New Calibration Algorithm Leads to Robust CGM Performance with Reduced Number of Calibration Measurements

Since CGM sensors tend to change their behavior over time, frequent calibrations (usually twice a day) using SMBG samples are required. We discuss here the newly proposed JKU calibration algorithm able to reduce the total number of calibrations over a sensor lifetime. In the available data of clinical studies consisting of 176 records, patients performed the first calibration measurement three hours after insertion, the second one two hours afterwards, and subsequently two measurements every day (one in the morning and one in the evening) over a period of 7 days. Based on this basic setup the impact of reducing the number of calibration measurements has been studied. We compared the newly developed JKU algorithm with the manufacturer’s state of the art (SOA) algorithm, and with an algorithm from the literature based on a Bayesian Framework (BF) and specifically devoted to calibration with less SMBG measurements. From the results presented in Fig.1, one can see that the SOA algorithm cannot be used with less calibration points without a loss of performance. However, the other two algorithms, BF and JKU calibration, are more robust towards the reduction of calibration measurements. Moreover, for the JKU calibration a comparable quality can be reached with different calibration schedules (see the MARD distribution over different schedules in Fig. 2). The new calibration method can achieve the performance (in terms of MARD) of the manufacturer’s algorithm by using approximately half of the calibrations and is robust with respect to calibration schedules.