Modeling of Effect of Glucose Sensor Errors on Insulin Dosage and Glucose Bolus Computed by LOGIC-Insulin

T. Van Herpe, B. De Moor, G. Van den Berghe, D. Mesotten
2014 Clinical Chemistry  
BACKGROUND: Effective and safe glycemic control in critically ill patients requires accurate glucose sensors and adequate insulin dosage calculators. The LOGIC-Insulin calculator for glycemic control has recently been validated in the LOGIC-1 randomized controlled trial. In this study, we aimed to determine the allowable error for intermittent and continuous glucose sensors, on the basis of the LOGIC-Insulin calculator. METHODS: A gaussian simulation model with a varying bias (0%-20%) and CV
more » ... 0% to ϩ20%) simulated blood glucose values from the LOGIC-1 study (n ϭ 149 patients) in 10 Monte Carlo steps. A clinical error grid system was developed to compare the simulated LOGIC-Insulin-directed intervention with the nominal intervention (0% bias, 0% CV). The severity of error measuring the clinical effect of the simulated LOGIC-Insulin intervention was graded as type B, C, and D errors. Type D errors were classified as acutely lifethreatening (0% probability preferred). RESULTS: The probability of all types of errors was lower for continuous sensors compared with intermittent sensors. The maximum total error (TE), defined as the first TE introducing a type B/C/D error, was similar for both sensor types. To avoid type D errors, TEs Ͻ15.7% for intermittent sensors and Ͻ17.8% for continuous sensors were required. Mean absolute relative difference thresholds for type C errors were 7.1% for intermittent and 11.0% for continuous sensors. CONCLUSIONS: Continuous sensors had a lower probability for clinical errors than intermittent sensors at the same accuracy level. These simulations demonstrated the suitability of the LOGIC-Insulin control system for use with continuous, as well as intermittent, sensors.
doi:10.1373/clinchem.2014.227017 pmid:25161144 fatcat:tgmy265okbfc3d526r32gh4uwm