First-Trimester Prediction of Gestational Diabetes Mellitus Based on Maternal Risk Factors

Abstract

Objective

To develop and validate a new first-trimester model for the prediction of gestational diabetes mellitus (GDM) based on maternal demographic characteristics and elements of medical history.

Design

Prospective cohort study.

Setting

Inner-city hospital.

Population

41 587 women with singleton pregnancies at 11⁺⁰–13⁺⁶ weeks' gestation, including 4231 (10.2%) who subsequently developed GDM.

Methods

Logistic regression model for GDM was developed and fivefold cross-validation was performed to assess the calibration and predictive performance of the model, assessed by the area under the receiver operating characteristic curve (AUROC) and detection rates (DRs) at different screen positive rates (SPRs).

Main Outcome Measure

GDM.

Results

In both parous women with a previous history of GDM and nulliparous women or parous women with no history of GDM, significant contributors to the prediction of GDM were maternal age, weight, height, ethnicity and family history of diabetes mellitus. In parous women with no previous history of GDM, there was a contribution from the birthweight z-score of the previous pregnancy. There was good agreement between the predicted risk and observed incidence of GDM (intercept 0.000, 95% CI: −0.034, 0.034; slope 1.000, 95% CI: 0.967, 1.033). The AUROC curve was 0.757 (95% CI: 0.749, 0.765). The performance was higher for GDM treated with insulin versus metformin or diet alone. At SPR of 40%, the DR of the insulin, metformin and diet alone group was 87.2% (95% CI: 84.9, 89.3), 80.0% (77.8, 82.0) and 61.5% (59.2, 63.7), respectively.

Conclusion

Assessment of risk for GDM can be achieved in the first trimester based on maternal risk factors.