Heritability of Proposed DSM-5 Autism Symptom Domains in a Large, Clinically-Ascertained Sample

Background:  Population-based twin studies have identified strong genetic influences for both global and specific autism symptoms. These studies have also found predominantly independent heritability, suggesting unique genetic influences across social communication/interaction (SCI) and restricted/repetitive behavior (RRB). Additionally, population studies have identified similar magnitudes of genetic effects across more and less stringent thresholds, implying comparable genetic influences across typical and pathological levels of autism symptoms. Together, these conclusions support a model of autism spectrum disorder (ASD) as the confluence of extreme scores on independent liability continua. While population studies are extremely useful for examining heritability of the full range of autism traits, they are limited by the presence of a minority of diagnosed ASD cases - even within extreme score ranges. Therefore, these studies may have limited power to test for distinct genetic influences across ASD and non-ASD cases.

An alternative model posits that ASD is a distinct syndrome composed of underlying SCI and RRB symptom continua. This alternative view would predict strong common heritable influences across domains and higher heritability for categorically diagnosed ASD than for individual difference levels of the trait. A third viewpoint, supported by a recent behavior genetic study of ASD diagnoses, posits substantial contributions of shared environment.

Objectives:  The present study evaluated these competing perspectives by estimating extreme group heritability, individual differences heritability, and shared environmental effects on proposed DSM-5 SCI and RRB symptoms.

Methods:  Twin data were obtained from the Interactive Autism Network (IAN; ASD-affected twin pairs N=369). Caregivers reported autism symptoms using the Social Responsiveness Scale (SRS) and/or the Social Communication Questionnaire (SCQ). Basic and augmented DeFries-Fulker regression models were computed in extreme groups and in non-ASD twin pairs using SCQ and SRS total scores and domain scores reflecting SCI and RRB symptoms.

Results:  Extreme group heritability (h_g²) was very large (smallest β=.92, SE=.15, p<.001) and consistent across total, SCI, and RRB scores. Shared environment (c²) and individual differences heritability (h²) were not significant when extreme scores were examined (c² largest β=.54, SE=.82, p>.10; h² largest β=.52, SE=.48, p>.10). In spite of being underpowered, the difference between extreme group heritability and individual differences heritability (h_g²-h²) was always large and reached significance for SRS total scores (z≥2.0; β=1.12, SE=.35, p=.002). Bivariate (cross-construct) heritability analyses - where twin pairs are selected for a high score on one symptom domain and group heritability is calculated for the second domain - indicated substantial genetic correlations for extreme SCI and RRB scores (smallest β=.81, SE=.15, p<.001).

Conclusions:  Recent analyses of symptom data have supported an alternative view of ASD as a distinct category with SCI and RRB sub-dimensions. The present results further support this view. Heritability was stronger for extreme scores than for individual differences, and SCI and RRB domains showed overlapping heritable influences. These findings should be considered very preliminary and will require confirmation with more extensive modeling. If confirmed, results support a strong investment in genomic research and suggest that genetic influences are likely to be pleiotropic, simultaneously driving SCI and RRB symptoms.