Comparing Prospective and Retrospective Measures of Early Language Regression in Children with Autism from a High-Risk Infant Cohort

Background: Autism spectrum disorders (ASD) have been described as having two differing onset patterns, 'early' and 'regressive'. Language regression has been reported in 20-47% of children with autistic disorder at a mean age of 18-21 months. Although retrospectively-reported regression has been corroborated by home video analyses, there are no prospective data to characterize this phenomenon.

Objectives: To compare early language trajectories in siblings of children with ASD, with and without parent-reported language regression.

Methods: Language development was assessed prospectively in 189 high-risk infants (siblings of children with ASD) and 85 low-risk comparison infants using the Mullen Scales of Early Learning - Expressive Language subscale (MSEL-EL) at 12, 24 and 36-42 months and the MacArthur Communication Development Inventory (CDI) at 12 and 18 months (direct and parent-report measures, respectively).  Language regression was assessed retrospectively by parent report at 36-42 months using the Autism Diagnostic Interview-Revised (ADI-R), defined as loss of at least 5 established communicative words for at least 3 months. Diagnostic outcomes at 36-42 months were determined using the ADI-R, the Autism Diagnostic Observation Schedule (ADOS) and DSM-IV-TR, blind to prior study assessments. ASD symptoms at 12 months were assessed using the Autism Observation Scale for Infants, and at subsequent assessments using the ADOS.

Results: Language regression was reported in no controls, but in 8 siblings at mean age 16.6 ± 4.1 months, including 6 of 20 with autistic disorder (AD; 30%), 0 of 25 (0%) with other ASDs, and 2 of 144 not diagnosed with ASD (1.4%); p<.001. Within the AD group, loss or plateau of skills (decrease or no change in MSEL-EL raw scores between yearly assessments) was observed in 4 of 6 (66.7%) with parent-reported regression, and 2 of 14 (14.2%) without regression (p=.037; Fishers exact test). Language development slowed (based on decline of 1 standard deviation or more in standard scores between assessments) in 6 of 6 (100%) with parent- reported regression, and 7 of 14 (50%) without (p=.051; Fishers exact). Children with and without parent-reported regression did not differ on language (as indexed by the MSEL and CDI) or non-verbal cognitive scores (as indexed by the MSEL) at 12 months, and had similar ASD symptom severity at all time points. Among the 2 high-risk infants with parent-reported regression who did not receive an ASD diagnosis at age 3, one child, diagnosed with language disorder at 3 years, had no change in MSEL-EL raw score (and a drop in standard scores from 82 to 55) between 12 and 24 months. The second child, who did not have any clinical diagnosis at age 3 years, had MSEL-EL raw scores that increased between 12 and 24 months, and stable/increasing standard scores.

Conclusions: In our prospective high-risk cohort, parents of 30% of children with AD reported language regression a the 3-year assessment. There is some agreement between retrospective and prospective measures of language regression, but trajectory patterns overlap, suggesting that ‘early’ and ‘regressive’ onset in AD might occur on a continuum that includes developmental slowing or plateau, as well as frank regression.