Development of an HbA1c prediction model using continuous glucose monitoring metrics in pediatric type 1 diabetes: insights into average glucose and recent glycemic trends

Lee, Hakyung; Yang, Mi; Kim, Hwa Young; Kim, Jaehyun

doi:.0.0.0

Ann Pediatr Endocrinol Metab > Accepted Articles

DOI: https://doi.org/10.6065/apem.2550214.107 [Accepted] Published online November 19, 2025.

Development of an HbA1c prediction model using continuous glucose monitoring metrics in pediatric type 1 diabetes: insights into average glucose and recent glycemic trends

Hakyung Lee¹, Mi Yang¹, Hwa Young Kim¹, Jaehyun Kim¹

¹Department of Pediatrics, Seoul National University Bundang Hospital, Seongnam, Korea
²Department of Pediatrics, Seoul National University College of Medicine, Seoul, Korea

Address for correspondence: Jaehyun Kim
Email: joyminer@gmail.com

Received: June 30, 2025 Revised: September 16, 2025 Accepted: October 3, 2025

Abstract

Purpose
Given the limitations of glycated hemoglobin (HbA1c), continuous glucose monitoring (CGM) metrics have been proposed as complementary indicators of glycemic control. This study evaluated the association between CGM metrics and HbA1c and developed HbA1c prediction models in Korean pediatric patients with type 1 diabetes (T1D).
Methods
We retrospectively analyzed CGM data from 85 patients aged 2–18 years using real-time CGM systems (G6 or G7, Dexcom, USA). CGM records over 12 weeks were segmented into five intervals (0–2, 0–4, 4–8, 8–12, and 0–12 weeks) prior to HbA1c measurement. Metrics included time-in-range (TIR), time-above-range (TAR), time-below-range (TBR), time-in-normoglycemia (TING), coefficient of variation (CV), and average glucose. HbA1c prediction models were constructed using ridge regression and validated in a separate test dataset.
Results
TIR consistently showed the strongest negative association with HbA1c, while TAR and average glucose showed the strongest positive associations. Among all intervals, 0–4 week CGM data demonstrated the strongest relationship with HbA1c (all P<0.05). Average glucose achieved the best explanatory power among all metrics (R²=0.83, AIC=84.34), and prediction models incorporating average glucose and TAR yielded the lowest mean squared error (MSE=0.15) and highest R² (0.83), with robust results in the test dataset.
Conclusion
Short-term CGM metrics, particularly average glucose during the 0–4 week preceding HbA1c testing, are strong predictors of HbA1c. These findings support the clinical utility of recent CGM data in optimizing the individualized glycemic management in pediatric patients with T1D.

Keywords: Continuous glucose monitoring, Glycated hemoglobin, Pediatrics, Type 1 diabetes