measures of refusal.
Another limitation with our measurement of food repertoire was that the food record only captured three days of intake, which may not adequately capture the variety of the typical diet. However, Falciglia et al32
compared a 3-day record with a 15-day record in children and found that the 3-day was accurate for assessment of variety in children, although more precise estimates were found when they used three non-consecutive days. We also made coding decisions that may have affected our estimate of variety. Participants reported eating approximately 900 different foods overall, so decisions about coding of food often required that we collapse several items into one group. We based our coding decisions on the structure of the NDSR software. For example, all breads were considered a single food, which means that white bread, whole wheat bread, and banana bread would all be considered bread. Although we instructed parents to report sandwiches by their ingredients, some did not, requiring the sandwich to be considered a single food. Any errors introduced in this way would be expected to affect both groups similarly.
Generalizability is an issue in all observational studies. Parents who describe their children as picky eaters are often concerned that their child’s lack of variety prevents them from getting optimum nutrition, and thus could be disproportionately attracted to a study like ours. In order to limit ascertainment bias, we developed recruitment materials that described our study purpose as understanding the mealtime and activity patterns of children with and without autism, without regard for whether parents had concerns in these areas. However, it is possible that parents of children with unusual eating patterns may have been more interested in participating in the study.
Our study operationalizes the definition of food selectivity and provides data from a moderate-sized sample that supports the notion that food selectivity is more common in children with ASDs than in typically developing children. Further, we have shown that one aspect of food selectivity—limited food repertoire—is associated with inadequate intake of nutrients. These findings, if confirmed, suggest that limited repertoire may be of concern. Future research is needed to determine the antecedents to food selectivity as well as to develop interventions that will increase food repertoire and decrease food refusal in children with ASDs. Longitudinal studies examining food selectivity are needed to understand whether food selectivity persists into adolescence and adulthood, the impact of prolonged food selectivity on nutritional status, and whether there are differences in persistence of this phenomenon between children with ASDs and typically developing children.