Early Pregnancy Glycemic Testing for Prediction of Gestational Diabetes Mellitus and Large for Gestational Age Birth Weight.
Level 3 - non-randomized controlled study
Prospective multicenter observational cohort study developing and validating predictive diagnostic models.
PubMed 42251767 · doi:10.2337/dc26-0302
What was done
In the prospective Glycemic Observation and Metabolic Outcomes in Mothers and Offspring (GO MOMs) cohort study across nine U.S. sites (2021-2025), 2,178 individuals with singleton pregnancies without pre-existing diabetes underwent blinded 75-g oral glucose tolerance testing (OGTT) and continuous glucose monitoring (CGM) at 10-14 weeks' gestation. Predictive models for gestational diabetes mellitus (GDM at 24-28 weeks) and large for gestational age (LGA) birth weight were developed using glycemic metrics plus clinical factors (maternal age, BMI, history of GDM or macrosomia). Models were derived on data from four sites and validated across five sites, evaluating positive predictive value (PPV) at a fixed negative predictive value (NPV >=96%) assuming 8.3% GDM and 10% LGA prevalence.
What was found
Among 2,178 participants, 93.3% of those pregnant at 24-28 weeks completed OGTTs (15.4% GDM rate), and 98.1% with live births had data to determine LGA (11.2% LGA rate). At NPV >=96% and 8.3% GDM prevalence, PPV for GDM was 12.2% (95% CI 11.0-13.3%) for clinical factors alone, 26.5% (95% CI 23.1-30.3%) for OGTT plus clinical factors, and 19.8% (95% CI 17.2-22.5%) for CGM plus clinical factors. For LGA, neither glycemic model improved PPV compared to clinical factors alone.
Why it matters
Adding first-trimester OGTT or CGM to clinical risk factors improves early prediction of GDM, though it does not improve prediction of LGA birth weight. This clarifies the diagnostic utility of early glycemic testing for maternal versus fetal outcomes.
Limits
The study was observational and did not test whether early identification of GDM risk leads to effective interventions or improved perinatal outcomes. Even with testing, positive predictive values remained low (maximum 26.5%). Results are modeled based on specific fixed prevalence rates and may vary in populations with differing baseline risks.