# Accuracy of a novel calibratable real-time continuous glucose monitoring device based on FreeStyle libre in- and out-of-hospital

## TL;DR

A new real-time glucose monitoring device, QT AIR, was developed and shown to be more accurate than existing systems when calibrated.

## Contribution

The novel QT AIR device, based on FreeStyle Libre, introduces real-time monitoring with improved accuracy through calibration.

## Key findings

- Calibrated QT AIR showed significantly lower MARD (12.39%) compared to uncalibrated QT AIR (20.63%) and FreeStyle Libre (18.33%).
- In-hospital calibration reduced QT AIR's MARD to 7.24%, with 95% of values in the safe Consensus Error Grid Zone A.
- Calibrated QT AIR had lower bias and variability compared to FreeStyle Libre according to CG-DIVA analysis.

## Abstract

Based on FreeStyle Libre, we designed QT AIR, an advanced real-time, calibrated Continuous Glucose Monitoring (CGM) device. This study aim to validate the consistency and clinical accuracy of the product by comparing the capillary blood glucose (CBG) with CGM data in both in-hospital and outpatient scenarios.

Results of CGM devices were compared with random capillary glucose values from users in both in-hospital and outpatient settings. The accuracy of CGMs was assessed through consistency analysis, Bland-Altman analysis, calculation of MARD and MAD, Consensus Error Grids, as well as analysis using the Continuous Glucose Deviation Interval and Variability Analysis (CG-DIVA).

In outpatient setting, 1907 values from 138 users were analyzed. FreeStyle Libre data, QT AIR calibrated and uncalibrated data showed strong positive correlations with capillary blood glucose values. The MARD values for the FreeStyle Libre, uncalibrated QT AIR, and calibrated QT AIR groups were 18.33%, 20.63%, and 12.39%, respectively. Consensus Error Grid, reference values in Zone A: FreeStyle Libre: 69.75%, QT AIR uncalibrated: 67.80%, QT AIR calibrated: 87.62%. The Bland-Altman analysis results suggest that FreeStyle Libre exhibitsed a systematic underestimation of blood glucose levels, while QT AIR almost rectified the differences. In the in-Hospital setting, the MARD of QT AIR after calibration was reduced to 7.24%. The Consensus error grid analyses of the in-Hospital data revealed that 95% of the calibrated QT AIR values fell within Zone A, a significantly higher proportion than that of other two group. The CG-DIVA analysis of the calibrated QT AIR device showed a median bias of -0.49% and a between-sensor variability of 26.65%, both of which are significantly lower than the corresponding values observed for the FreeStyle Libre device.

We successfully transformed a retrospective CGM system into a real-time monitoring device. The monitoring accuracy of the device could be improved by calibration.