The major complications of late pregnancy, preterm birth (PTB), preeclampsia (PE), small-for-gestational-age (SGA) and gestational diabetes mellitus (GDM), together afflict about 25% of first pregnancies and their incidence is rising. Despite intense effort, we are unable to predict which women are at risk in first pregnancies. Accurate prediction in early pregnancy would allow tailored antenatal care and facilitate new interventions for prevention. We aimed to develop algorithms combining clinical, lifestyle and SNP genotype data for use as screening tools in early pregnancy to identify risk for PTB, PE, IUGR and GDM.
Data from 2977 women and their partners who participated in the SCOPE study in Adelaide and Auckland were included in the study. Blood was sampled for DNA extraction and SNP genotyping. Clinical and lifestyle information and pregnancy outcomes were recorded for all women, including 123 PTB, 167 PE, 142 SGA and 89 GDM pregnancies. Two prediction models for each pregnancy complication were developed by penalized logistic regression and correction for FDR with final integration using Baye’s theorem. Risk stratification was obtained in 3 tiers: low, moderate and high risk for each complication.
Models included clinical, lifestyle and SNP genotype data for each complication with SNPs contributing up to 53% of weighted risk prediction. The percentage of women predicted at low risk who subsequently developed the complication, i.e. false negatives, was 1.3% PTB, 1.5% PE, 2.1% SGA and 0.7% GDM. Positive predictive values in women deemed at high risk were 23.6% for PTB, 22.9% for PE, 20.4% for SGA and 21.8% for GDM. Our models accurately predicted 88% of PTB, 91% of PE, 86.6% of SGA and 92% of GDM cases.
We now need to validate our models prospectively in new cohorts. These screening tools, if confirmed, may enable early identification of women at risk and early interventions.