CGM in healthy athletes: what a glucose sensor really shows, and what we made up

Three elite athletes. A world record in a relay ultra-race, a record everesting, and a 93-meter breath-hold dive. All three had a continuous glucose monitoring (CGM) sensor on the arm. None of them has diabetes. And not one of them was “in range” by clinical-table standards for the entire event.

What the sensors show at real events

A case series published in Sensors in 2026 describes these three events — and three completely different glucose patterns.

Race Across the West: 44 hours 20 minutes, 1329 km, 12,700 m of climbing, a two-person relay with 20–40 minute pulls. Average glucose in the saddle — 5.1 ± 1.3 mmol/L, at rest — 6.4 ± 1.4. Overall range across the race: from 3.0 to 9.4 mmol/L. The athlete spent 9.15% of his time on the bike below 3.9 mmol/L versus 1.23% during rest. And that is despite eating: 2591 g of carbohydrate, averaging 58 g/h (92 g/h on day one, 53 on day two, 31 on day three). Fueling happened only between pulls — meaning every return to the course after eating reproduced the classic reactive hypoglycemia scenario.

Everesting: 6 hours 40 minutes of continuous work, 76 climb repeats at an average gradient of 14.2%, 8848 m of vertical, average power 296 W. Glucose — 8.9 ± 0.3 mmol/L, coefficient of variation 3.5%. Practically a straight line. Except that line ran above 7.8 mmol/L exactly 100% of the time. Carbohydrate — 149 g/h, noticeably above the standard recommendation of 90 g/h.

Freediving: 93 meters, 3 minutes 55 seconds. After surfacing, glucose held at 10.4 ± 1.0 mmol/L and did not drop below 11.1 mmol/L for about 50 minutes — even though the athlete had not eaten for 12 hours. Lactate meanwhile shot up from 0.7 to 13.4 mmol/L.

The authors' conclusion: deviations from “euglycemia” in elite sport are commonplace and reflect an adaptive response to extreme stress, not broken regulation.

What CGM cannot do

A review in Performance Nutrition (2025) explains why you cannot build a nutrition strategy out of these graphs.

The sensor does not measure blood. CGM reads glucose in interstitial fluid. There is a 5–10 minute lag between it and blood, and that lag grows precisely when glucose is changing fast — that is, during intense work and after eating. Accuracy drops at high intensity and at values below 4.0 mmol/L. MARD is 9–14% in the clinic and <15% under load, and the limits of agreement with lab analysis are roughly ±1.5 mmol/L. One and a half units is the difference between “everything is fine” and “I am falling apart.”

The sensor does not see fuel. Blood glucose is only 20–30% of available carbohydrate sources. Muscle glycogen, the main store, physically never leaves the muscle and never enters the blood. CGM measures neither glycogen nor the flow of carbohydrate from the gut. A well-fueled athlete holds normal glucose thanks to the liver — meaning a flat line says nothing about how much fuel is left.

No link to performance was found. There is no causal evidence that a particular glucose level or its stability affects performance. Studies have shown no association between glycemic metrics and finishing time.

How to use it sensibly

It works for finding patterns in training. How you respond to a specific breakfast 30, 60 and 90 minutes before the start. Whether you crash in the first ten minutes after a pre-race gel. What coffee on an empty stomach does. Where the slide begins on a three-hour long ride. These are hypotheses, later checked against how you feel and your pace — not the other way round.

It does not work for making decisions from a number during a race. Between reality and the screen sit the lag, the error, and the fact that the number is not about fuel. Take the gel on plan, not on the sensor.

Myths worth burying:

“Flat sugar = good shape.” The everesting record holder had 3.5% variability — with glucose at 8.9 mmol/L for all six and a half hours. Flat and high.
“CGM will tell you when it is time to eat.” It will not: by the time glucose starts heading down, the glycogen is already spent, and the sensor lags behind.

That is why a carbohydrate strategy is still calculated in grams per hour — based on duration, intensity and how well your gut is trained, not on sensor readings. The 60–90 g/h guideline is about throughput, which the sensor fundamentally does not measure.

Limitations

This is a series of three cases with no hormone control, no glycogen measurement and no synchronization of glucose with power. It cannot be generalized. The devices themselves are approved for diabetes management — use in healthy people is off-label. Add compression artifacts during sleep, loss of connection with the phone (in the everesting the interval dropped from 1 to 15 minutes) and a high risk of misinterpretation — up to unnecessary dietary restrictions because of a normal post-meal peak.

Separately: the sensor is not a diagnostician. But if outside training you consistently see fasting or post-meal values that worry you, that is a reason to get proper lab tests from a doctor, not to build theories from a graph.

Key points

At real events, glucose in healthy elite athletes ranged from 3.0 to 11+ mmol/L — and that is a response to load, not a pathology.
The pattern depends on the type of load: a broken-up relay produced crashes, a continuous climb produced stable hyperglycemia, apnea produced a delayed rise after surfacing.
CGM measures interstitial fluid with a 5–10 minute lag, and accuracy drops precisely during intense work.
The sensor does not see muscle glycogen or carbohydrate flow — it is not a “fuel gauge.”
The link between glucose level or stability and performance has not been causally proven.
Useful as a reconnaissance tool in training, useless and harmful as an advisor in a race.
Count fueling in g/h to plan; worrying numbers outside training go to a doctor, not to a forum.

Sources: Sensors, 2026 — Beyond Euglycemia: Case Studies Using Continuous Glucose Monitoring in Elite Athletes Without Diabetes During Record Athletic Events; Performance Nutrition, 2025 — Application potential of continuous glucose monitoring (CGM) in elite endurance athletes without diabetes. https://doi.org/10.3390/s26051624 · https://doi.org/10.1186/s44410-025-00013-7