Accuracy and performance of continuous glucose monitors in athletes

Two Ipro2 and one Guardian Real-time CGM devices (Medtronic Minimed, Northridge, CA, USA) were inserted into 10 subjects (resting HR < 60 beats per minute (bpm), training 6-15 h per week). For each participant a fasting continuous exercise test was carried out until failure, ∼90 min, and glucoses boluses were given at 30 min (0.5 g/kg) and failure (1 g/kg). Reference BG measurements were taken every 10 min for the first 60 min, every 5 min until failure + 30 min and every 10 min until failure + 60 min with an Abbott Optimum Xceed glucometer. Pre-glucose bolus, all sensors perform better compared to results seen in diabetic cohorts with median mean absolute relative difference (MARD) of 9.7%, 9.6% and 11.1% for the two Ipro2's and the Real-time, respectively. However, there is increased error post-bolus likely due to the gradient of BG change being higher, so the delay in transport to interstitial fluid and sensor results in a larger discrepancy from reference values. CGM devices agree very well with each other during rigorous exercise with median cross-correlation coefficients between 0.88 and 0.97 for the different sensor pairings. This good correlation between all three signals suggests the error between glucose measured by CGM and from blood is not random, but likely due to transport/uptake effects. As the interstitial fluid is the medium from which glucose enters muscle cells, this CGM value might be more useful than BG in determining glucose availability for athletes.