Regression Rates Differ According to the Operational Definition Employed and ASD Subgroup Status

Background:  A recent meta-analysis reported different rates of regression depending on whether the term regression was operationalized as encompassing mixed, language, social, or language/social (Barger, Campbell, & Donald, 2011). This work is extended by investigating whether reported regression rates differ between subgroups of children with autism spectrum disorder (ASD) for each of these operationalized regression terms. 

Objectives:  We performed meta-analyses of the literature to determine whether (a) reported rates of regression differed for children with a diagnosis of Autism (N= 3149), Pervasive Developmental Disorder (PDD, N=1650), and Asperger’s (N=610) and (b) reported rates of regression differed between girls (N=652) and boys (N=3,050) with autism spectrum disorders (ASD). 

Methods:  Autism diagnosis and regression data were extracted from 21 published studies across 5,771 participants. Twelve studies reported data on mixed, 11 studies reported data on language, 1 study reported data on social, and 1 study reported data on language/social regression. Gender and regression data were extracted from 19 published studies across 3,702 participants. Ten studies reported data on mixed, 8 studies reported data on language, 1 study reported data on social, and 3 studies reported data on language/social regression. 

Results:  Regarding regression and diagnostic status, across all regression types, children with a diagnosis of Autism had a higher rate of reported regression (31%) than those with a diagnosis of PDD (26%) or Asperger’s (14%). This same pattern was found for mixed (autism=39%; PDD=31%; and Asperger’s=19%) and language (Autism=36%; PDD=25%; and Asperger’s=13%) regression. Prevalence of social and language/social regression could not be calculated due to a lack of data. Furthermore, reported rates of mixed and language regression did not differ statistically for children with Autism (39% and 36% respectively), but reported rates of mixed regression were higher than language regression for both children with PDD (31% and 25% respectively) and Asperger’s (19% and 13% respectively). Regarding regression and gender, across all regression types, girls (31%) did not differ from boys (34%). This same pattern was found for language (girls=31%; boys=33%) and language/social (girls=53%; boys=41%) regression; however, genders did differ in regards to rates of mixed (girls=37%; boys=43%) regression. Data were insufficient to calculate social regression. Furthermore, for girls with an ASD, rates of regression were higher for language/social (53%) compared to rates of mixed (37%) and language (31%) regressions, which did not differ from one another. For boys with an ASD, rates of regression were lower for language regression (33%) compared to rates of mixed (43%) and language/social (41%) regression, which did not differ from one another.

Conclusions:  These findings indicate that reported rates of regression may differ among and within subgroups of children with ASD, depending on the operationalization of the term regression. Across regression types, children with autism had a higher rate of regression compared to PDD and Asperger’s, which did not differ from one another.