Statistical methods
Statistical analyses were performed using SAS Enterprise Guide 8.2. Maternal and infant demographics were analysed using t-tests or Mann-Whitney U tests for continuous variables. Chi-square or Fisher’s Exact tests were performed on categorical data. Early pregnancy biomarkers were assessed for normality and log-transformed where the data was skewed. Outliers were not deleted, as the referent values for pregnant women are currently unknown for the chosen biomarkers. This resulted in wide confidence intervals for some biomarkers. In addition, the lowest limit of detection was used for observations that fell below this level. Biomarker univariable analysis was performed using t-tests. Significant biomarkers in univariable analyses were included in a multivariable model, adjusting for maternal characteristics associated with metabolic health (age, BMI, smoking, alcohol use and socioeconomic status). We performed two additional sensitivity analyses excluding all women who (1) developed GDM or hypertensive disease during pregnancy, and (2) those with low haemoglobin (defined as <110 g/L) at 15±1 weeks gestation.
Results:
One hundred and thirty eight women of Indian, and 5,240 women of European ethnicity were included in this study.