There’s a white plastic disc on your brother-in-law’s arm at the barbecue, and he’s telling you that grapes spike him. Somebody else at the table has one too. And you’re standing there holding a paper plate, thinking about the lab printout on your kitchen counter — the one where the glucose number was 103, and the note said “watch this,” and nobody explained what watching it would actually involve.
So you look it up, and the internet offers you a $99 sensor, an $89 monthly plan, and a great deal of confident advice about bananas. This issue is about what that sensor can and can’t tell you, and about the question almost nobody asks out loud: not whether to buy one, but how long to keep it. The answer changes what you spend by a factor of ten.
The goal here is the usual one: technology that makes your life better, not more complicated — and the things you might not realize would help.
The short version
What matters: A continuous glucose monitor — a CGM, and that’s the only piece of jargon I’ll ask you to carry this week — is sold as a monitor. For anyone who isn’t taking insulin, it isn’t one. It’s a two-week course in how your own body responds to your own meals. Courses end. Subscriptions don’t. That gap is where the money goes.
Who should care: Anyone whose last blood test came back in the prediabetes range and who was sent home with no plan. That’s a large group — the CDC estimated 96 million American adults, more than one in three, had prediabetes in 2019.
Who should skip this: If you take insulin, these no-prescription sensors are not for you — the makers say so themselves, and this is one of the few places in consumer tech where the warning label is the most honest sentence on the page. Same if you have a history of problematic low blood sugar; Stelo’s own site says do not use it. And if you don’t yet know your A1C, skip the whole purchase for now and read the supporting section instead. That number is probably already sitting in your patient portal, and it costs nothing.
What to do next: Look up your last A1C and fasting glucose before you buy anything. Five minutes, no device, and it decides whether the rest of this issue applies to you.
You’re not buying a monitor. You’re buying two weeks of tuition.
Start with what the thing physically is, because almost every misunderstanding downstream comes from getting this part wrong.
A CGM is a small disc you press onto the back of your upper arm. A filament about the width of a hair sits just under the skin. It does not touch your blood. It sits in the fluid between your cells — interstitial fluid — and measures glucose there, then sends a reading to your phone every few minutes.
That distinction sounds like a technicality. It is the single most important fact about the device.
And while we’re looking at words: “continuous” is doing something to you. It sounds like thoroughness — nothing gets missed, it’s watching the whole time, so it must beat a once-a-year blood test. But “continuous” describes only how often the thing samples. It makes no claim about whether any given reading is right, and it makes no claim that reading number four thousand tells you anything reading number four hundred didn’t. It is a sampling rate wearing the costume of a quality rating. Hold onto that, because it explains both of the next two sections.
Glucose arrives in that fluid after it arrives in your blood. The delay is around five minutes when your levels are steady, and reported ranges run from roughly 5 to 20 minutes depending on where the sensor sits and how fast things are changing. When glucose is moving quickly — right after a meal, during exercise — the gap widens. So the graph on your phone is not what your blood is doing. It is what your blood was doing, a few minutes ago, filtered through your tissue.
Now here’s the part that should change how you read that graph.
It runs high in people who are fine
In February 2025, researchers at the University of Bath published a study in the American Journal of Clinical Nutrition comparing a commercial CGM — an Abbott FreeStyle Libre 2 — against finger-prick tests in healthy, non-diabetic volunteers. The finger-prick is the reference standard.
The sensor overstated how much time these healthy people spent above a normal blood-sugar threshold — by roughly fourfold as measured, and still about twofold after the researchers adjusted for differences between the two methods. Against the specific threshold Diabetes UK uses, the overstatement ran to nearly 400%. It also rated a smoothie’s glycemic index — the measure of how fast a food raises blood sugar — at 69 when the finger-prick put it at 53, and whole fruits got misclassified as medium or high-impact foods when the reference test said they were low.
Professor Javier Gonzalez, one of the authors, put the implication plainly: for healthy people, “relying on CGMs could lead to unnecessary food restrictions or poor dietary choices.”
Size it honestly before you lean on it: 15 volunteers. That is a small study, and I’d rather you hear that from me than find it yourself and wonder what else I skipped. It doesn’t settle the question. What it does is give a physical explanation — the sensor is reading the wrong fluid, a few minutes late — for something people already report, which is that these devices make ordinary food look alarming.
One disclosure you should have, because I’d want it: that study was funded by a smoothie company — innocent drinks, lowercase i, that’s the brand. Read it knowing that. I’ll say why I still trust the finding: the funder’s interest points toward smoothies looking harmless, and the result went against the device, which is not the direction money usually bends a study. But you get to weigh that yourself, and you couldn’t if I didn’t tell you.
So: your brother-in-law’s grapes may not have done what the graph said they did. The sensor runs hot, and whole fruit is exactly the category it got wrong.
Why the value is front-loaded
This next part is my read, not a study, and I’ll flag it as such.
Think honestly about how you eat. Most of us rotate through maybe ten to fifteen meals. The same two breakfasts. The same lunch four days a week. Three or four dinners in heavy rotation. A weekend pattern that’s different from the weekday one.
A CGM’s entire job for a non-diabetic is to show you how your body handles those meals — and what a walk, a bad night’s sleep, or a stressful afternoon does on top of them. Which means that after you’ve run your actual rotation past it once or twice, and tried the two or three experiments worth trying, the sensor starts telling you things you already know. The information doesn’t stop; the new information does.
That’s a course. You take a course, you finish it, you keep the notes.
The pricing tells you the industry knows this, if you look at it sideways. Stelo, from Dexcom, sells a two-sensor pack for $99 as a one-time purchase — each sensor runs up to 15 days, so that’s about a month of wear — or $89 a month if you subscribe. Abbott’s Lingo sells a single 14-day sensor for $49, two for $89, six for $249. The subscription is the cheaper unit price, which is the standard shape of every subscription ever built. It’s only cheaper if you needed the twelfth month.
What the guidelines actually say — and where they stop
The American Diabetes Association’s Standards of Care for 2026 expanded CGM recommendations meaningfully: for the first time, CGM is recommended for adults with type 2 diabetes who are on glucose-lowering therapies other than insulin.
Read the shape of that. The body with the deepest evidence base on this device just widened its recommendation — and stopped at the edge of diagnosed diabetes. It did not step across into “everyone should wear one.” When a group that has every reason to endorse a tool declines to endorse it for a population, that silence is data. It’s not a prohibition. It’s an absence of evidence, which is a different and more honest thing.
The manufacturers drew the same line themselves, which is the part I find genuinely persuasive. Dexcom and Abbott between them own this category. When each of them launched a no-prescription version, they had every commercial reason to make it available to the widest possible group — and both excluded the same people: anyone taking insulin. That is the largest, most motivated, most price-insensitive group of glucose-watchers in the country, and both companies wrote them out of the label voluntarily. A company does not turn down its best customers for fun. It does it because the FDA makes it prove what it claims, and it decided not to claim that.
Medicare draws the same line with money. Part B covers a CGM for people with diabetes who are insulin-treated or have a history of problematic low blood sugar, with a doctor’s visit inside six months to confirm it. Learning about your metabolism is not on the list. So the person who wants the sensor for education pays cash, and the person whose sensor is covered already has the diagnosis it would have taught them about.
(I’m deliberately not covering Medicare’s rules across the rest of this category this week — alert buttons, blood-pressure cuffs, the whole home-health aisle, where the coverage logic is genuinely strange and worth real money. That’s its own issue and I want to do it properly.)
Bobby’s Verdict
Best for: Someone with a prediabetes-range lab result and no medication yet, who wants to see which of their own habits actually move the number — and who will take the sensor off when the learning stops.
Skip if: You take insulin (the makers exclude you explicitly). You have a history of problematic low blood sugar — Stelo’s site says do not use it. You don’t know your A1C yet. Or you’re prone to reading a number as a verdict on yourself; this device generates a number every five minutes, and there are people for whom that is not a good trade.
Setup difficulty: Low, with one prerequisite worth checking before you order. It needs a smartphone — Stelo’s app runs on both iPhone and Android — plus an account you create, and a phone recent enough to run the current app. If the person this is for doesn’t carry a smartphone, or carries one they don’t use, stop here; the sensor has no screen and no reason to exist without the phone. Past that: an applicator, an app, ten minutes. The hardest part is the first hour of resisting the urge to check it constantly.
Monthly cost: $99 one-time for a Stelo 2-pack, roughly a month of wear; $89/month if you subscribe. Not covered by insurance. HSA/FSA eligible — see Quick Win 2.
Privacy note: I read the policy so you don’t have to, and this is the one part of the purchase I’d think hardest about. Stelo is covered by Dexcom’s general privacy policy, not a Stelo-specific one. It says the company collects glucose readings, device identifiers, medical history and treatment information, and specific geolocation, among other things. On sharing, it says two things that sit oddly together, and you should have both: that disclosure to “advertising networks, social networks” and similar “qualifies as the 'sale' of Personal Data” under certain state privacy laws — and, separately, that “we do not 'sell' any consumer data in the traditional sense for monetary compensation.” Both sentences are theirs. The policy also states plainly that it is distinct from the HIPAA notice that governs your doctor’s records, which is the clearest confirmation you’ll get that this is not the same regime as your medical file. None of that is unusual for a consumer health product, and none of it is a reason to panic. It is a reason to decide before you create the account rather than after a month of glucose data exists. And know the exit: Dexcom offers a right to deletion at privacyrequest.dexcom.com or 1-866-384-4277. If you’re only using the sensor for two weeks, deleting the account afterward is a reasonable end to the project.
My take: Buy one 2-pack. Wear both sensors. Write one page. Take it off, and don’t set up the auto-ship. If a year from now something real has changed — a new medication, a new diagnosis, a doctor asking a specific question — buy another pack then, with a question to answer.
The two-week protocol
The point of this is to leave with a page of notes instead of a habit.
Before you open the box, write down five questions. Real ones. “Does oatmeal do more to me than eggs?” “Does the after-dinner walk actually help or do I just feel virtuous?” “Does a bad night’s sleep change the next morning?” If you can’t write five, you’re not ready to buy — and that’s a $99 save, not a failure.
Sensor one: change nothing. Eat your ordinary two weeks. No experiments, no performing for the graph. You’re establishing what normal looks like, and you cannot do that while also trying to impress it.
Sensor two: run the experiments, one variable at a time. Same breakfast two days running, walk after it on one. Same dinner, one with a 10-minute walk after. Note the days you slept badly. One change at a time or you’ll learn nothing.
Write the page. Three things that moved your numbers, two that didn’t, one habit worth keeping. That page is what you bought.
Take it off and cancel nothing, because you never subscribed.
Take the page to your next appointment. “Here’s what I saw over a month” is a far better opening than “I read something online.”
What to watch — and what not to assume
Watch patterns, not points. A single high reading means almost nothing, especially given the lag and the overstatement problem above. The same meal producing the same shape three times is worth something.
Expect some sensors to quit early. Dexcom’s own site reports a study in which 77.9% of Stelo sensors lasted the full 15 days — so roughly one in five may not go the distance. Budget for that rather than being outraged by it.
A rise after eating is not damage. Glucose going up after a meal is what glucose does. The wellness internet has quietly redefined a normal physiological response as an event to be prevented, and the sensor’s tendency to overstate it pours fuel directly onto that fire.
Don’t cut foods based on the graph alone. That is the specific harm the Bath researchers warned about, and whole fruit is the exact food their sensor got wrong.
It is not a diagnosis, and it doesn’t replace a meter. Stelo’s own guidance says if your readings don’t match your symptoms, a blood glucose meter may be an option, and that you shouldn’t take medical action on the device’s output without talking to a professional. When the manufacturer’s marketing page tells you to go get a different device, believe it.
👉 Stelo by Dexcom, 2-sensor pack on Amazon — the one-time pack, deliberately, not the subscription. If you buy this, buy it once, run the two weeks above, and let it end. That’s the version of this purchase I’d recommend to my own family, and it’s the only version I’d recommend at all.
This newsletter uses affiliate links. If you buy through one, I may earn a small commission at no extra cost to you. It never changes what I recommend or what I tell you to skip.
The two numbers that cost nothing, and probably already exist
Before you spend $99 to generate thousands of readings, it’s worth knowing that the two numbers that actually determine whether you have a problem are cheap, standardized, and quite possibly already in your file.
Fasting glucose is a single measurement after not eating overnight — the number on that lab printout. A1C is different and more useful: it’s a blood test that reflects your average blood sugar over roughly the past three months. One is a snapshot; the other is the trend. The sensor on your arm is neither — it’s a high-resolution movie of two weeks, which is a genuinely different thing from both.
And here’s the part that gets lost when people treat the cheap tests as the budget option: the old tests do something the expensive one cannot. A sensor you wore for four weeks cannot tell you your three-month average, because it wasn’t there. The finger-prick can’t be beaten either — it’s the reference standard the sensor gets validated against, which is precisely why the Bath researchers used it to catch the sensor running high. The cheap, boring, decades-old tests are the ones with a settled relationship to an actual diagnosis. The expensive continuous one is the one that has to be interpreted before it means anything.
Which points at the real mismatch. Everything about a CGM — the five-minute refresh, the alerts, the live graph — is engineered for someone who needs to know their glucose right now in order to do something about it in the next ten minutes. That’s an insulin user. If your actual question is “is my average drifting the wrong way over years,” you have been handed a tool built around somebody else’s question. It can still teach you things. It just isn’t shaped like your problem, and the blood test you may have already had is.
The thresholds are not secret and not in dispute. The CDC and the American Diabetes Association put prediabetes at a fasting glucose of 100 to 125 mg/dL, or an A1C of 5.7% to 6.4%. Below that is the normal range; above it is the diabetes range. I’m giving you the numbers so you can find your own on the page — not so you can diagnose yourself. Which side of a line you land on, and what to do about it, is a conversation with your doctor, and it’s a short one if you walk in already knowing your value.
This week: log into your patient portal and find your most recent basic metabolic panel or comprehensive metabolic panel. Look for “Glucose.” Then look for “Hemoglobin A1C” — it may be on a different report, and it may not have been ordered at all. If it wasn’t, that’s the single most useful thing to ask for at your next blood draw. It’s a standard, inexpensive test that gets added to routine bloodwork all the time.
If both numbers are comfortably normal, you’ve just answered the question this issue opened with, for free, and you can skip the purchase entirely with a clear conscience.
Quick wins
1. Walk for ten minutes after your biggest meal of the day.
If you did buy the sensor, this is the intervention it would most likely have taught you anyway — so you may as well have it now, at no cost.
The research here is unusually consistent. Reviews pooling many studies find that walking after eating lowers the post-meal glucose rise, in people with diabetes, with prediabetes, and with neither. The timing matters more than the duration: a short walk soon after the meal does more for that meal’s rise than a longer walk at some other point in the day, and even a couple of minutes of gentle movement beats sitting still.
Not a workout. Not a program. The dishes, the mail, once around the block after dinner. Attach it to something you already do and it stops requiring willpower.
2. Check your HSA or FSA card before you check the price.
Stelo is HSA/FSA eligible. So if you have a health savings account or a flexible spending account sitting there — and this audience often does, sometimes with money that expires at year-end — the $99 comes out of pre-tax dollars. Depending on your bracket, that’s a meaningful discount that requires nothing but using the right card at checkout.
It’s a small thing. It’s also the kind of small thing nobody tells you, and it applies to a lot more of the health aisle than most people use it for.
Skip this for now
Any watch or ring that says it reads your blood sugar without breaking the skin
This one isn’t a matter of taste, and it’s the reason I wanted to write this issue carefully.
On February 21, 2024, the FDA issued a safety communication with an unusually blunt title: do not use smartwatches or smart rings to measure blood glucose levels. The agency stated it has not authorized, cleared, or approved any smartwatch or smart ring intended to measure or estimate blood glucose on its own. Its surveillance had turned up dozens of companies and multiple brand names selling exactly that. The stated risk is direct: an inaccurate reading, acted on, can lead to a dangerous dose of insulin or medication.
The distinction that matters, because it trips people up: a watch that displays readings from a real, FDA-authorized sensor is fine. That’s just a screen for a device that pierced the skin. What the FDA is warning about is a watch or ring claiming to do the measuring itself, through unbroken skin. Nobody has cleared that. Not Apple, not Samsung, not the ad you saw on Facebook for $89.
And the claim keeps mutating into new vital signs. The tech press has been running pieces about a smartwatch that will measure “real blood pressure” — a different measurement with its own complications, but the same promise, made in the same shape. When a sensor claim arrives without a regulatory clearance attached, the right move is to wait. Nothing bad happens to you if you’re twelve months late to a good device. Something bad can happen if you’re early to a wrong one.
Which one are you?
You already have a number you’re watching — a glucose, an A1C, a blood pressure — and you’re trying to move it.
Nobody has ever given you a number, and you’re not sure whether you should go looking.
Hit reply with “watching” or “looking.” One word is genuinely enough. Those two groups need completely different issues from me, and right now I’m guessing which one you’re in.
If you take one thing from this: the question with any sensor you wear is not whether it works. It’s how long it stays useful — and whether the price is structured around that answer or around the opposite one. That test applies well beyond glucose.
Next week: Google just raised the price of nearly every Nest camera and doorbell. If you put one up to keep an eye on a parent’s front door, or your own, I’ll have what changed and whether it’s worth switching.
Reply and tell me which one you are. I read all of them, and this newsletter gets better in exactly the direction you point it.
— Bobby