Accuracy and reliability of a continuous glucose monitoring system with a focus on hypoglycaemia

Glucose monitoring is increasingly based on continuous glucose monitoring (CGM) systems. However, data on the accuracy of CGM within the hypoglycaemic range is sparse. This study investigated CGM accuracy within the hypoglycaemic range in three different clinical settings of hypoglycaemia.

Ninety-two people with various causes of hypoglycaemia were analysed: (i) people during insulin tolerance testing (ITT) (n = 63); (ii) people with insulinoma (n = 16); and (iii) people with diabetes receiving subcutaneous insulin therapy (n = 13). CGM accuracy was evaluated for subgroups and different glucose rates of change (RoC) using mean absolute relative difference (MARD), percentage of glucose values within ±20 mg/dL of point-of-care glucose (%20/20), diabetes technology society error grid and Bland-Altman analysis (BAA).

Four hundred sixty-four CGM/POC glucose pairs were obtained (39.7% level 1, 60.3% level 2 hypoglycaemia). CGM accuracy decreased from people with diabetes receiving subcutaneous insulin therapy (MARD: 13.9%; %20/20: 81.5%) to those with ITT (MARD: 50.8%; p < 0.01, %20/20: 67.5%; p .02). CGM accuracy significantly decreased with higher RoC. Proportion of CGM values with moderate risk in failing to detect potentially dangerous hypoglycaemia increased from people with diabetes (1.5%) and insulinoma (1.2%) to people with ITT (14.2%, p < 0.01). BAA revealed a significantly increasing bias of -3.3 + 10.7 mg/dL in people with diabetes receiving insulin therapy to -15.2 ± 13.6 mg/dL in people with ITT (p < 0.01).

CGM accuracy can vary in different clinical hypoglycaemia scenarios. While it remains acceptable in people with diabetes receiving subcutaneous insulin therapy and people with insulinoma, it appears to be inaccurate for glucose monitoring during ITT.