The utility of pregnancy hemoglobin A1c (HbA1c) during pregnancy for predicting gestational diabetes mellitus (GDM) is limited due to altered erythrocyte lifespan and kinetics. Continuous glucose monitoring (CGM), however, has shown promise for tracking and potentially diagnosing GDM. We aimed to compare the predictive value of early pregnancy HbA1c and CGM-derived metrics for GDM diagnosis in mid-to-late pregnancy.

We studied 289 multi-ethnic Asian pregnant women with overweight/obesity (body mass index [BMI] ≥23 kg/m^2) from a hospital-based prospective cohort. At 11-13 weeks of gestation, blinded CGM devices and random plasma collection were performed. HbA1c was measured enzymatically. CGM-derived parameters (e.g., J-Index, mean of daily differences [MODD], mean amplitude of glycemic excursions [MAGE]) were computed using EasyGV software. GDM screening at 24-28 weeks of gestation followed the International Association of Diabetes and Pregnancy Study Group (IADPSG) criteria, with women later initiated on GDM treatment also classified as GDM. Student t-test, Mann-Whitney U test, Chi-square test, and Fisher’s exact test compared maternal characteristics and CGM metrics between GDM and non-GDM subjects. Logistic regression examined associations of HbA1c and CGM metrics (per standard deviation [SD] increase) with GDM risk, adjusting for maternal age, ethnicity, BMI and systolic blood pressure [SBP] assessed at first trimester). Stepwise regression was applied for model prediction selection.

68 out of 289 (23.5%) participants were diagnosed with GDM. First-trimester HbA1c and most CGM metrics varied significantly between GDM and non-GDM subjects. In regression analysis, per SD increase in HbA1c (0.304%) was associated with a 3.11-fold in higher GDM incidence (95% CI 2.02, 4.80). Per SD increase in mean glucose (0.52 mmol/L), midnight average glucose (0.51 mmol/L), glucose variability (4.41%), time in high/very-high range (1.85%), and MODD (0.25) were all strongly associated with GDM risk (all p<0.001). Prediction models using CGM-derived metrics (e.g., J-Index, MODD, and MAGE) outperformed both traditional risk models (e.g., maternal age, ethnicity, BMI, SBP) and HbA1c-only models, showing the highest area under curve (AUC) (0.803 vs. 0.687 vs. 0.773), and the best R^2 (0.230 vs. 0.0754 vs. 0.177).

First-trimester CGM-derived metrics demonstrated superior predictive ability for GDM compared to traditional risk factors and HbA1c. CGM may provide earlier and more precise risk stratification in pregnant women with overweight/obese.