Do Continuous Glucose Monitors Overestimate Your Blood Sugar Spikes?

A continuous glucose monitor and a finger-prick test were fed the same commercial fruit smoothie in a randomized crossover trial out of the University of Bath, and they disagreed on what the food was. The Abbott FreeStyle Libre 2 put the smoothie's glycemic index at 69, in the medium-to-high band. Capillary blood, the reference method, put it at 53, which is low (see [1]). Whole fruit got the same treatment: low-GI by fingerstick, medium or high by sensor. Across the protocol the monitor overestimated time spent above 7.8 mmol/L by roughly 3.8-fold, and still by about 2-fold after correcting for a baseline offset (see [1]). Fifteen healthy adults, one device, a consistent direction of error: up.

The mechanism is not a defect, it is the measurement site. A CGM reads glucose in the interstitial fluid around your cells, not in blood, so its numbers lag and drift from the venous value. In this trial the sensor ran about 0.9 mmol/L high both fasting and after meals (see [1]). That bias sits on top of the noise floor these devices already carry in people without diabetes. In normo-glycemic adults the mean absolute relative difference between sensor and venous glucose has been measured at 17.6 percent, with a per-person correlation of only 0.68 (see [2]). Seventeen percent error is wider than most of the food-to-food differences an optimizer is trying to detect.

That is the protocol problem. If you rank two meals by a 15-point spike and the instrument's error band is wider than 15 points, you are tuning to noise, not signal. The device is not useless, but its valid output is a within-day trend on the same body under the same sensor, not an absolute verdict that a banana is a high-GI food. If you want to know whether a specific meal moves you, the resolving measurement is paired capillary testing, not a prettier sensor trace. In Wellness Project I would hold any single reading loosely and weight the multi-week pattern in your logs instead.

The trial authors were direct: CGM is not a valid way to classify a food as high or low GI, and capillary sampling should be prioritized when the number has to be right (see [1]).

This is general information, not medical advice. Talk to a qualified clinician about your own situation.