Title : Gestational diabetes is associated with a distinct urinary metabolomic signature in the latter half of pregnancy
Introduction: Gestational diabetics (GD) have both impaired insulin sensitivity and secretion, which may lead to perturbations in multiple metabolic pathways. Currently, screening paradigms for gestational diabetes mellitus (GDM) rely on demonstration of hyperglycemia after an oral glucose load, and are cumbersome for patients. A metabolomics approach may reveal a unique maternal metabolic signature in GD that enables an alternative screening strategy. This study was undertaken to assess whether the relative levels of late pregnancy urinary metabolites of GD differ to those of normal gravidas (NG) and to determine whether the proposed metabolites have utility to identify GDM in the latter half of gestation.
Methods: This nested case-control study involved 46 GD and 46 NG, who were matched for maternal age, pre-pregnancy BMI and gestational age (GA) at urine collection. Exclusion criteria included multiple gestation and metabolic or cardiovascular disorders. The Global Alliance to Prevent Prematurity and Stillbirth supplied the urine samples and demographic data. Practitioners at 3 separate medical centers diagnosed GDM by glucose challenge and glucose tolerance test, according to local criteria. A metabolomics platform (Metabolon, Inc) analyzed the osmolality- corrected levels of 626 untargeted endogenous small molecules (<1000 Daltons) in urine via ultra-performance LC/MS and GC/MS. Multivariate methods (random forest accuracy, random forest GINI and boosting relative importance) were used to screen for metabolites simultaneously distinguishing GD from NG. A classification tree using the metabolites identified by screening provided the final algorithm for predicting GD vs NG.
Results: There were no significant demographic differences between GD and NG. Values displayed as mean (SD):Maternal age (years) = 32.3 (4.7) in GD; 31.8 (4.2) in NG. BMI (Kg/m2)= 31.5 (6.8) in GD; 29.9 (6.3) in NG. Gestational age (weeks) = 30.8 (3.6)in GD; 30.5 (3.0) in NG.
Three multivariate criteria simultaneously identified 8 metabolites distinguishing GD from NG. A 5-level classification tree incorporating 4 of these metabolites predicted GDM with a sensitivity of 87%, specificity of 91% and unweighted accuracy (average of sensitivity and specificity) of 89%.
Conclusion: This preliminary study reveals that the metabolic profile of random urine samples in the latter half of pregnancy was highly accurate in identifying GD versus NG. These promising results require confirmation via a larger validation study.