Predictive Metabolomic Markers in Early to Mid-pregnancy for Gestational Diabetes Mellitus: A Prospective Test and Validation Study

Abstract

Gestational diabetes mellitus (GDM) predisposes pregnant individuals to perinatal complications and long-term diabetes and cardiovascular diseases. We developed and validated metabolomic markers for GDM in a prospective test-validation study. In a case-control sample within the PETALS cohort (GDM n = 91 and non-GDM n = 180; discovery set), a random PETALS subsample (GDM n = 42 and non-GDM n = 372; validation set 1), and a case-control sample within the GLOW trial (GDM n = 35 and non-GDM n = 70; validation set 2), fasting serum untargeted metabolomics were measured by gas chromatography/time-of-flight mass spectrometry. Multivariate enrichment analysis examined associations between metabolites and GDM. Ten-fold cross-validated LASSO regression identified predictive metabolomic markers at gestational weeks (GW) 10–13 and 16–19 for GDM. Purinone metabolites at GW 10–13 and 16–19 and amino acids, amino alcohols, hexoses, indoles, and pyrimidine metabolites at GW 16–19 were positively associated with GDM risk (false discovery rate <0.05). A 17-metabolite panel at GW 10–13 outperformed the model using conventional risk factors, including fasting glycemia (area under the curve: discovery 0.871 vs. 0.742, validation 1 0.869 vs. 0.731, and validation 2 0.972 vs. 0.742; P < 0.01). Similar results were observed with a 13-metabolite panel at GW 17–19. Dysmetabolism is present early in pregnancy among individuals progressing to GDM. Multimetabolite panels in early pregnancy can predict GDM risk beyond conventional risk factors.