Machine Learning-Derived Predictive Risk Score for Prediabetes and Type 2 Diabetes Development in Parous Women

Area of Science:

• Reproductive Health
• Endocrinology
• Data Science in Medicine

Background:

• Pregnancy offers a critical opportunity to identify women at risk for chronic diseases like type 2 diabetes (T2D).
• Existing T2D prediction models primarily target women with gestational diabetes mellitus (GDM), leaving a gap in risk stratification for the general parous population.
• This study addresses the need for a predictive model for prediabetes or T2D in all women after childbirth, using routine pregnancy data.

Purpose of the Study:

• To develop and validate a predictive model for prediabetes or T2D risk in parous women.
• To create a simple, clinically applicable risk index derived from pregnancy variables.
• To assess the model's performance in stratifying women into different risk categories for chronic disease.

Main Methods:

• A machine-learning approach was employed to develop a risk prediction model using data from the Genetics of Glucose Regulation in Gestation and Growth (Gen3G) cohort.
• The model was used to derive the Gestational 4-variable Prediabetes/type 2 diabetes (G4PD) index.
• Validation was performed in the Project Viva cohort at 3 and 17 years postpartum.

Main Results:

• The G4PD index incorporates gestational weight gain, pre-gestational BMI, first-trimester maternal age, and GDM hyperglycemia severity.
• The model demonstrated moderate predictive performance, with an area under the receiver operating characteristic curve (ROC-AUC) of 0.696 in the Gen3G cohort and 0.682 in the 17-year Project Viva dataset.
• The G4PD index effectively stratified women into clinically relevant risk categories, identifying low-risk individuals and those with substantially elevated risks years after delivery.

Conclusions:

• The G4PD index, utilizing readily available clinical pregnancy variables, offers a practical tool for predicting prediabetes and T2D risk.
• This index provides moderate, long-term risk stratification for chronic disease in the general parous population.
• The findings support the use of pregnancy as a key period for identifying and managing future chronic disease risks in women.