# Accuracy and Clinical Safety of the FreeStyle Libre 2 System in an Internal Medicine Intermediate Care Unit

## TL;DR

The FreeStyle Libre 2 glucose monitoring system is clinically safe and accurate for use in hospitalized patients, with most readings falling within acceptable error ranges.

## Contribution

This study provides real-world evidence of the FreeStyle Libre 2 system's accuracy and safety in an intermediate care unit setting.

## Key findings

- 99.03% of glucose readings fell within clinically acceptable zones on the Parkes Error Grid.
- MARD did not differ significantly between diabetic and non-diabetic patients or across glycemic ranges.
- Lower MARD was observed in patients receiving intravenous fluids or presenting with shock.

## Abstract

Introduction: Hyperglycemia and glycemic variability in hospitalized patients are strongly associated with adverse outcomes, including increased mortality and infection rates. While continuous glucose monitoring (CGM) offers the potential for real-time surveillance, its accuracy in critical patient settings remains under scrutiny. This study evaluates the real-world accuracy and clinical safety of the FreeStyle Libre 2 system in an Internal Medicine Intermediate Care Unit (IMICU).

Methods: We conducted a prospective observational study of adult patients admitted to the IMICU at Unidade Local de Saúde de São João, located in Porto, Portugal. Participants utilized the FreeStyle Libre 2 system, and sensor glucose values were paired with reference capillary point-of-care (POC) measurements. Accuracy was assessed using the mean absolute relative difference (MARD) and point accuracy thresholds (percentage of readings within 15%/15 mg/dL, 20%/20 mg/dL, and 30%/30 mg/dL of reference). Clinical safety was evaluated using the Parkes (Consensus) Error Grid Analysis (EGA). Subgroup analyses were performed to assess performance across glycemic ranges, diabetes status, and clinical comorbidities.

Results: Twenty-three patients (mean age 72.39 ± 15.51 years; 14 (60.9%) male) were included, providing 414 paired glucose readings. The overall MARD was 18.62 ± 11.87%. On the Parkes Error Grid, 226 (99.03%) paired readings fell within Zones A and B, with one (0.24%) in Zone E. Point accuracy analysis showed 46.14% of readings within 15%/15 mg/dL, 61.11% within 20%/20 mg/dL, and 85.75% within 30%/30 mg/dL. MARD values did not differ significantly across hypoglycemic, euglycemic, and hyperglycemic ranges (p = 0.843) or between patients with and without diabetes (p = 0.401). Subgroup analysis revealed lower MARD in patients receiving intravenous fluid therapy (p < 0.001) and who presented with shock (p < 0.001).

Conclusion: Despite a higher numerical MARD typical of comparisons against POC reference methods, the FreeStyle Libre 2 system demonstrated a high degree of clinical safety (99.03% in acceptable zones) in the IMICU. Its consistent performance in both diabetic and non-diabetic cohorts supports its potential as an adjunctive surveillance tool for detecting glycemic trends in critical patient settings.