230-OR: A Real-Time Recalibration Algorithm to Improve the Accuracy of CGM Sensors in Newborns

Introduction and Objective: Over one third of infants in the U.S. are considered at-risk for neonatal hypoglycemia (NH), which is associated with long-term neurodevelopmental sequelae. Currently, neonatal blood glucose is assessed intermittently, missing over 25% of NH events. Continuous glucose monitoring (CGM) sensors, providing glycemic data every 5 minutes, are standard care for individuals with diabetes and represent a possible solution to improve the monitoring of NH. However, CGM sensors are not optimized for the physiology of neonates. Here we present interim data aimed to develop and validate an algorithm that allows the real-time recalibration of CGM sensors when applied to newborns. Methods: Pregnant women with diabetes and non-diabetic controls were recruited in the third trimester to participate in this IRB-approved study. A factory calibrated Dexcom G6 sensor was placed within 2 hours of birth and point of care blood glucoses (POCBG) were measured per standard of care guidelines with an Abbott Freestyle Precision Pro glucometer. We developed a recalibration algorithm, based on a linear regression model, that exploits the first 2 POCBG and can be applied right after the second measurement. The accuracy of original and recalibrated CGM was measured via Mean Absolute Relative Difference (MARD). Results: On the 12 individuals, a total of 59 POCBG references were available to assess sensor accuracy. The MARD of original CGM data was 40.8%, with a mean error (ME) of 25.2 mg/dL. The recalibration algorithm reduced the MARD of CGM to 12.5% and the ME to 3.5 mg/dL, which was a statistically significant improvement (p<0.01). Conclusion: Application of CGM devices for newborns is key for NH management, but original CGM data do not provide sufficient accuracy to guide clinical care. The proposed real-time recalibration algorithm enhances the accuracy of CGM, approaching the approved accuracy of the devices, potentially enabling a better detection of NH and evaluation of glycemic-control metrics. Disclosure F. Prendin: None. S. DelFavero: Research Support; Dexcom, Inc. Other Relationship; Dexcom, Inc. S. Cherkerzian: None. A. Coburn-Sanderson: None. R. Abdel-Rahman: None. C. Monthe-Dreze: None. A. Galderisi: None. S. Sen: Other Relationship; Dexcom, Inc., nova. A. Facchinetti: None.