Clinical trial validation of the STAR-3D glycemic control framework

Glycemic control (GC) to regulate dysglycemia in critically ill patients is associated with improved outcomes. However, it has been hard to achieve safely due to metabolic variability. Protocol design is critical to provide safe, effective control to all patients. STAR is a unique model-based, patient-specific, and risk-based glycemic control framework accounting for inter- and intra- patient variability using a clinically validated digital twin model. This study presents the first clinical trial results of STAR with a new 3D stochastic model, which was shown to improve predictions of patient-specific insulin sensitivity variability and value in validated virtual trials, and can thus potentially improve glycemic control outcomes. In total, 616 patients are considered in this analysis, totaling 37489 hours of control. Overall, 77% of blood glucose (BG) measurements were in target, with a median BG of 6.7 mmol/L. There was no incidence of severe hypoglycemia (BG < 2.2 mmol/L) and only 0.3% of mild hypoglycemia (BG < 4.0 mmol/L). These results were achieved with a median [25th 75th percentiles] of 4.0 [2.0 6.0] U/h of insulin, and nutrition rates with median 100 [83 100] % of goal feed. These results were consistent when considering per-patient statistics. Using this new tri-variate stochastic model in the STAR glycemic control framework successfully provided extremely safe, effective GC for all patients, with high nutrition intake, despite targeting normoglycemic ranges. High safety from hypoglycemia, high %BG in normoglycemia, and high nutrition delivery are all associated with improved outcome in ICU patients.