Testing & Measurement · Updated September 2026
A CGM is a genuinely transformative device for people with diabetes. For people with normal glucose regulation it is an interesting and expensive source of data whose value has not been demonstrated in outcome trials.
A CGM is a filament inserted into subcutaneous tissue that measures glucose in interstitial fluid, not blood, and reports a reading every few minutes. Interstitial glucose tracks blood glucose closely at steady state and lags it by roughly 10 to 15 minutes when glucose is changing quickly. That lag is not an error; it is the physiology of diffusion, and it means a reading during a sharp rise systematically understates the concurrent blood value.
Accuracy is usually reported as MARD — mean absolute relative difference against a laboratory reference. Modern consumer sensors report MARD figures in the single digits to low teens, which is clinically excellent for managing diabetes. It is also loose enough that a difference of a few mg/dL between two readings, or between two foods, is inside the noise.
The honest case is educational rather than clinical, and it is not trivial. A CGM converts abstract nutritional advice into personal, immediate feedback: you learn which of your habitual meals produce large excursions, how much a walk after eating flattens a curve, how badly a poor night's sleep degrades the next day's response, and how your particular body handles specific carbohydrate sources.
Several of those are genuinely useful and several are individual enough that generic advice cannot supply them. People who wear one for a month often make durable dietary changes, and that is a real outcome even though it was produced by education rather than by the device itself.
No large trial has demonstrated that CGM use in people without diabetes improves clinical outcomes — not cardiovascular events, not diabetes incidence, not mortality.
Glucose variability in healthy people is normal. Post-meal excursions are the system working, and interpreting every peak as damage is a misreading of the physiology.
The reference ranges displayed by consumer apps are frequently tighter than clinical normal ranges, which manufactures alarm out of ordinary readings.
There is a documented pattern of anxiety and food-avoidance developing in metabolically healthy people who wear one continuously, which is a real harm and not a rare one.
Sensor-to-sensor variation means two sensors worn simultaneously can disagree by more than the difference you are trying to detect between two breakfasts.
The most common analytical error is treating a single sensor reading as a precise measurement. If you are comparing oats against eggs on the strength of one morning each, you are reading noise. Repeated measurement across several days is the only way to extract a real signal, and almost nobody does that.
| Question | Can a CGM answer it? |
|---|---|
| Does this specific meal spike me more than that one? | Yes, if you repeat each several times — not from a single comparison |
| Does walking after dinner flatten my curve? | Yes — a large, repeatable, well-documented effect |
| Does poor sleep worsen my glucose response? | Yes — also large and consistent enough to see |
| Do I have undiagnosed diabetes or prediabetes? | Not properly — that is HbA1c and a fasting or tolerance test, ordered by a doctor |
| Is my metabolic health improving over months? | Weakly — HbA1c and lipids are the established markers |
| Should I be worried about a 160 mg/dL post-meal peak? | Almost certainly not, in someone with normal HbA1c |
The reader or app is not the expense. Sensors last 10 to 15 days and are consumables, so continuous wear means roughly two to three sensors a month, indefinitely. Over a year this is a subscription in all but name, and it is frequently several times the headline price of a starter kit.
This is the strongest practical argument for the approach most sensible users converge on: wear one for two to four weeks, learn what your body does, make the dietary changes it reveals, and stop. The educational value is front-loaded, and the marginal information from month six is close to zero for someone with normal regulation.
Regulatory status varies by country and has been changing. Some sensors are available over the counter for general wellness use, others require a prescription, and several of the popular consumer “metabolic health” services are a subscription wrapper around a prescription sensor with an app and coaching attached. Check what you are actually buying, because the software layer is frequently where the margin sits and the underlying hardware may be available more cheaply.
Change one variable at a time and repeat it. Eat the same breakfast three mornings running and look at the average shape, not one curve. Then test the intervention — a 15-minute walk afterwards, or the same meal with protein added first — three more times. This is tedious and it is the difference between learning something about yourself and generating an attractive graph.
Log context. Sleep, alcohol, illness, menstrual cycle phase, stress and caffeine all move glucose curves substantially, and a reading without context is uninterpretable. Most people who conclude a food is a problem have actually detected a bad night's sleep.
See a doctor and get HbA1c, fasting glucose and a lipid panel. Those are the validated, inexpensive tests that clinical decisions are actually made on, and a CGM is not a substitute for any of them. A consumer sensor that suggests something is wrong should send you to those tests rather than replacing them; a consumer sensor that suggests everything is fine is not sufficient reassurance either.
People with diabetes, for whom this technology is transformative and not in question. People with diagnosed prediabetes making a serious dietary change, where the feedback loop can drive adherence. Athletes managing fuelling around long-duration events. And people who are simply curious and can afford it, provided they go in knowing they are buying education rather than a health outcome, and knowing that a subset of users develop genuine anxiety around eating.
Accuracy stated as mean absolute relative difference from a laboratory standard, not a vague “clinical grade” claim.
The recurring sensor price, not the starter kit price, is what this costs. Any service that obscures the per-month sensor cost is worth scrutinising.
Apps that flag ordinary post-meal excursions as problems are manufacturing alarm from normal physiology.
The ability to export your own readings, rather than only seeing them through a scoring layer that may be tuned to keep you subscribed.
Ranked by how much the result can inform a decision. The cheapest item on this list is the one a doctor would order first, which is worth stating plainly in a category priced like a subscription.
The test clinical decisions are actually made on, and a fraction of a month of CGM sensors. HbA1c reflects average glycaemia over roughly three months and is the established marker for diagnosing and tracking prediabetes and diabetes. If the question you want answered is whether your glucose regulation is a problem, this answers it and a CGM does not.
View on AmazonThe mainstream consumer sensor: a filament in interstitial fluid reporting every few minutes to a phone. Genuinely useful for two to four weeks of learning — which meals move you, what a post-meal walk does, how sleep changes the next day. Treat it as an education purchase with a defined end date rather than an ongoing subscription, because the marginal information after the first month is small.
View on AmazonMeasures blood directly rather than interstitial fluid, so there is no lag, and strips cost pennies. For the specific question “does this meal spike me?” a fasting reading plus one at 60 and 120 minutes answers it for a tiny fraction of CGM cost. What you lose is the overnight and between-meal picture, which is genuinely the CGM's advantage.
View on AmazonSensors are the recurring cost, and one lost to a shower, a shirt sleeve or a gym session part-way through its wear period wastes the largest single line in this budget. An overpatch costs almost nothing and reliably gets a sensor through its full period. Unglamorous, and it materially changes the cost per day of data.
View on AmazonGlucose is one input among several, and a panel covering lipids, liver enzymes and kidney function gives a far more complete picture of metabolic health than continuous glucose data does on its own. If the underlying motivation for buying a CGM is “am I metabolically healthy,” this panel plus HbA1c answers that question better and more cheaply.
View on AmazonIt can be genuinely educational for a few weeks — you learn which meals move you, how much a post-meal walk helps, and how sleep affects the next day. What has not been shown is any clinical benefit: no large trial demonstrates improved outcomes from CGM use in people without diabetes. Wear one to learn, with a planned end date, and be aware that a subset of users develop real anxiety around eating.
In someone with normal HbA1c, a post-meal rise is the system working, not evidence of damage. Consumer apps often display reference ranges tighter than clinical normal ranges, which turns ordinary physiology into an alert. The informative signal is the pattern over weeks — how high, how long elevated, and whether it is improving — not any single peak.
Modern sensors report MARD figures in the single digits to low teens against laboratory reference, which is excellent for managing diabetes. It is still loose enough that small differences between two readings, or between two foods measured once each, sit inside the noise. Two sensors worn simultaneously can disagree by more than the difference many people are trying to detect.
A CGM measures glucose in the fluid between cells, and glucose takes roughly 10 to 15 minutes to diffuse there from blood. At steady state the two track closely; when glucose is changing fast, the sensor reads behind. So a peak appears later and lower than the true blood peak, which matters if you are comparing sharp responses between foods.
The starter kit price is misleading. Sensors last 10 to 15 days and are consumables, so continuous wear means roughly two to three sensors a month indefinitely, which over a year is a subscription in all but name and typically several times the headline figure. This is the strongest practical argument for wearing one for a defined learning period rather than permanently.
Take the 7-question quiz to get your personalized supplement and lifestyle stack — tailored to your age, fitness level, and longevity concerns.
Take the QuizA 28-day fillable sleep log built around the seven levers that actually move sleep, so you test one per week and see what it did.