Data Management
YY, IH and YZ independently extracted information from the included studies using a predefined data coding sheet. Authors were contacted to obtain more information if related data was ambiguous or not included in the publications. In cohorts reporting different diet quality scores, the results based on the largest sample size were included in the meta-analysis to avoid sample overrepresentation. When diet quality measures were provided on equal sample sizes, results for the diet quality score with the most items were included. In cohorts where diet quality was reported at different time points in pregnancy, second trimester measures were selected. For each of these decisions, sensitivity analyses were then performed using other reported results from the same cohort to test the robustness of the findings. For each study, results from the model adjusting for the most confounding variables were selected.
YY and IH independently assessed the quality and risk of bias of each included study with the Newcastle–Ottawa Scale (NOS)29. The sum of points was used to categorize overall study quality as either high (7-9), moderate (4–6) or low (0-3).