Endocrine Abstracts

JOINT676

Introduction: HbA1c has long been used as the gold standard marker for reflecting average blood glucose levels over the previous three months. As continuous glucose monitoring (CGM) systems are becoming a part of the standard care for type 1 diabetes (T1D), there is greater recognition of the limitations of HbA1c in revealing hyperglycemia, hypoglycemia, and glycemic variability. Meanwhile, CGM metrics such as Time in Range (TIR) have emerged as important parameters for glycemic control. In this study, we aimed to explore the relationship between CGM metrics and HbA1c in Korean pediatric patients with T1D and to develop and validate the HbA1c prediction models based on different CGM metrics and time intervals.

Methods: A total of 85 children and adolescents with T1D, using CGM (G6 or G7, Dexcom, USA), were included. Twelve weeks of CGM records were analyzed to evaluate the relationship between HbA1c and CGM metrics, including TIR, Time Above Range (TAR), Time Below Range (TBR), Coefficient of Variation (CV), and Average Glucose, across time intervals of 0–2 weeks, 0–4 weeks, 4–8 weeks, 8–12 weeks, and 0–12 weeks prior to HbA1c measurement. HbA1c prediction models were developed using Ridge regression modeling and cross validation with CGM metrics as the training data set. Additionally, to assess the performance of the HbA1c prediction models, a separate test data set comprising 12-week CGM records from 80 patients was used.

Results: Average Glucose had the greatest impact on A1c across all time periods, and was identified as the most suitable metric for HbA1c prediction with higher R² and lower AIC values than TIR and TAR. Validation using test data confirmed that regression models based on Average Glucose demonstrated the best performance, with the lowest MSE (MSE=0. 1425) and highest R² (R²=0. 8444). Among time periods, the 0–4 week interval consistently had the largest coefficients, suggesting it is the most predictive of HbA1c compared to other intervals.

Conclusion: Average Glucose, TIR, and TAR are highly related with HbA1c, with Average Glucose being the most predictive metric. The 0–4 week period showed the strongest association with HbA1c changes, highlighting the importance of closely reviewing recent glycemic data, especially in the 4 weeks preceding an HbA1c test, to better understand glycemic trends and inform clinical decisions.

Keywords: Continuous glucose monitoring; CGM metrics; HbA1c; prediction modeling