Bauhaus et al. 2023, accuracy of continuous glucose monitoring during exercise

Study in the International Journal of Environmental Research and Public Health comparing a continuous glucose monitor against a laboratory glucose analyser in athletes, at rest and during exercise. Agreement with the lab reference was better at rest (mean absolute relative difference, MARD, about 17%) than during exercise, and the limits of agreement widened substantially during exercise (roughly −58 to +67 mg/dL, versus −13 to +52 mg/dL at rest). During running, MARD was about 22% at moderate intensity (65% of maximum heart rate) and about 18% at high intensity (85%), both well outside the roughly 10–15% usually considered clinically acceptable. The authors concluded that relying on interstitial-glucose data during exercise can lead to carbohydrate-intake mismanagement in healthy athletes. A small sample, but the key validity evidence: the device is least accurate exactly when a runner would want to act on it.