Prediction of gestational diabetes mellitus using continuous glucose monitoring metrics.
Journal
Diabetes research and clinical practice
Journal Volume
231
Start Page
113044
ISSN
1872-8227
Date Issued
2026-01
Author(s)
Abstract
Aims: We evaluated continuous glucose monitoring (CGM)-derived metrics for predicting gestational diabetes mellitus (GDM).
Methods: We analyzed data from 167 pregnant participants who had ≥ 3 days of CGM data at 18-24 weeks' gestation and who underwent 75-gram oral-glucose-tolerance-tests at 24-28 weeks in a multi-ethnic prospective cohort. Predictive performance of CGM metrics was assessed using the area-under-the-receiver-operating-characteristic-curve (AUROC) with 20 repetitions of 5-fold cross-validation; optimal cut-points were determined using Youden's index. Results: There were 30 (18 %) GDM cases. The strongest predictors were %time-above-7.8-mmol/L (%TA7.8) [AUROC (95 % CI): 0.862 (0.780, 0.945); cut-point: 1.23 %; sensitivity: 0.800; specificity: 0.847] and the hyperglycemia-component-of-the-Glycemic-Risk-Index (Hyper-GRI) [0.862 (0.779, 0.945); cut-point: 0.79; sensitivity: 0.767; specificity: 0.883]. J-index, standard deviation (SD), and mean-amplitude-of-glucose-excursions(MAGE) also achieved AUROCs > 0.80. The predictive performance of these metrics was stronger in women with BMI < 23 kg/m2 (n = 89; AUROC range: 0.813-0.882) than in those with BMI ≥ 23 kg/m2 (n = 78; AUROC range: 0.657-0.756). Among Chinese participants (n = 142), %TA7.8 and J-index had AUROC > 0.80; in non-Chinese participants (n = 25), SD performed best (AUROC: 0.845). Adding individual CGM metrics to a model including maternal age, pre-pregnancy BMI, job status, ethnicity, history of GDM, and family history of diabetes improved the AUROC from 0.642 to 0.895 (%TA7.8), 0.867 (Hyper-GRI), 0.877 (J-index), 0.868 (SD), and 0.848 (MAGE). Conclusions: CGM-derived metrics show good performance in predicting GDM and potential for earlier detection of adverse pregnancy glycemic profiles.
Methods: We analyzed data from 167 pregnant participants who had ≥ 3 days of CGM data at 18-24 weeks' gestation and who underwent 75-gram oral-glucose-tolerance-tests at 24-28 weeks in a multi-ethnic prospective cohort. Predictive performance of CGM metrics was assessed using the area-under-the-receiver-operating-characteristic-curve (AUROC) with 20 repetitions of 5-fold cross-validation; optimal cut-points were determined using Youden's index. Results: There were 30 (18 %) GDM cases. The strongest predictors were %time-above-7.8-mmol/L (%TA7.8) [AUROC (95 % CI): 0.862 (0.780, 0.945); cut-point: 1.23 %; sensitivity: 0.800; specificity: 0.847] and the hyperglycemia-component-of-the-Glycemic-Risk-Index (Hyper-GRI) [0.862 (0.779, 0.945); cut-point: 0.79; sensitivity: 0.767; specificity: 0.883]. J-index, standard deviation (SD), and mean-amplitude-of-glucose-excursions(MAGE) also achieved AUROCs > 0.80. The predictive performance of these metrics was stronger in women with BMI < 23 kg/m2 (n = 89; AUROC range: 0.813-0.882) than in those with BMI ≥ 23 kg/m2 (n = 78; AUROC range: 0.657-0.756). Among Chinese participants (n = 142), %TA7.8 and J-index had AUROC > 0.80; in non-Chinese participants (n = 25), SD performed best (AUROC: 0.845). Adding individual CGM metrics to a model including maternal age, pre-pregnancy BMI, job status, ethnicity, history of GDM, and family history of diabetes improved the AUROC from 0.642 to 0.895 (%TA7.8), 0.867 (Hyper-GRI), 0.877 (J-index), 0.868 (SD), and 0.848 (MAGE). Conclusions: CGM-derived metrics show good performance in predicting GDM and potential for earlier detection of adverse pregnancy glycemic profiles.
Subjects
Continuous glucose monitoring
Gestational diabetes mellitus
Glycemic variability
Insulin resistance
Predictive modelling
Pregnancy
Publisher
Elsevier Ireland Ltd
Type
journal article