CGM Lowered A1c Without Insulin. The Sensor Is Not the Treatment.

A continuous glucose monitor does not lower glucose.

It observes glucose. It shows when it rises, how long it stays elevated and what happens after a meal, walk, medication dose or ordinary Tuesday afternoon.

That distinction matters when interpreting the CONNECT trial, published October 14 in The New England Journal of Medicine. The study found that adults with type 2 diabetes who used continuous glucose monitoring achieved considerably better glucose control than those receiving routine care—even though none were using insulin.

The result is important. It suggests that better feedback can change outcomes in a population for whom continuous monitoring has often been treated as optional.

But the sensor itself was not the intervention in the way a drug is an intervention. It supplied information. The benefit depended on what people and their clinicians did with it.

What the trial tested

CONNECT enrolled 283 adults with type 2 diabetes across 22 primary-care practices in the United States. Participants ranged from 21 to 84 years old and were not being treated with insulin.

Their average starting A1c was approximately 8.8%, indicating that this was a population with substantial room for improvement—not a group of metabolically healthy adults investigating small post-meal fluctuations.

Researchers randomly assigned 145 participants to use a Dexcom G7 continuous glucose monitor and 138 to routine care. Both groups continued receiving diabetes care, and the study followed them for 26 weeks.

The primary outcome was A1c, a measure reflecting average glucose exposure over roughly the preceding two to three months.

At the end of the trial:

• Average A1c fell from 8.8% to 7.2% in the CGM group.
• It fell from 8.7% to 8.0% in the routine-care group.
• The adjusted difference between groups was 0.9 percentage points in favor of CGM.

That is not a trivial separation.

Participants using CGM also spent 21 percentage points more time in the glucose range of 70 to 180 mg/dL. That translates to approximately five additional hours per day in range.

Importantly, this improvement did not come with more hypoglycemia. Time spent below both 70 and 54 mg/dL was slightly lower with CGM, and the trial reported no severe hypoglycemic events.

Why seeing the data may change the outcome

Traditional glucose monitoring offers isolated measurements. A finger-stick reading might show what glucose is doing at one particular moment, but it can miss the shape of the day: the delayed rise after dinner, the prolonged elevation after a sedentary afternoon or the different response to two superficially similar meals.

CGM turns those moments into a pattern.

That pattern can make cause and effect more visible. It may help someone recognize that a meal produces a larger or longer rise than expected, that physical activity changes the curve or that glucose remains elevated overnight. It can also give clinicians more information when assessing whether the existing plan is working.

The CONNECT trial did not isolate precisely which decisions produced the improvement. It tested access to the CGM system as a package: the device, its feedback and the behavior or clinical conversations that followed.

That is why “CGM lowered A1c” is convenient shorthand, but incomplete. A more accurate interpretation is that access to continuous feedback helped participants achieve better control within a diabetes-management setting.

Information became useful when it changed action.

This is stronger evidence than a wellness experiment

Consumer interest in CGM has expanded well beyond diabetes. Healthy users now wear sensors to compare breakfasts, identify “spikes” or pursue a flatter glucose trace.

CONNECT should not be used to validate all of those practices.

The participants had established type 2 diabetes and an average A1c close to 8.8%. The trial measured improvement in clinically elevated glucose over six months. It did not enroll healthy adults, establish an ideal glucose curve for people without diabetes or show that suppressing every post-meal rise improves health.

A temporary increase in glucose after eating is normal physiology. Its significance depends on magnitude, duration, overall metabolic context and whether glucose regulation is already impaired.

There is also a meaningful difference between using data to solve a defined problem and collecting data because more measurement feels inherently better.

For someone with inadequately controlled type 2 diabetes, CGM may reveal actionable patterns that an occasional measurement misses. For a healthy person, the same stream of information can create noise, food anxiety or unnecessary attempts to micromanage normal variation.

The value of a monitor depends on the question it is being asked to answer.

What the study does not establish

CONNECT was a strong randomized trial, but its boundaries matter.

First, it lasted 26 weeks. That is long enough to demonstrate a sustained A1c difference across more than one red-blood-cell cycle, but not long enough to show whether participants maintained the improvement for years.

Second, it measured biomarkers rather than clinical complications. Lower A1c and less time in hyperglycemia are meaningful, but the study did not test whether CGM reduced heart attacks, kidney failure, vision loss, neuropathy or mortality.

Third, participants and clinicians knew which group they were in. Blinding is not realistically possible for an intervention that displays live glucose data, but awareness may influence engagement and other behavior.

Fourth, the trial was funded by Dexcom. Several authors were affiliated with the company, and the tested device was the Dexcom G7. Industry funding does not invalidate a randomized result, but it belongs in the interpretation—especially when questions about durability, comparative devices and cost remain.

Finally, these participants began with glucose levels high enough that there was a clear signal to improve. The trial tells us much less about people whose A1c is already near target, people with prediabetes or people without a glucose disorder.

The more useful question: what decision will the data change?

The practical case for CGM is strongest when three things are present:

1. There is a defined metabolic problem.
2. The monitor can reveal information that current measurements are missing.
3. Someone is prepared to use that information in a structured way.

That could mean identifying persistent overnight hyperglycemia, comparing repeatable meal patterns or giving a clinician more context when evaluating an existing management plan.

It does not mean reacting to every individual peak. CGM readings have measurement error and physiological lag, and a single unusual trace is rarely the most useful unit of interpretation. Repeated patterns matter more.

A monitor is most valuable when the feedback loop is clear:

Observe a pattern. Form a reasonable hypothesis. Make an appropriate change. Then look for a repeatable response.

Without that structure, continuous data can become continuous distraction.

Practical takeaways

• This was a meaningful result. In adults with type 2 diabetes who were not using insulin and began with an average A1c near 8.8%, CGM improved A1c by an additional 0.9 percentage points over routine care.

• The benefit extended beyond the average. Participants spent about five more hours per day in the 70–180 mg/dL range without an increase in hypoglycemia.

• The study does not apply cleanly to healthy users. It provides evidence for management of established type 2 diabetes, not proof that healthy adults benefit from flattening every post-meal glucose rise.

• The device supplied feedback, not treatment by itself. The useful mechanism is likely the decisions, behaviors and clinical adjustments enabled by better information.

• Long-term outcomes remain unanswered. CONNECT did not show whether CGM use in this population prevents complications or whether the benefit persists for years.

• A defined question should come before a new stream of data. The most useful reason to monitor is that the result could change a meaningful decision.

The bottom line

CONNECT moves continuous glucose monitoring further beyond its original role in insulin dosing. For adults with type 2 diabetes and glucose meaningfully above target, real-time feedback can improve control even when insulin is not part of the treatment plan.

That is a substantive expansion of the evidence.

It is not, however, a mandate for universal glucose tracking. The trial’s lesson is not that every person needs a sensor. It is that measurement can be powerful when it is attached to a real problem, an interpretable pattern and a decision worth changing.

Sources

• Oser TK, Martens T, Ruedy KJ, et al. Continuous Glucose Monitoring in Non–Insulin-Treated Type 2 Diabetes. New England Journal of Medicine. Published online October 14, 2026. DOI: 10.1056/NEJMoa2608571.
• PubMed record and abstract. PMID: 42878622.
• CONNECT trial registration. ClinicalTrials.gov identifier: NCT06601166.

A clearer next step

Health data is most useful when it answers a defined question. Explore The Modern Bio Testing Lab for evidence-aware guidance on A1c, glucose testing and other biomarkers—including what each test can clarify and where its limits begin.

Get one deeply researched, evidence-aware article each week with the free Modern Bio Brief.