The use of parent‐completed questionnaires to investigate developmental outcomes in large populations of children exposed to antiseizure medications in pregnancy.

Objective Methods Results Significance This study was undertaken to assess the utility of the Ages and Stages Questionnaire–3rd Edition (ASQ‐3) and the Vineland Adaptive Behavior Scales–2nd Edition (VABS‐II) as neurodevelopmental screening tools for infants exposed to antiseizure medications in utero, and to examine their suitability for use in large‐population signal generation initiatives.Participants were women with epilepsy who were recruited from 21 hospitals in England and Northern Ireland during pregnancy between 2014 and 2016. Offspring were assessed at 24 months old using the Bayley Scales of Infant Development–3rd Edition (BSID‐III), the VABS‐II, and the ASQ‐3 (n = 223). The sensitivity and specificity of the ASQ‐3 and VABS‐II to identify developmental delay at 24 months were examined, using the BSID‐III to define cases.The ASQ‐3 identified 65 children (29.1%) as at risk of developmental delay at 24 months using standard referral criteria. Using a categorical approach and standard referral criteria to identify delay in the ASQ‐3 and BSID‐III at 24 months, the ASQ‐3 showed excellent sensitivity (90.9%) and moderate specificity (74.1%). Utilizing different cut‐points resulted in improved properties and may be preferred in certain contexts. The VABS‐II exhibited the strongest psychometric properties when borderline impairment (>1 SD below the mean) was compared to BSID‐III referral data (sensitivity = 100.0%, specificity = 96.6%).Both the ASQ‐3 and VABS‐II have good psychometric properties in a sample of children exposed to antiseizure medications when the purpose is the identification of at‐risk groups. These findings identify the ASQ‐3 as a measure that could be used effectively as part of a tiered surveillance system for teratogenic exposure by identifying a subset of individuals for more detailed investigations. Although the VABS‐II has excellent psychometric properties, it is more labor‐intensive for both the research team and participants and is available in fewer languages than the ASQ‐3. [ABSTRACT FROM AUTHOR]