Numerical and clinical accuracy of the freestyle libre 2 system : a diagnostic accuracy study In critically Ill patients

Introduction: Poor glycemic control in critically ill patients increases mortality and the risk of hypoglycemia, especially among those receiving insulin therapy. Although continuous glucose monitoring (CGM) has shown clinical benefits, its use in intensive care units (ICUs) is not currently recommended, and existing evidence primarily comes from older devices or less severely ill populations. The accuracy of the FreeStyle Libre 2 system in critically ill patients needing insulin therapy remains inadequately defined. This study assessed the numerical and clinical accuracy of FreeStyle Libre 2 compared with capillary glucose monitoring, using central laboratory glucose as the reference standard, obtained from either central venous catheter or arterial line samples. Methods: We conducted a diagnostic accuracy study using simultaneous measurements of FreeStyle Libre 2, capillary glucose, and central laboratory glucose (venous or arterial) in adult ICU patients with diabetes or stress hyperglycemia requiring insulin therapy and ventilatory or vasopressor support. Numerical accuracy was assessed using ISO 15197:2013 criteria and Mean Absolute Relative Difference (MARD). Clinical accuracy was evaluated using Clarke and Parkes error grid analysis. Results: A total of 123 paired measurements were collected from five critically ill patients. FreeStyle Libre 2 showed a MARD of 10.98%, similar to capillary glucose (7.29%), with no significant difference. Neither method met ISO 15197:2013 accuracy standards. In clinical accuracy analyses, FreeStyle Libre 2 classified 100% of measurements within zones A+B on both Clarke and Parkes grids, while capillary glucose achieved 97.6% and 100%, respectively. Two hypoglycemic episodes were recorded; FreeStyle Libre 2 detected one, while capillary glucose detected none. Conclusions: FreeStyle Libre 2 showed numerical accuracy comparable to capillary blood glucose and consistently reliable clinical performance in critically ill patients needing insulin therapy. These results support its potential as an alternative to capillary glucose monitoring in the ICU; however, more studies with a higher number of hypoglycemic events are necessary to confirm its effectiveness in this critical range.