Background: Using traditional risk factors to select pregnant women to undergo an early oral glucose tolerance test misses some women and identifies many not at risk.
Objective: To develop an early risk prediction model for GDM based on a combination of clinical and lifestyle predictors plus single nucleotide polymorphisms (SNPs).
Methods: A training cohort comprised 2315 nulliparous participants (3% n=71 GDM, n=2244 non-GDM) from the Adelaide and Auckland SCOPE cohorts. Internal validation involved repeated random sub-sampling using 70% training and 30% validation data. External validation was in the independent STOP cohort comprised of 196 GDM (16%) and 1024 non-GDM pregnancies. Using a tiered modelling approach, two models were developed based on predictors at 15 weeks’ gestation. Tier 1 had high sensitivity while adding SNP predictors in Tier 2 resulted in higher positive predictive value. Final risk classification into low, moderate and high risk was obtained by an integrated probability from both tiers.
Results: Training data showed the model achieved a sensitivity of 91.3±8.3% (83.7% in external cohort) in Tier 1 with a NPV of 99.6±0.8% (91.7% in external cohort). At Tier 2, a further 8-10% of women were identified at high risk of GDM with a PPV of 21.7±8.1% (28.6% in external cohort).
Interpretation: The tiered model provides classification of risk for GDM into three levels, which may improve and reduce selection of women for early OGTT. All predictors can be obtained at 12-15 weeks’ gestation, allowing modifiable risk factors to be addressed and potential early interventions to reduce GDM.