Prognostic biomarkers of future diabetes in South Asian women diagnosed with gestational diabetes: a prospective cohort study

Background — Gestational diabetes (GDM) confers an increased risk of future type-2 diabetes (T2D). We aimed to identify determinants of the progression of GDM to T2D in South Asian women and develop a precision prognostics model for potential future clinical application.

Methods — This study included 247 South Asian women with GDM from the prospective South Asian Birth Cohort (START) study in Ontario, Canada. Metabolomics was performed on 2nd trimester fasting serum samples by multisegment injection−capillary electrophoresis−mass spectrometry. We determined incidence of postpartum T2D through validated diagnostic codes using health administrative data. We used multivariable logistic regression to identify predictors of incident T2D through backward elimination. We assessed diagnostic performance of models using area under the receiver operating characteristic curve (AUC ROC) and 5-fold cross-validation to assess model stability.

Results — Of 247 South Asian women diagnosed with GDM with a mean age of 30.9 years and median total follow-up of 9.7 years, 45 (18.2%) developed T2D within a median of 4.9 years. Pre-pregnancy body mass index, area-under-the-curve glucose, and level of education, were identified as the strongest determinants of T2D post-GDM (AUC ROC = 0.83 [95% Confidence Intervals (CI), 0.78–0.88]). Addition of metabolite biomarkers—specifically, the hypoxanthine-to-creatinine ratio—only modestly improved predictive performance (AUC ROC = 0.853 [CI, 0.80–0.91]). Folds mean (SD) of the AUC ROC from 5-fold cross-validation models was 0.81 (0.12).

Conclusions — We developed a parsimonious five-factor model combining clinical, anthropometric, and metabolic markers that accurately predicts T2D progression after GDM in South Asian women. This model can be potentially deployed at the time of GDM diagnosis to improve risk assessment and target intensive postpartum interventions.