Now that a continuous glucose monitor can be bought without a prescription, a particular claim has spread fast: wear one, take a supplement, watch the curve, and you will see whether it works. It is an appealing idea because it feels like evidence. A flat line after berberine looks like proof. It is not, and understanding why is more useful than any single reading the sensor will give you.
What people think they are doing
The reasoning goes like this. Baseline meal on Monday, same meal on Wednesday with the supplement, compare the two peaks. Lower peak means it worked.
Every step of that is reasonable. The problem is that it is a study design with one participant, no control, no blinding and no repetition, and each of those omissions can produce the result on its own.
Why the comparison does not hold
There is no control condition
To attribute a difference to the supplement, everything else has to be equal. In practice, between Monday and Wednesday you changed your sleep, your stress, your previous meal, your activity in the preceding hours, your hydration, and where in your cycle you are if that applies.
Any of those moves a glucose curve. You cannot hold them still, and you cannot measure most of them.
Your own variation is bigger than you think
The same person eating the same standardised meal on different days produces meaningfully different glucose responses. This is well documented in the personalised nutrition literature, and it is the single most common reason a self-test appears to show an effect.
If your natural day-to-day variation is larger than the effect you are looking for, you will find effects that are not there. You will also miss ones that are.
The sensor itself has error
Consumer CGMs report accuracy as MARD, the average gap from a laboratory measurement. Stelo reports 8.3% and Lingo 9.3%, both manufacturer-measured.
At that accuracy, a displayed 100 mg/dL could reasonably be anywhere from about 91 to 109. Two sensors worn on the same arm at the same time will disagree with each other. Comparing two peaks that differ by ten points is comparing numbers inside the noise.
Regression to the mean
People start testing after a reading that alarmed them. Extreme values tend to be followed by less extreme ones for purely statistical reasons, with or without an intervention. If you begin your supplement the day after your worst curve, the improvement was partly going to happen anyway.
Knowing you are watched changes what you do
This one is rarely admitted. People eat differently while wearing a CGM. Portions shrink, snacks get skipped, walks happen after dinner. Those changes are real and they are good, but they are not the supplement.
What a CGM is genuinely good for
None of this makes the device useless. It makes it useful for different questions than the ones people ask of it.
Finding your own repeatable patterns. If oatmeal reliably sends you higher than eggs across five separate mornings, that is a pattern worth acting on. Repetition is what turns a reading into information.
Catching a surprise. Some people discover that a food they assumed was fine produces a large response. That is worth knowing, and no average from a lab will tell you.
Making the feedback loop immediate. The behaviour change a sensor prompts is arguably its main benefit, and it does not require the data to be precise.
Seeing the effect of something large and obvious. A big change in meal composition, or starting a GLP-1 medication, produces a shift big enough to see through the noise. A supplement usually does not.
What would count as evidence
If you want to know whether a compound affects glucose, the answer comes from randomised, controlled, ideally blinded trials with enough participants to separate signal from variation, and from meta-analyses of those trials.
That is why our own write-ups name the studies. Berberine vs chromium for blood sugar compares what the trials found, rather than what one person’s sensor showed. Is berberine really nature’s Ozempic takes apart a claim that spread on exactly the kind of anecdote this article is about.
If your interest is your own numbers rather than a compound’s general effect, the tool that answers it cheaply is a lab test. An at-home A1C test gives a three-month average for a fraction of a CGM subscription.
Questions people ask
So a CGM is useless for supplements?
Not useless, but it cannot establish that a supplement works. It can show you whether your own glucose pattern changed over weeks, which is a weaker and more honest claim.
What if I repeat the test ten times?
Repetition genuinely helps and is far better than a single comparison. You still have no control condition and no blinding, so you can describe what happened but not why.
Could I see a real effect if it were big enough?
Yes. Large, consistent effects show through noise. Most supplement effects on glucose are small, which is precisely why they need trials to detect.
Which is more accurate, Stelo or Lingo?
Stelo reports 8.3% MARD against Lingo’s 9.3%, both manufacturer-reported. The difference is smaller than the variation between two sensors on the same arm. The full comparison is in Stelo vs Lingo.
Should I still buy one?
Possibly, for pattern-finding rather than compound-testing. Is Stelo worth it works through who genuinely benefits.
The bottom line
A continuous glucose monitor is a good instrument pointed at the wrong question when it is used to test supplements. It measures your glucose accurately enough to show patterns and not precisely enough to isolate causes.
Use it to learn what your own meals do to you. Use published trials to learn what a compound does to people in general. Those are two different questions, and only one of them is answered by a sensor on your arm.
