Familial Aggregation of Regression Status and ADOS Parameters Among Individuals with ASD From the AGRE Collection

Background: It is well established that autism is one of the most heritable among behaviorally defined disorders, with concordance rates in monozygotic twins (MZ) ranging from 60—90%,  when the criteria are more narrowly or broadly defined, and the rate in dizygotic twins (DZ) at approximately 10%. Concordance rates in non-twin siblings vary between 5—8%. Despite these numbers, identification of etiologically relevant genes in autism spectrum disorders (ASD) has progressed slowly, which has prompted attempts to identify more phenotypically homogeneous subgroups of affected individuals for molecular-genetic analyses. Yet it is not clear whether select grouping variables (e.g., presence of a language delay, regressive onset of the disorder) are influenced more by genetic or environmental factors. Various investigators have discovered familial aggregation among children with ASD for traits such as nonverbal communication per the Autism Diagnostic Interview—Revised (ADI-R; Silverman et al., 2002, MacLean et al., 1999), reciprocal social interaction (Mazefsky et al., 2008), and indices of verbal/nonverbal IQ and adaptive functioning (Goin-Kochel et al., 2008). However, heritability for other important parameters, including regression status and core-deficit scores per the Autism Diagnostic Observation Schedule (ADOS), are unknown among this population and could prove equally enlightening.
Objectives: To explore familial aggregation of developmental regression, communication skills, social interaction, and stereotyped behaviors/restricted interests among siblings with ASD.
Methods: Data from the Autism Genetic Resource Exchange (AGRE) were analyzed for full, biological siblings who met criteria for an ASD per the ADOS and ADI-R (N = 1358, 78.4% male; M age = 8.0 years, SD = 4.5; 81.7% white). Frequencies of skill loss (regression) were calculated by type of loss (e.g., language, social, fine/gross motor, self-help, play), and twin correlations were computed for all MZ (n = 38) and DZ (n = 37) twin pairs to assess the potential for heritability of regression in ASD per type of loss. Subsequent Intraclass Correlation Coefficients (ICCs) were computed to make full use of this large sample and account for differences in the numbers of affected siblings per family (range = 2—5). Current analyses focus on twin correlations for the ADOS subscale-domain scores for communication skills, reciprocal social interaction, and stereotyped behaviors/restricted interests.
Results: Among this sample, 33.4% had experienced some form of skill regression; the most common losses were for language and social-engagement, experienced by 22.5% and 22.0% of the sample, respectively. Tetrachoric twin correlations for language loss were .85 (MZ) and .17 (DZ) and for social-engagement loss, .80 (MZ) and .67 (DZ); however, the ICC for language loss was .11 and that for social-engagement loss, .16. Comparable results for the ADOS subscale domains are forthcoming, as are heritability estimates for all parameters.
Conclusions: Results from the twin correlations and ICCs for regression items were inconsistent, which may indicate that the genetic mechanisms that were represented in this small DZ twin sample were not representative of the usual mix of genetic mechanisms in the AGRE population. Differences in assisted-reproduction rates between twin and nontwin siblings are currently being explored as one possible explanation.