Introduction

Maternal metabolic risk factors, including high pre-pregnancy body-mass-index(BMI), excessive gestational weight gain(GWG), and gestational hyperglycaemia, are known to induce adverse pregnancy outcomes. These include stillbirth, pre-term delivery, low or high birthweight, pre-eclampsia, and maternal postnatal diabetes/cardiovascular disease(1, 2). The prevalence of gestational diabetes mellitus(GDM) varies from 2 to 25 percent worldwide, and has increased over the last decades in parallel with the increasing obesity prevalence of women in child-bearing age (3, 4).
A variety of strategies targeting metabolic management during pregnancy have been developed to prevent endocrine-related adverse pregnancy outcomes. However, a recent individual patient data meta-analysis suggested that diet and lifestyle interventions in pregnancy only achieved modest success in reducing GWG, but have no effects on composite maternal and foetal outcomes, such as pre-term birth, macrosomia, and foetal adiposity(5). It might be partly attributed to the alterations of maternal/placental function and metabolic programming in preconception and early pregnancy, which occurs prior to when most interventions are initiated. Other than maternal weight and glucose level, maternal lipid levels during pregnancy are recognised as ignored risk factors for adverse pregnancy outcomes recently(6).
The mechanism of how maternal metabolic dysfunction is linked with neonatal health remains uncertain, however, most researchers believe it may be explained by the unfavourable intrauterine nutritional environment(7, 8). Previous studies mainly focused on one specific metabolic trait or on the number of metabolic disorders, without assessing the underlying interacted effect of the natural metabolic network (9, 10). This is due to the limited analytical ability of classical statistical methods to analyse multidimensional data. Understanding how the metabolic network influences neonatal health is crucial to future interventional studies and potentially to antenatal/prenatal advice.
In this study, we investigated the independent association of maternal metabolic traits with birthweight and cord blood insulin in the Born in Guangzhou Cohort Study(BIGCS). The inter-dependency of maternal metabolic risk factors and their association with birthweight and cord blood insulin was then mapped by Additive Bayesian Network(ABN) analysis, a robust data-driven structure discovery model, which has been widely used in other disciplines(11, 12).