Predicting risk for pregnancy complications

Abstract

For years, it has been a challenge to identify women at risk of Preeclampsia (PE) and Preterm Birth (PTB), one of the leading causes of maternal and perinatal morbidity and mortality. Despite an increasing number of clinical and statistical prediction models being developed, which have been shown to outperform traditional approaches based on maternal history, due to complex underlying relationships and gene-environment interactions, identifying women at risk based on a single time-point, especially during early stages of pregnancy, remains a challenge. Therefore, this study not only aims to identify potential predictors for pregnancy outcomes and develop prediction models based on combinations of clinical measurements and Single-nucleotide polymorphisms (SNP) predictors, but also to establish a tiered prediction system by integrating risk estimates at various stage of pregnancy. This thesis contains both theoretical development and practical application of the models, with results of best models written as manuscripts for future publication. Critical issues in real-life statistical analysis, including subgroup differences, and model and variable selection (with FDR control) were discussed, as well as novel strategies on the tiered prediction model development. The results from tiered models provide prediction for PE and spontaneous preterm birth (SPTB) that not only outperform traditional approaches, but also provide an earlier prediction applicable to all pregnant women, including healthy nulliparous women. This approach also allows for regular monitoring and revision of predicted risk throughout pregnancy. This may assist in providing tailored antenatal care or interventions that could benefit both the mother and child, and to avoid unnecessary interventions for low-risk individuals, while modifiable predictors could also be addressed to reduce the risk or severity of PE or PTB.