Reliability of real-time continuous glucose monitoring in infants

Background Neonatal hypoglycemia is a common and treatable risk factor for neurological impairment. Real-time continuous glucose monitoring (RT-CGM) can show glucose concentration in real time. Using an RT-CGM alarm, physicians can be alerted and intervene in hypoglycemia. No reports, however, have evaluated the reliability of RT-CGM at low glucose levels in infants. This study therefore investigated the difference between blood glucose (BG) and RT-CGM sensor data at low glucose levels and assessed the optimum method of using a hypoglycemic alarm in infants. Methods We enrolled infants whose glycemic management was difficult. We calculated the mean absolute difference (MAD) and mean absolute relative difference (MARD) between BG and RT-CGM sensor data. We compared the MAD and MARD between the low BG fluctuation and high BG fluctuation groups. Results We used RT-CGM for 12 patients (29 times) and investigated 448 pairs of BG and RT-CGM sensor data. The MAD between these pairs was 9.3 +/- 8.9 mg/dL, and the MARD was 11.5%. The MAD at low glucose was 7.7 +/- 6.0 mg/dL, and the MARD was 16.2%. The MAD and MARD were 6.8 +/- 5.4 mg/dL and 7.8% in the low fluctuation group and 10.1 +/- 9.5 mg/dL and 12.7% in the high fluctuation group, respectively. Conclusions The difference between BG and RT-CGM sensor data changes with the degree of fluctuation in BG. When physicians set the hypoglycemic alarm, consideration of this difference and a change in the alarm setting according to the degree of fluctuation in BG may be useful.