Objective. To better understand the autistic regression phenotype and test whether previous definitions have been too narrow.

Method. Taken from an epidemiological sample, 271 subjects with autism between 6 and 19 years of age were categorized into one of three onset groups: (1) Narrowly-defined regression—subjects with expressive language loss, (2) Broadly-defined regression—subjects with partial language loss and/or losses in non-language domains, or (3) Early Onsey—subjects without regression. Using a combination of parental report and data collected from California's Regional Center system, a range of early development and outcome measures, regression characteristics, demographic variables, and potential etiological factors were compared between the onset groups.

Results. Almost no differences were found among the onset groups on the outcome variables examined (i.e., autism severity, social, communicative, and adaptive outcomes). Similarly, the two samples of subjects with regression did not differ on the frequency of social and adaptive/motor loss or demographic variables. A substantial proportion of subjects with regression (81% of Narrowly-defined regression subjects and 55% of Broadly-defined regression subjects) indicated having delayed language or social development prior to the onset of their regression. No significant differences were found between the onset groups in any of the potential etiological factors examined (i.e., rate of seizures, MMR vaccination history, pregnancy complications and risk behaviors during pregnancy, and gastrointestinal symptoms).

Conclusions. Regression involving losses other than language is very common and represents a sizeable portion of the onset types in autism. Few differences were found between the two samples of subjects with regression, thus supporting the combination of the two regression groups into one when classifying autistic regression. In addition, early developmental delays are common in individuals with regressive autism, further supporting the need to broaden the regression definition to include such early delays. Finding virtually no group differences among a range of outcome measures between children with and without regression may indicate that, once the symptoms of autism start, the course of the disorder is quite similar and differences do not magnify over time in the course of development.