A longitudinal item response model for Aberrant Behavior Checklist (ABC) data from children with autism

Area of Science:

• Pharmacometrics
• Psychometrics
• Autism Spectrum Disorder Research

Background:

• The Aberrant Behavior Checklist (ABC) is widely used in autism clinical trials.
• Characterizing longitudinal changes in ABC data presents analytical challenges due to its heterogeneous nature.

Purpose of the Study:

• To develop and validate the first item response theory (IRT) model within a pharmacometric framework.
• To characterize longitudinal changes in ABC data for children with autism.
• To provide an alternative analytical approach to traditional methods for ABC data.

Main Methods:

• Utilized data from 120 children with autism, comprising 20,880 observations of 58 ABC items over three months.
• Developed longitudinal IRT models incorporating five latent disability variables based on ABC subscales (irritability, lethargy, stereotypic behavior, hyperactivity, inappropriate speech).
• Modeled observed ABC item scores as a function of subject disability over time.

Main Results:

• The IRT pharmacometric models accurately described longitudinal changes in patient disability.
• Estimated distinct time-courses of disability for each ABC subscale.
• Demonstrated a reduction in model-estimated disability post-therapy initiation, particularly for hyperactivity.

Conclusions:

• The developed IRT pharmacometric framework offers a robust method for analyzing longitudinal ABC data in autism.
• IRT effectively captures the heterogeneity of ABC data, leading to more accurate analyses than traditional approaches.
• This model can serve as a valuable alternative for analyzing autism clinical trial data.