Abstract
Stress hyperglycemia is common in intensive care unit (ICU) patients and is independently associated with increased morbidity and mortality. Traditional glycemic management relying on intermittent capillary blood glucose measurement and standard bolus calculators often leads to glycemic variability and elevated risks of hypoglycemia. This prospective, multicenter randomized controlled trial evaluated the efficacy and safety of a continuous glucose monitoring (CGM)-driven dynamic insulin dosing algorithm compared to standard bolus calculators in critically ill patients. A total of 340 adult ICU patients requiring intravenous insulin therapy across four academic medical centers were randomized 1:1 to either the CGM-driven algorithm group (n=170) or the standard bolus calculator group using point-of-care capillary blood glucose (n=170) over a 72-hour intervention period. The primary endpoint was time in target glycemic range (TIR: 100–180 mg/dL). Secondary endpoints included severe hypoglycemia (<54 mg/dL), glycemic coefficient of variation (%CV), and ICU length of stay. The CGM-driven algorithm group achieved a significantly higher TIR compared to the control group (78.4% ± 8.2% vs. 62.1% ± 11.5%, p < 0.001). Incidences of severe hypoglycemia were markedly reduced in the algorithm group (1.2% vs. 5.3%, p = 0.004), alongside lower glycemic variability (%CV: 18.3% vs. 27.6%, p < 0.001). Nursing point-of-care testing frequency was significantly lower in the CGM group. CGM-driven dynamic insulin titration significantly improves glycemic control, minimizes hypoglycemia, and reduces glycemic variability in ICU patients compared to standard bolus calculators.
