Impact of calibration algorithms on hypoglycaemia detection in newborn infants using continuous glucose monitors
Neonatal hypoglycaemia is a common condition that can cause seizures and serious brain injury in infants. It is diagnosed by blood glucose (BG) measurements, often taken several hours apart. Continuous glucose monitoring (CGM) devices can potentially improve hypoglycaemia detection, while reducing the number of BG measurements. Calibration algorithms convert the sensor signal into the CGM output. Thus, these algorithms can have a direct impact on measures used to quantify excursions from normal glycaemic levels. The aim of this study was to quantify the effects of calibration sensor error and non-linear filtering of CGM data on measures of hypoglycaemia (defined as BG < 2.6mmol/L) in neonates. CGM data was recalibrated using an algorithm that explicitly recognised the high accuracy of BG measurements available in this study. Median filtering was also implemented either before or after recalibration. Results for the entire cohort show an increase in the total number of hypoglycaemic events (161 to 193), duration of hypoglycaemia (2.2 to 2.6% of total data), and hypoglycaemic index (4.9 to 7.1µmol/L) after recalibration. With the addition of filtering, the number of hypoglycaemic events was reduced (193 to 131), with little or no change to the other metrics. These results show how reference sensor error and thus calibration algorithms play a significant role in quantifying hypoglycaemia. In particular, metrics such as counting the number of hypoglycaemic events were particularly sensitive to recalibration and filtering effects. While this conclusion might be expected, its potential impact is quantified here, in this case for at-risk neonates for whom hypoglycaemia carries potential long-term negative outcomes.