Identifying understudied correlations between autism & phenotypic attributes in a large family dataset.

BACKGROUND

Autism Spectrum Disorder (ASD), a neurodevelopmental condition marked by restricted, repetitive behaviors and social communication difficulties, is one of the fastest-growing pediatric behavioral health concerns in the United States. Long-term outcomes significantly improve with early intervention, but diagnosis and treatment are complicated by the large range of phenotypic presentations that can be moderated by identity factors like gender and culture. Many physical and behavioral characteristics associated with the autism phenotype are not included in the screening and diagnostic instruments used in research.

METHODS

We have built a multi-site registry of diverse families with children with autism to collect longitudinal data on their physical and behavioral attributes to study the heterogeneous autism phenotype. Our KidsFirst registry contains 6,951 participants (hereafter "children") from 4,120 families, 1,865 of which have more than one child. In addition to collecting standard clinical instruments such as the Social Communication Questionnaire (SCQ), we have collected information on the phenotypic attributes of hearing issues, noise sensitivity, vision challenges, irregular sleep, impaired motor skills, metabolic disorders, gastrointestinal (GI) problems, infections, seizures, and premature birth for both ASD and non-ASD children. After validating parent-reported diagnoses against SCQ scores, we analyzed the association of each attribute with the ASD diagnosis and the other attributes using a logistic regression model and permutation tests.

RESULTS

Noise sensitivity, impaired motor skills, irregular sleep, GI problems, infections, and seizures attributes were significantly associated with autism diagnosis. These attributes also share correlation structures amongst themselves, suggesting that groupings of attributes may help to define subtypes of autism.

LIMITATIONS

The attributes analyzed in this study are not a comprehensive list of suspected traits of autism. Parent-reported diagnoses may not always be accurate, although we validated diagnoses. Despite accounting for family structure in our experiments, the relationships between attributes and diagnosis are likely stronger in the general population because our "control" sample is comprised of biological siblings who may still possess subclinical autistic traits, given the heritability of autism.

CONCLUSIONS

A more expansive conceptualization of the autism phenotype is likely to be useful to both researchers and families for identifying a more targeted approach to intervention.