A Comparison of Prospective and Retrospective Methods for Measuring Regression

Background: This study examined when and how behavioral signs of autism spectrum disorders (ASD) emerge in the first years of life. Most previous investigations have been retrospective, relying on parent report of earlier development or analysis of home videotape of infants later diagnosed with ASD. The existing literature suggests that behavioral signs of autism emerge in two different patterns, an early onset and a regressive course. Recently, however, the adequacy of a dichotomous classification of onset has been questioned. Questions about when and how behavioral signs of autism emerge may better be answered through prospective, rather than retrospective, studies. Objectives: The present investigation used a longitudinal infant sibling design to examine symptom emergence in the first years of life in infants at low and high risk for autism. One specific aim was to compare traditional retrospective methods of measuring regression to prospective methods. Methods: All 25 infants in the sample who were later diagnosed with ASD were matched on gender to 25 low-risk children determined to have typical development at 36 months. Prospective measures of onset, collected at 6, 12, 18, 24, and 36 months of age, included frequencies of gaze to faces, social smiles, and directed vocalizations, which were both coded from video and rated by examiners. The retrospective measure, collected at 36 months, was the ADI-R. Results: The frequency of gaze to faces, shared smiles, and vocalizations to others were highly comparable between groups at 6 months of age, but significantly declining trajectories over time were apparent in the group later diagnosed with ASD. Group differences were significant by 12 months of age on most variables. We computed 95% confidence intervals for the change from one visit to the next for the typically developing sample. Then we counted how many of the participants with ASD outcomes displayed declines in social-communication behaviors that were greater than the 95% confidence interval for change in the TD group. This analysis documented loss of skills in 86% of the sample with ASD. In contrast, only 17% of parents reported a regression in either language skills or social development on the ADI-R. Conclusions: These results suggest that behavioral signs of autism are not present at birth, as once suggested by Kanner, but emerge over time through a process of diminishment of key social-communication behaviors. More children may present with a regressive course than previously thought, but parent report methods do not capture this phenomenon well. Implications for onset classification systems and clinical screening will be discussed.