Specific Language Impairment As Autism Endophenotype: A Meta-Analysis of Infant Sibling Studies

Background:    The characterization of language impairments in the siblings of children with autism has considerable implications for understanding the genetic transmission of autism spectrum disorder (ASD), as well as the relationship between social cognition and language development. Previously, we observed the co-occurrence of language delay and subclinical autistic social impairments in a large sample of undiagnosed siblings of individuals with autism.

Objectives:   To clarify the evidence that early language impairment, with or without accompanying social impairments, constitutes an autism endophenotype.

Methods:  A meta-analysis was performed incorporating the scientific literature and unpublished data from the Infant Brain Imaging Study (IBIS), an ongoing multi-site study tracking brain and behavioral development in siblings of children with ASD and a low-risk control group. Seventy-four articles were identified via a search using the keywords “autism,” “language,” and “sibling” for studies published in the PubMed database during the previous 10 years.  Seventeen satisfied the following inclusion criteria: 1) implementation of standardized assessments of language and/or ASD, 2) inclusion of siblings of children with ASD (sibs-ASD) and typically developing children (TD), and 3) language data for both sibs-ASD and TD. A meta-analysis was performed incorporating the 7 published studies involving independent participants under 36 months (range 12-27 months) and data from IBIS.  Among these 8 studies, 5 categorically identified subjects with and without language impairment (sibs-ASD: n=364, TD: n=228) while 5 reported expressive and receptive language scores from the Mullen Scales of Early Learning (MSEL) (sibs-ASD: n=389; TD: n=222).  The Mantel-Haenszel (MH) method was used for dichotomous data and a random effects model with inverse variance weighting was used for continuous data. In the IBIS sample, the relationship between social motivation, as indexed by a subset of items from the Autism Observation Schedule for Infants (AOSI) at 12 months, and MSEL language scores was explored using a partial correlation controlling for the MSEL Early Learning Composite.    

Results: The MH statistic indicated increased language impairment in sibs-ASD (χ²_MH=4.36, 1 d.f., p=0.037).  Both receptive and expressive language scores showed positive effect sizes when comparing TD to sibs-ASD, signifying lower language scores in sibs-ASD.  The summary effect size was moderate for receptive language, 0.540 (95% confidence interval: 0.450-0.630), and small for expressive language, 0.210 (95% confidence interval: 0.0043-0.416). None of the published studies examined correlations between language and social impairment.  In IBIS, a significant correlation was observed for receptive language and social motivation items on the AOSI (r=-0.200, p=0.016), demonstrating an association of greater receptive language with decreased social deficits.

Conclusions:  There is a dearth of studies characterizing language phenotypes in sibs-ASD. Current data support that language impairment comprises an endophenotype for ASD during early childhood.  Receptive language appears more affected than expressive language, similar to previous reports of ASD populations, although the effect on expressive language may be independent of social function. More detailed characterization of this language impairment is needed to refine genetic investigations of ASD and our understanding of factors influencing the developmental progression of autistic traits.