Descriptive Analysis of Autistic Regression in Families From the Autism Genetic Resource Exchange

Background: Given the heterogeneity of autism spectrum disorders (ASD), it’s important to identify subgroups of individuals that may present with an etiologically distinct phenotype.  Children with regression may represent a subgroup of individuals with ASD that may provide insight into the biological and genetic causes of the disorder.
Objectives: To provide a descriptive analysis of individuals with ASD enrolled in the Autism Genetic Resource Exchange who report a developmental regression.  This analysis provides an evaluation of cognitive and social functioning as well as co-morbid medical conditions.
Methods: Subjects with confirmed diagnoses of autism according to the Autism Diagnostic Interview (ADI-R) and the Autism Diagnostic Observational Schedule (N=895) from 2-21 years of age were assessed as part of a study to collect data and biological samples for research into autism biology and genetics. The ADI-R provided early developmental histories. A measure of cognitive functioning was provided by the Stanford-Binet V (SBV) and Ravens scales on a subset of the groups.   
Results: The subjects were initially divided into two groups, those with a history of regression and those with no history of regression, according to the ADI-R.  598 (66.97%) subjects had no reported regression (AUT-NR) while 295 (33.03%) had some history of regression (AUT-R).  We found no significant differences in age of onset between the AUT-NR (18.6 months) and AUT-R (19.23 months) groups.   We did find significant differences in full scale IQ scores on a subset of subjects that had data from the SBV (n = 198).  We found a significant difference (p = 0.0005152) between the full scale IQ of the AUT-NR vs AUT-R groups. 
We subdivided the AUT-R group into three groups according to the focus of loss.  Seventy two subjects had reported regression of language but not social skills (AUT-RL), 96 with loss of social skills but not language (AUT-RS), and 127 with loss of social and language skills (AUT-RLS). We found that full scale IQ scores in these groups differed significantly (Kruskal-Wallis chi-squared = 13.846, df = 3, p = 0.003122).
As previously reported, we found that subjects without regression had a later age of first word then those with regression (AUT-NR, mean = 34.58, SD = 18.08; AUT-R, mean = 22, SD = 16.55).  When we divided the groups as previously shown, we found that the difference was provided by subjects with regression of language (AUT-RL, mean = 18.64, SD = 12.47) and (AUT-RLS, mean = 17.14, SD = 12.33) rather than from those with regression of social skills alone (AUT-RS, mean = 34.23, SD = 19.69).  This is significantly later than previously reported.
Finally, we examined whether regression was associated with a medical illness.  We found that approximately 26% of subjects who had regression had an associated illness whether the regression was focused on a loss of social skills, language skills, or both.
Conclusions: This study contributes to an understanding of social and language skills in children with autism with and without regression and supports the notion that children with regression may represent an etiologically-distinct phenotype.