Modeling Disordered Eating in Terms of Core Autism Symptoms Increases Heritability Estimates
Background: Although aberrant eating behaviors are a well-established problem among individuals with autism, it remains unclear whether those behaviors are related to core autism symptomatology. The answer to this question has implications both for our basic understanding of the neurobiology of autism and for the design of therapeutic interventions. If aberrant eating behaviors and core autism symptomatology arise from a shared latent neuropathology, one might expect that treatment affecting one domain could also show efficacy in the other.
Objectives: Using a sample of over 5,000 families with autism, we aimed to model aberrant eating behaviors in terms of core autism symptomatology, and determine whether these models could demonstrate a significant association between the two domains. We further aimed to quantify the SNP-heritability of traits related to aberrant eating behaviors, thus laying the groundwork for identifying biological risk mechanisms.
Methods: In partnership with the nationwide (US) genetic study, SPARK for Autism, we distributed online self-report surveys assessing sleep, eating, and gastrointestinal patterns to SPARK participants (children with autism and their parents). The analysis presented here is our first look at the NIAS data from this study, a nine-item instrument for measuring aberrant eating behaviors. The NIAS has three sub-scales: picky eating, diminished appetite, and fear of eating consequences. We trained Random Forest classifiers to predict each NIAS sub-scale score in children, using item-level data from the SCQ (measuring social communication) and the RBS-R (measuring restricted and repetitive behaviors) as predictor variables. SNP-heritability estimates were carried out on a sub-sample of 2,627 children who also had complete genetic data available.
Results: Both SCQ- and RBS-R-based multivariate models of NIAS subscales showed significant association between disordered eating and core autism symptoms (all three NIAS subscales p<0.000001). Family history of eating disorders corresponded with a significant increase in total NIAS scores in both unaffected adults (t=7.38, p<0.0001) and affected individuals (t=3.86, p<0.001). Additionally, total child NIAS scores were significantly associated with mothers’ scores on the EAT-26 (n=3676 dyads, r=0.10, p<0.0001). Despite this evidence suggestive of familiality, none of the three NIAS subscales, as reported by participants, showed significant SNP-heritability in this sample. However, the fear subscale, as predicted by the SCQ and RBS-R Random Forest, showed significant SNP-heritability (h2=0.28; S.E.=0.16). Interestingly, this subscale was the only NIAS measure significantly linked to a family history of specifically anorexia nervosa (t=4.38, p<0.00001).
Conclusions: These results suggest that disordered eating in autism is significantly linked to its core symptomatology (social communication deficits and restricted and repetitive behaviors). The lack of SNP-heritability present in NIAS subscales, and its emergence when disordered eating is modeled in terms of core autism symptoms, may suggest that further work is needed to develop instruments that measure eating behaviors in a way that is both biologically and clinically meaningful. Work is now underway to further characterize the nature of the interplay between traits related to eating, socialization, and restricted and repetitive behaviors.