Influence Of Calibration Method On The Accuracy Of Continuous Glucose Monitoring System In Critically Ill Patients

Objective: To determine the accuracy of real-time continuous glucose monitoringsystem (CGMS) in critically ill patients, especially for the influence of calibrationmethod on accuracy.Methods: In this prospective study, we used a real-time CGMS in critically illpatients. CGMS sensors were calibrated against glucose concentration in deep venousblood sample measured by either ICU-based blood gas/glucose analyzer (BG group)or handheld point-of-care glucose analyzer (POC group). During CGMS monitoring,deep venous blood samples were obtained every2hours and glucose concentrationswere measured by standard central laboratory device (CLD), blood gas/glucoseanalyzer (BG) and handheld point-of-care glucose analyzer (POC). CGMS glucosevalues were documented simultaneously. With CLD measurement as reference,accuracy of CGMS, BG and POC glucose measurement were evaluated bycalculating relative absolute difference (RAD), Bland-Altman limits of agreementanalysis and Clarke error grid analysis (EGA).Results: Thirty-three patients were enrolled,18in BG group and15in POC group.With CLD measurement as reference, RAD of CGMS in BG group (12.6%) wassignificantly less than that in POC group (21.2%)(P<0.001). In Bland-Altmananalysis, bias in BG group (0.27mmol/L) was significantly less than POC group(1.93mmol/L)(P<0.001), and distance between upper and lower limit of agreement was narrower in BG group. The percentages of matched points in EGA zone A andzone A plus B in BG group (70.2%and98.2%) were significantly higher than those inPOC group (47.1%and92.7%)(P<0.001). The absolute difference between CLD andCGMS was increased significantly in process of time after each CGMS sensorcalibration (P=0.015). However, they were significantly lower in BG group than thosein POC group at the same time point (P<0.001). In comparison of accuracy betweenBG and POC glucose measurements, RAD in BG measurement (5.6%) wassignificantly lower than that in POC measurement (11.6%)(P<0.001). InBland-Altman analysis, bias in BG measurement (-0.26mmol/L) was significantlyless than POC measurement (1.10mmol/L)(P<0.001), but distance between upperand lower limit of agreement was similar in two measurements. The percentage ofmatched points in EGA zone A in BG measurement (94.5%) was significantly higherthan that in POC group (76.7%)(P<0.001).Conclusions: Real-time subcutaneous CGMS is reliable in glucose monitoring inICU patients. ICU-based blood gas/glucose analyzer is a good choice for CGMSsensor calibration and glucose monitoring in critically ill patients. CGMS sensorcalibration time could influence the accuracy of CGMS monitoring. Further study isneeded to identify the optimal time interval of CGMS sensor calibration.