Work

Dual‐center prospective validation of an early second‐trimester predictive model integrating BMI trajectories and pathophysiological biomarkers for gestational diabetes risk stratification

Year: 2025

Type: article

Abstract: OBJECTIVE: To develop and validate an early second-trimester predictive model integrating body mass index (BMI) trajectories and pathophysiological biomarkers for gestational diabetes mellitus (GDM) risk stratification, with the aim of reducing weight-monitoring frequency while maintaining predictive accuracy. METHODS: In this prospective dual-center cohort study, 1,450 pregnant women (<12 weeks gestation) without preexisting diabetes were enrolled from two Beijing maternal-child hospitals. Serial anthropometric (BMI at 6-10, 12-14, 15-19, and 24-28 weeks) and metabolic biomarkers (fasting glucose, C-peptide, lipids, uric acid) were analyzed. GDM was diagnosed via 75 g OGTT at 24-28 weeks. Multivariable logistic regression identified predictors using a 7:3 training-test split. Model performance was assessed by area under the ROC curve (AUC), calibration, and decision curve analysis. RESULTS: , 1.01-1.12), fasting glucose (OR = 1.86/mmol/L, 1.45-2.40), fasting C-peptide (OR = 1.87/ng/mL, 1.28-2.73). The model demonstrated moderate discrimination (training AUC = 0.684; testing AUC = 0.685) with 63.7-67.2% sensitivity and 59.8-61.4% specificity at the optimal threshold (0.237). Negative predictive values exceeded 82%, enabling effective risk exclusion. CONCLUSIONS: This dual-center model pioneers GDM risk stratification by 12-14 weeks using clinically accessible metrics, reducing weight-monitoring frequency without compromising prognostic value. While AUC limitations (0.68-0.69) suggest unmeasured contributors, its operational simplicity and robust negative predictive capacity support implementation in resource-constrained settings. The findings redefine antenatal care paradigms by shifting focus to early metabolic dysregulation rather than late diagnostic thresholds.

Source: Journal of Diabetes Investigation

DOI: 10.1111/jdi.70179