Glucose management indicator: Do we need device-specific equations?

Glucose management indicator (GMI) may not perform equally well for different continuous glucose monitoring (CGM) systems. Thus, we aimed to develop device-specific GMI for the Guardian 3 and 4 sensors, compare them to the original GMI, and investigate the association between the glycaemic gap (=HbA1c-GMI) and HbA1c.

In a single-centre, observational study of adult type 1 diabetes patients using Guardian Sensor 3 (G3) and 4 (G4) CGM devices, we estimated HbA1c using CGM-derived mean glucose for both CGMs using linear mixed models. We compared the estimates and their residuals (G3-gap and G4-gap) to the original GMI and its residuals (Bergenstal-gap) using regression and Bland-Altman plots.

We included 120 adult type 1 diabetes patients (90 with G3 and 30 with G4) and 194 measurement points. We found that for G3 and G4 sensors, GMI significantly underestimated glycaemia in high HbA1c ranges, reaching the clinically significant 0.5 %[4.4 mmol/mol] difference at 7.6 % [60 mmol/mol] for G3 and 8.3 % [67 mmol/mol] for G4 sensors. For G4, GMI significantly overestimated glycaemia in the lower HbA1c range. We found a strong relationship between all 3 gaps and HbA1c, and the slope was steeper for the Bergenstal-gap versus the sensor-specific G3 and G4 gaps. The G3-gap was approximately half as large as the Bergenstal-gap for HbA1c > 7 % [53 mmol/mol], and the G4-gap is approximately half of the Bergenstal-gap for the whole HbA1c range.

Device-specific GMI equations could reduce the risk of clinically significant under- and overestimation of HbA1c, improving clinical decision-making.