Statistical analyses
Multiple regressions were used to estimate the effects of prenatal
maternal anxiety on birthweight and gestational age. In a full cohort
where all subjects with the exposures, outcomes and adjusting variables
were included, we regressed birthweight and gestational age on maternal
anxiety score at 17th and 30thgestational week, their interactions (Crude in Table 2-A and 2-B), and
further included adjusting variables step by step. The step1 encompassed
maternal smoking and alcohol intake, and the step 2 was those in the
step 1 and parity, birth complication, child’s sex and preeclampsia. The
step 3 was those in the step 2 and maternal age, education and marital
status. Regression analyses were performed in R 4.0.0.
In the sibling comparison design, we selected all available sibling
pairs (For families participating more than three times, the sibling
pairs were randomly selected) and computed the difference values in each
variable between a sibling and the other. Similar to the analyses in the
full cohort, we regressed the sibling-difference values in the outcomes
on those in the exposures, controlling for family shared effects. Here,
the three steps of adjustments were made in the same manner as in the
full cohort.
We performed multiple imputations to handle the missing data points in
the exposures, outcomes and adjusting variables. The mice R package was
used; the number of imputations was set to be 5 (m=5), and the
imputation method in use was predictive mean matching (method=“pmm”).
All the analyses in the full cohort and sibling-design were re-run using
the imputed data (Supplementary File x).