The short version
- In 153 healthy non-diabetic people aged 7 to 80 wearing a blinded Dexcom G6, mean average glucose was 98 to 99 mg/dL and median time in the 70-140 mg/dL range was 96%.
- Those healthy participants spent a median of 2.1% of the day — about 30 minutes — above 140 mg/dL, and 1.1% of the day below 70 mg/dL.
- A 2018 Stanford study found that people classed as normoglycaemic by standard measures reached prediabetic glucose ranges 15% of the time and diabetic ranges 2% of the time.
- A 2015 study of 800 people and 46,898 meals found high variability in glucose response to identical meals between individuals, which is the strongest published case for personal measurement.
- None of these studies tested whether reducing post-meal glucose peaks in a person with normal HbA1c changes any later health outcome.
Somebody straps on a sensor, eats a banana, watches the line go to 150, and concludes that bananas are a problem. That sequence happens thousands of times a day, and almost none of it survives contact with the reference data.
What normal actually looks like
In 2019, the T1D Exchange network published exactly the study this question needed. 153 healthy, non-diabetic, non-obese people aged 7 to 80 wore a blinded Dexcom G6 for up to ten days. Blinded matters — they could not see the numbers, so they could not change their behaviour in response to them.
Mean average glucose was 98 to 99 mg/dL in every age group except the over-sixties, where it was 104. Median time between 70 and 140 mg/dL was 96%, interquartile range 93 to 98.
Which means: median time above 140 mg/dL was 2.1% of the day — about 30 minutes. Median time below 70 was 1.1%, about 15 minutes. Mean within-person coefficient of variation was 17%.
So a healthy person spends half an hour a day above the line that consumer apps colour red, and a quarter of an hour in what an app would call a low. That is not dysfunction. That is the reference range for people with no metabolic disease at all.
The paper that gets quoted the other way
The Stanford glucotypes study from 2018 is the one the CGM industry cites, and it found something genuinely interesting: people classed as normoglycaemic by standard testing reached prediabetic glucose ranges 15% of the time and diabetic ranges 2% of the time, and they clustered into distinguishable patterns of variability.
That is a real finding about population heterogeneity, and it is a good argument that a single fasting draw is a crude instrument. What it is not is evidence that your individual peaks are hurting you. The study characterised patterns. It did not follow those people for a decade to see who developed anything, and it did not test whether flattening a curve changes an outcome.
That distinction — describing variation versus demonstrating consequence — is where almost all CGM marketing lives.
The one thing a CGM is genuinely good for
Personal variability, and it is well documented. The 2015 Cell study monitored 800 people for a week, capturing responses to 46,898 meals, and found high variability in the response to identical meals between individuals. A machine-learning model built on blood parameters, dietary habits, anthropometrics, activity and gut microbiota predicted personal responses, was validated in a separate 100-person cohort, and a blinded randomised dietary intervention based on it produced lower post-meal responses.
So the personalisation premise holds up: your response to a given food is genuinely yours, and universal food rules have limited utility. If you want to know how you handle oats versus toast, a sensor will tell you and nothing else will.
What no study here shows is that a person with a normal HbA1c who successfully flattens their post-oat curve is better off in any way that shows up later. That trial has not been reported, and until it is, the honest framing is that a CGM measures something real whose significance in healthy people is unproven.
The failure mode we see most
People start eliminating foods. Fruit goes first, then oats, then anything with a starch, and within a month a reasonably varied diet has become five foods chosen for the shape of a line on a phone. The person is not healthier. They are more anxious and eating less fibre.
If you are going to wear one, set the terms first. Two weeks, not two years. A handful of specific questions — does eating protein first change my lunch curve, does a walk after dinner change my evening — rather than continuous surveillance. And then take the sensor off and keep whichever two habits earned their place.
The same discipline applies to whatever you try alongside it. Cinnamon's fasting-glucose effect vanishes between one review and the next. Chromium does nothing measurable in people without diabetes. Time-restricted eating flattens overnight glucose largely because you did not eat. A sensor will make all three of those look more impressive than the pooled trials say they are, because a sensor shows you the acute response and none of the follow-through.
What we'd actually tell you
If your HbA1c and fasting glucose are normal, a CGM is a curiosity purchase, and a legitimate one. Buy it as an experiment, not as a monitoring system, and do not let it recruit you into fearing food.
If they are not normal, this is a clinical situation and a sensor is not the answer to it. Diabetes and prediabetes are diagnoses, and your doctor has both better tests and better tools. Berberine is the one supplement in this space whose trials produced numbers a lab panel can actually check, and even there the trials were run in people already diagnosed with something, the absorption is poor, and it belongs in a conversation with your doctor before it belongs in your cart — particularly if you take anything else that moves glucose.
Good questions
Should I wear a CGM if I do not have diabetes?
As a short experiment, it can be genuinely informative, because your response to a given meal really is individual. As ongoing monitoring, we would say no. Healthy people spend about half an hour a day above 140 mg/dL, so a device that flags every excursion will show you a lot of events with no demonstrated significance.
Is a glucose spike after eating bad for me?
In a person with normal HbA1c, nobody has shown that it is. Reference data from blinded sensors show healthy people routinely go above 140 mg/dL after meals and spend a median of 30 minutes a day there. Studies have characterised who spikes and how much; none of them followed people to see whether flattening those peaks changes anything.
Why does my CGM show different numbers than my finger stick?
Sensors read interstitial fluid rather than blood, so they lag behind and sit alongside rather than on top of a capillary reading. That gap is normal and expected. It also means the precise height of a peak is less reliable than the shape and timing of it, which is another reason not to make food decisions on individual readings.
Will a CGM help me lose weight?
There is no good evidence that it will, and it is not what these devices were built for. What it may do is make you notice what and when you eat, and attention to eating is genuinely useful. If that is the mechanism you are buying, a food log costs nothing and does the same job without teaching you to fear fruit.
Does berberine flatten glucose spikes on a CGM?
It may show up acutely, but that is a weak reason to buy it. Berberine's trial evidence comes from people already diagnosed with type 2 diabetes, hyperlipidaemia or metabolic syndrome, its oral absorption is poor, and it belongs in a discussion with your doctor if you take anything else affecting blood glucose. Judge it on a lab panel over three months, not on a sensor over three days.
I cut out fruit because of my CGM. Was that right?
Probably not, and this is the most common way sensors go wrong. Whole fruit comes with fibre, potassium and micronutrients, and a transient rise after eating it is normal physiology in people without diabetes. If a device is steadily shrinking the list of foods you eat, it has stopped giving you information and started giving you rules.
Sources
- Shah VN, DuBose SN, Li Z, et al. Continuous Glucose Monitoring Profiles in Healthy Nondiabetic Participants: A Multicenter Prospective Study. J Clin Endocrinol Metab, 2019. View study
- Hall H, Perelman D, Breschi A, et al. Glucotypes reveal new patterns of glucose dysregulation. PLoS Biol, 2018. View study
- Zeevi D, Korem T, Zmora N, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell, 2015. View study
These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.
