Sub-group and sensitivity analyses
To address possible bias by exposure misclassification, we replicated
the main analyses for SSRI/SNRI ever exposure in a sub-sample of women
enrolled in MoBa since 2004, and compared “truly SSRI/SNRI-exposed”
with “truly unexposed” pregnancies, based on concordant exposure
information between self-reported and prescription dispensation data. To
document confounding by maternal pre-existing psychiatric illness, we
estimated the independent association of this factor with child ADHD,
and further adjusted the weighted effect estimates for SSRI/SNRI
exposure by this covariate. To assess the robustness of the findings, we
carried out additional sub-group and sensitivity analyses, as described
in detail in the Supplement.
Up to 16.5% of the pregnancies had missing values in at least one of
the sufficient confounders. Under the assumption that data were missing
at random, we imputed incomplete data via multiple imputation with
chained equation (ten replications).31-33 More detail
is provided in the Supplement.