Association of Internalising Traits and Autistic Traits in Adolescence in a Community-Based Twin Sample

Background: Autism spectrum disorders (ASD) show a high degree of comorbidity, with an estimated 70% of individuals with ASD having an additional psychiatric disorder (Simonoff et al. 2008, JAACAP 47: 921-929). Anxiety disorders in particular are among the most commonly co-occurring conditions with ASD with a prevalence of ~40% (van Steensel et al 2011, CCFPR 14:302-317).

Objectives: The current study aimed to estimate the degree to which the overlap between autistic traits and internalising traits in adolescence in the general population is explained by genetic and environmental influences. Further analysis investigated the extent to which internalising traits were associated with specific domains of autistic traits.

Methods: Participants were part of the Twins Early Development Study (TEDS), a UK-based community sample. Parents of 12-14 year old twins completed the emotional symptoms scale of the Strengths and Difficulties Questionnaire (SDQ, Goodman 1997, JCPP 38: 581-586) and the Autism Spectrum Quotient (AQ, Baron-Cohen et al 2006, JADD 36: 343-350). Degrees of covariation between internalising traits and autistic traits (N= 2,005 twin pairs) were assessed using bivariate twin model-fitting, differentiating shared genetic and environmental influences. In a further analysis, exploratory factor analysis was conducted on the AQ. Five subscales descriptive of specific domains of autistic behaviour were derived from the factor analysis: 1) numbers/patterns, 2) social/non-social inflexibility, 3) poor mentalising, 4) solitariness, 5) poor imagination. Overlap between internalising traits and these five subscales of the AQ on genetic and environmental influences was assessed.

Results: The phenotypic correlation between internalising traits and overall autistic traits was r = 0.32 (p<0.001) and in univariate analyses, both showed moderate heritability (52% and 50%, respectively). Cross-twin-cross-trait correlations were r_MZ = 0.28 for monozygotic and r_DZ= 0.24 for dizygotic twins. In an ACE bivariate model genetic and non-shared environmental correlations were modest (r_g = 0.22, r_e = 0.16), and a high shared environmental correlation was found (r_c = 0.84).   Univariate twin model-fitting on the AQ individual factor analysis-based subscales revealed modest heritability for numbers/patterns (33%) and poor imagination (42%) and moderate-to-high heritabilities for social/non-social inflexibility (52%), poor mentalising (69%) and solitariness (78%).  In bivariate analyses of internalising traits with each AQ subscale, inflexibility showed the highest amount of shared genetic influences, while all other subscales showed low or non-significant genetic overlap with internalising traits.  There was some evidence that shared and non-shared environment explained part of the covariation between each AQ subscale and internalising traits.

Conclusions: These results suggest that both genetic influences and shared environmental influences play a role in the association between autistic traits and internalising traits during adolescence. It is important to consider separately the specific domains of autistic trait-like behaviours in relation to co-occurring internalising traits, and their different aetiologies illustrate the complexity proposed by the fractionable autism triad hypothesis (Happé & Ronald 2008, NR 18:287-304). The present work provides new information on the relationship of internalising traits and autistic traits in early adolescence and particularly about more narrowly defined sets of autistic trait-like behaviours.