Continuous Glucose Monitoring Metrics Are Failing People With Diabetes Mellitus

Continuous glucose monitoring (CGM) systems have transformed the management of diabetes mellitus (DM) and now serve as a cornerstone of modern diabetes care. For individuals with type 1 diabetes (T1D) using automated insulin delivery (AID) systems, CGM is not merely a helpful tool—it is the foundational data source that drives real-time insulin dosing decisions. Every few minutes, algorithms process interstitial glucose readings from sensors to dynamically adjust insulin delivery. As such, the quality of the glucose signal has direct and immediate therapeutic implications. Despite this, the most widely used metric for assessing CGM performance—the Mean Absolute Relative Difference (MARD)—has become inadequate for current clinical demands. MARD reflects the average deviation of CGM readings from a reference method (venous or capillary) under ideal conditions. While it is simple, easy to communicate, and historically valuable, it overlooks the clinical realities that matter most: when and how sensor errors occur, and what real-world consequences they produce. A CGM system with a “good” MARD can still be affected by factors such as study design and methodology, duration of sensor wear, placement site, electrochemical interference, and lag in glucose detection—all of which can significantly impact safety and effectiveness.