Model-Based Glycaemic Control in Critically Ill Diabetes Mellitus Patients: Monte Carlo Sensitivity Analysis.

Insulin resistance and sensitivity variabilities exacerbated diabetes mellitus (DM) and non-diabetes mellitus (NDM) patients’ conditions in the intensive care unit (ICU). This problem has been affiliated with glycaemic control performance and external errors, thus, influencing the blood glucose (BG) monitoring in those patients. A model-based glycaemic control was proposed as it offers a non-invasive observation of DM patients’ insulin sensitivity (SI) in the ICU. This model-based glycaemic control used the Intensive Care Insulin Nutrition Glucose (ICING) model that combines stochastic targeted (STAR) protocol which was developed in Christchurch enabling the estimation of SI. However, lower SI in Malaysian cohorts has led to ICING model enhancement, giving better SI estimation to represent each critically ill DM and NDM patient's metabolic parameter. To identify the enhanced ICING model robustness, BG sensitivity error was added with 5% ±1 of noise error then simulated 100 times with Monte Carlo simulations. A total of 131 patients (170 DM and 101 NDM episodes) from the STAR trial in a general ICU was simulated producing 17000 and 10100 Monte Carlo simulations. The Monte Carlo analysis results showed with model enhancement, the model-based glycaemic control for Malaysian DM and NDM is robust and most importantly safe to be used with less than 0.1% of mild and severe hypoglycaemias. The median BG level, the % BG 6.0 – 10.0 mmol/L with and without Monte Carlo for DM and NDM cohort were in the target. In conclusion, through this validation, the enhanced ICING model is robust, optimised and safe to be used for glycaemic control within the DM and NDM in Malaysian ICUs.