CGM for Runners: What Glucose Sensors Really Tell You

A small sensor on the upper arm, a graph on the phone, and suddenly every runner can see what their body does with breakfast during a long run. Continuous glucose monitoring (CGM) has gone from diabetes technology to a trend tool in endurance sport within a few years — standard in pro teams, only just arriving in the recreational field. The decisive question is not whether the technology works. It does. The question is what you can actually do with the numbers without getting lost in a curve that says more about your last gel timing than about your fitness.

What a CGM sensor measures — and what it does not

A CGM sensor is a small filament inserted a few millimetres under the skin, measuring in the tissue. More precisely: in the interstitial fluid between the cells. That is not the same as blood. Glucose diffuses from the capillary into this tissue, and that transport takes time. In practice this means a delay of roughly five to fifteen minutes between what happens in the blood and what the sensor shows.

For an easy long run that is irrelevant. For an interval session with fast changes it is decisive: if you take a gel the moment you see a drop, you are reacting to a value from ten minutes ago — and possibly eating against a trend that is already heading up again. This is exactly where the rollercoaster curves come from that many beginners stumble over: they treat a delayed value as a current one, reach for sugar too early and provoke the next overshoot.

Accuracy is reported as MARD, the mean absolute relative difference against a blood reference. Current systems sit at eight to nine percent — a good figure, but it means that at a displayed 100 mg/dl the real value may be several points off. Two further quirks come on top. First, the compression effect: if you lie on the sensor at night, perfusion in the tissue drops and the value appears to fall without anything having happened. Second, post-meal inertia: the rise in tissue comes later and flatter than in blood.

Anyone who knows these three effects — delay, measurement scatter, compression — can work with the data. Anyone who does not will mistake their metabolism for clockwork, which it is not.

Why glucose became a topic in endurance sport at all

In efforts beyond two hours, carbohydrate availability is the most important limiting factor — more important than how your lactate threshold happens to look. The body stores glycogen in liver and muscle, roughly 400 to 600 grams in an average endurance athlete. At moderate intensity you burn around 60 to 90 grams of carbohydrate per hour. The arithmetic is unforgiving: after two to three hours the store is empty unless you refuel.

At the same time, fuelling practice at the elite level has shifted markedly upward over the past decade. Where 60 grams per hour was long considered the ceiling the gut can absorb, 90 to 120 grams is now the target for long races — possible because glucose and fructose are taken up through different transporters and therefore add up. Not every stomach tolerates those amounts without training. And this is precisely where CGM becomes interesting for non-diabetics: not as training control, but as feedback on how your body actually handles a specific fuelling strategy.

The second reason for the boom is mundane: sensors have become cheaper, smaller and available without a prescription. What was a medical device ten years ago is now a 14-day disposable for 60 to 70 euros. Add pro teams publishing their data openly, which creates the impression that glucose control is the next level of performance management. Is it? Only partly.

The typical patterns: what healthy runners see on the display

The first aha moment usually arrives on day two or three, and it is sobering: blood glucose reacts far more strongly to what you eat than to what you train. A breakfast of muesli with fruit juice pushes many people to 160 or 180 mg/dl within 40 minutes, followed by a drop below baseline towards the end of the second hour. That reactive dip is why some runners feel sluggish two hours after breakfast even though they have not eaten anything for ages.

The second pattern concerns the sessions themselves. During easy running in zone 1 to 2, glucose stays remarkably stable in most people, often slightly elevated: the liver releases glucose while the muscles take it up, and the balance holds. Above threshold the picture flips. Hard intervals drive the value up sharply in the short term — stress hormones such as adrenaline and cortisol mobilise glucose even without any intake. A reading of 170 mg/dl in the middle of a 1,000 m interval is not a sign of bad nutrition in a healthy fasting runner; it is a normal stress response.

The third pattern is the late crash. After long sessions, especially beyond 90 minutes, a dip into the 65 to 75 mg/dl range often shows up at night or the next morning. The cause is muscle glycogen repletion: the muscle pulls glucose out of the blood to refill its stores and the liver cannot keep up with release. It is physiological and self-limiting, but it feels like exhaustion. Knowing this, you eat a carbohydrate-rich dinner with enough protein instead of panic-sucking glucose tablets.

Who these patterns are genuinely actionable for: people with type 1 diabetes, anyone with conspicuous readings outside training, athletes who want to test race fuelling systematically, and runners with recurring symptoms such as a performance drop despite good form, dizziness or waking at night. For everyone else the insight is real but rarely decisive.

CGM as a fuelling test device: the most practical benefit

If there is one application where CGM delivers a clear advantage in recreational sport, it is testing race nutrition. The standard recommendation of 60 to 90 grams of carbohydrate per hour is a population average. What your gut actually tolerates, whether you do better with 80 grams in liquid form than with 60 grams as gels, and whether your timing works — that can only be established individually, and the glucose trace adds a layer of information that gastrointestinal symptoms alone do not provide.

A concrete approach for a test run of 2.5 to 3 hours: pick a route with an even profile and a pace close to your planned race pace but not above it. Fuel exactly to plan — for example 30 grams of carbohydrate every 30 minutes as gels with water. Record not only the glucose values but above all how you feel, stomach pressure, and whether you are still able to perform at the end. Repeat with a second strategy, say 80 grams per hour via drink and bars. Comparing the two curves shows you which strategy you handle more stably.

What CGM does not deliver here: it does not tell you how much to top up. The glucose value does not reflect glycogen stores. A stable 100 mg/dl after three hours does not mean your muscles are still full — it only means you are currently supplying enough to hold blood glucose while the stores are long since emptying. Mistake that for the same thing and you still hit the wall at kilometre 32. Details on this in Hitting the wall in the marathon and Fuelling for the marathon.

What studies actually show

The evidence is more sober than the marketing tone suggests. For people with type 1 diabetes the benefit of CGM is well established: fewer hypoglycaemic events, better time in range, demonstrated in randomised trials and anchored in guidelines. For type 2 diabetes there is a similar if smaller effect, mainly with intensified therapy.

For healthy, non-diabetic athletes there is to date no study demonstrating a performance advantage from CGM use. What exists are observational studies showing that continuous glucose monitoring changes fuelling behaviour: people who see their values eat more regularly and earlier during long sessions instead of reacting only when symptoms appear. That indirect effect — better timing, fewer glucose swings — is plausible and often the real gain in practice. But it is a behavioural advantage, not a measurement one.

Scepticism is also warranted for the popular claims migrating from the CGM world into sport. That a flat glucose curve automatically means "metabolic health" is not established. That post-meal spikes are inherently harmful holds for diabetics, not for metabolically healthy people — healthy humans have produced these spikes forever and reliably bring them down within two hours. And the idea that CGM can identify the training zone where you burn maximum fat does not survive scrutiny: blood glucose says nothing about substrate use in the muscle.

The limits of the technology in training

Three limitations you should know before rebuilding your training around glucose values. First: the sensor is not a training control instrument. The established measures for load management — heart rate, pace, power, RPE, the ACWR from your load history — capture what training does to the body. Glucose captures what metabolism is negotiating at that moment. Those are two different questions.

Second: the within-person scatter is larger than the smooth graph suggests. The same run with the same fuelling can produce different curves on two days — depending on sleep, the previous day's load, cycle phase, outside temperature and carbohydrate availability the evening before. Anyone deriving a rule from a single curve is building on sand.

Third: sensor placement. On the upper arm, where most sensors sit, runners see more disturbance than sedentary people — from sweat, movement and chafing against clothing. For triathletes in a wetsuit and runners in tight shirts, detachment is a realistic problem. An IPX8 water resistance rating covers swimming, not two hours of friction under a tight shirt.

When CGM makes sense for runners — and when it does not

A pragmatic grid that has proven itself in practice:

• Worth it: you have type 1 diabetes. Then CGM is not an optional tool but the basis for how you coordinate training and insulin dose. Fundamentals in Diabetes and endurance sport.
• Worth it: you are preparing for a marathon, a long-distance triathlon or a stage race and want to test your fuelling systematically instead of trying it for the first time in the race.
• Worth it: you have had repeated performance drops or symptoms that fit hypoglycaemia and want to know whether your fuelling is too thin.
• Conditionally worth it: you want to understand how your daily routine and meal timing interact with training. Two to four weeks is plenty for that.
• Not worth it: you are looking for a training control device. Pace, heart rate, power and RPE are the right tools — a glucose value replaces none of them.
• Not worth it: you want to measure your aerobic fitness. That is what lactate tests, step tests and field tests are for.

Cost belongs to the honest picture too: 60 to 70 euros per 14-day sensor adds up to more than 1,500 euros a year with continuous use. For healthy recreational athletes that is not a sensible priority — measured in benefit per euro, a second strength session per week is far ahead, as is an extra rest day or an eight-hour sleep window. If you have the money and the curiosity, test for two to four weeks and then make a deliberate decision.

Common mistakes when starting out

• Treating the curve as the goal. Anyone chasing a "flat" trace eats small amounts constantly and never trains the ability to run steadily at lower values. The body is not a control loop you set to a target value.
• Eating against the delay. A falling value on the display does not automatically justify immediate sugar. Wait two readings and look at the slope, not the point.
• Steering fasted training by the curve. In fasted running the value drops physiologically — that is the point of the session, not a reason to abort. Act only on symptoms (dizziness, shakiness, loss of concentration).
• Interpreting at night without knowing the compression effect. A night-time 65 mg/dl that returns to 90 within 15 minutes of turning over was not a dip — it was a pressure artefact.
• Reading single outliers as patterns. Without several repetitions of the same situation under the same conditions, every "insight" is coincidence.
• Using the technology without medical context. Conspicuous values outside sport belong with a doctor, not in a training forum.

How Peakora handles metabolic data

Peakora steers your training with what actually represents load: TSS from the Training Stress Score, CTL, ATL and TSB as the form curve, plus heart rate and pace zones derived from your own data. When you test fuelling, that feeds into the plan — as a note attached to the session, as a deliberate test run scheduled in the plan instead of a random experience in the race. What Peakora does not do: steer training by glucose values. That would be false precision that physiology does not support. The terms around load management are collected in the glossary, and the framework for race preparation on the marathon pillar page.

Bottom line: a good test device, not a training device

CGM is honest technology with a clear use case and clear limits. The use case is fuelling: finding out which carbohydrate amount, which form and which timing your gut tolerates under race load. The limit is training control: glucose has no place there, because it does not represent the quantities that matter — load, recovery, adaptation.

Approach it with that expectation and 60 to 70 euros buys you a fairly precise picture of how your body handles race fuelling. Expect more and you buy an expensive curve plus a new field of worry. The decision is yours — but it should be an informed one.

Frequently asked questions about CGM for runners

What does a CGM sensor measure while running?

A CGM sensor measures glucose in the interstitial fluid beneath the skin, not in the blood. The displayed value therefore lags blood glucose by roughly five to fifteen minutes — relevant during fast pace changes, irrelevant for the trend across an hour.

Do healthy runners need a glucose sensor?

No — for training control a CGM is not necessary in healthy runners. The benefit comes mainly from testing race fuelling for long runs and races, because you can see how your body handles 60 to 90 grams of carbohydrate per hour.

How accurate are CGM sensors?

Current systems reach a mean absolute relative difference of about eight to nine percent against a blood reference. Accuracy is good in the normal range and worse during rapid changes and at night when you lie on the sensor. Check against a finger prick whenever the reading does not match how you feel.

How much does a CGM sensor cost?

A 14-day sensor costs roughly 60 to 70 euros in Austria depending on the system, and healthy recreational athletes pay for it themselves. A limited test period of two to four weeks makes sense; continuous year-round use does not.