Maternal plasma diacylglycerols and triacylglycerols in the prediction of gestational diabetes mellitus
Guixue Hou, Ya Gao, Liona C. Poon, Yan Ren, Chunwei Zeng, Bo Wen, Argyro Syngelaki, Liang Lin, Jin Zi, Fengxia Su, Weiwei Xie, Fang Chen, K. H. Nicolaides
Abstract
OBJECTIVE: To define the lipidomic profile in plasma across pregnancy, and identify lipid biomarkers for gestational diabetes mellitus (GDM) prediction in early pregnancy. DESIGN: Case-control study. SETTING: Tertiary referral maternity unit. POPULATION OR SAMPLE: Plasma samples from 100 GDM and 100 normal glucose tolerance (NGT) women, divided into a training set (GDM first trimester = 50, GDM second trimester = 40, NGT first trimester = 50, NGT second trimester = 50) and a validation set (GDM first trimester = 45, GDM second trimester = 34, NGT first trimester = 44, NGT second trimester = 40). METHODS: weeks), and tested by ultra-high-performance liquid chromatography coupled with electrospray ionisation-quadrupole-time of flight-mass spectrometry; The GDM prediction model was established by the machine-learning method of random forest. MAIN OUTCOME MEASURES: Gestational diabetes mellitus. RESULTS: In both the GDM and NGT group, lyso-glycerophospholipids were down-regulated, whereas ceramides, sphingomyelins, cholesteryl ester, diacylglycerols (DGs) and triacylglycerols (TGs) and glucosylceramide were up-regulated across the three trimesters of pregnancy. In the training dataset, seven TGs and five DGs demonstrated good performance in the prediction of GDM in the first and second trimesters (area under the curve [AUC] = 0.96 with 95% confidence interval [CI] of 0.93-1 and AUC = 0.97 with 95% CI of 0.95-1, respectively), independent of maternal body mass index (BMI) and ethnicity. In the validation dataset, the predictive model achieved an AUC of 0.88 and 0.94 at the first and second trimesters, respectively. CONCLUSIONS: Our results have proposed new lipid biomarkers for the first trimester prediction of GDM, independent of ethnicity and BMI.