Most people with type 2 diabetes who don’t use insulin learn about their blood sugar in snapshots: a finger-prick reading now and then, and an HbA1c blood test every few months.
That’s like trying to learn to drive by checking a photo of the road once a week.
A continuous glucose monitor, or CGM, is a small sensor worn on the arm that reads glucose every few minutes and shows it on a phone. Until now, the strongest evidence for these devices came from people who use insulin.
A new randomized trial tested them in people who don’t. Seeing their blood sugar in real time lowered their average glucose about as much as many people see from adding a diabetes medication.
283 adults wore a sensor or kept using finger pricks for six months
The CONNECT trial was run across 22 primary care practices in the US, with statistics handled by the Jaeb Center for Health Research, an independent coordinating centre.
- 283 adults with type 2 diabetes who didn’t use insulin were randomly assigned to wear a sensor (145 people) or to keep using a standard finger-prick meter (138 people).
- Everyone had poorly controlled blood sugar: an HbA1c of at least 7.5%, averaging 8.8%. Most guidelines aim for below 7%.
- Their average age was 60. About 40% were already taking a GLP-1-type drug, such as semaglutide or tirzepatide, and 37% an SGLT2 inhibitor.
- The sensor group wore a Dexcom G7 for 26 weeks.
Both groups got the same diet and exercise education, the same visit schedule and a glucose meter. The only difference was whether they could see their glucose continuously.
HbA1c, the main outcome, reflects average blood sugar over the previous two to three months.
HbA1c fell 0.9 points more with the sensor
HbA1c fell much further with the sensor.
- Sensor group: from 8.8% to 7.2%
- Routine care: from 8.7% to 8.0%
- Extra improvement with the sensor: 0.9 percentage points (95% CI 0.6 to 1.1)
More time in the healthy range. Glucose stayed between 70 and 180 mg/dL (3.9–10 mmol/L) 62% of the time with the sensor, compared with 41% with routine care. That 21-point difference (95% CI 16 to 26) adds up to about five extra hours a day in the healthy range.
More people hit their targets. In the sensor group, 82% improved their HbA1c by at least half a point, and 46% got below 7%.
It helped on top of existing drugs. Benefits appeared in every subgroup reported, including people already taking GLP-1 drugs or SGLT2 inhibitors.
No one had a severe low blood sugar episode. People wore the sensor a median of 97% of the time and reported less diabetes-related distress and more satisfaction than the routine-care group.
The drop is comparable to adding a diabetes pill
For comparison, adding a typical diabetes pill often lowers HbA1c by somewhere around half a point to a point (our general context, not from this trial).
The gain was largest for those who needed it most. People who started with an HbA1c above 10% dropped 3.1 points with the sensor, 2.1 points more than routine care. For those starting below 8%, the extra benefit was small and uncertain. People diagnosed in the past ten years also benefited more than those with longer-standing diabetes. These subgroup results are less certain than the main finding.
Feedback appears to be the treatment
The change came fast. Within the first week, time in range in the sensor group jumped from about 29% to roughly 44%. That’s too quick to be explained by new prescriptions taking effect.
A more likely explanation is feedback. With a sensor, you can see what a bowl of pasta does to your glucose an hour later, or how a 15-minute walk after dinner flattens the curve. The biggest differences between the groups appeared in the evenings and overnight, when snacking and late meals often push glucose up.
Advice like “eat fewer carbs” or “move more” is easy to understand and hard to act on. A live graph turns it into something you can test yourself: What happens if I do this instead?
That’s how feedback works in many areas of life. People save more electricity when they can see their usage in real time, and drivers slow down when a sign shows their speed. The information doesn’t force a change, but it makes the consequences visible.
Interestingly, standard questionnaires about diet and exercise didn’t show clear differences between the groups. Whatever changed may have been small and specific, like a swapped snack or a walk after dinner, that broad questionnaires miss.
The comparison group improved too, without a sensor
The routine-care group improved too, from 8.7% to 8.0%, without a sensor.
They got the same education, phone calls and clinic visits. That drop shows that regular attention and support move blood sugar on their own. The sensor added a large extra benefit on top of that, but it worked alongside care, not instead of it.
What the study doesn’t show
This is a peer-reviewed randomized trial in a top medical journal, but it has limits:
- It was funded by Dexcom, the sensor’s maker, although an independent centre ran the analysis.
- It wasn’t blinded. Everyone knew whether they had a sensor.
- It lasted six months. A 12-month extension is underway, so we don’t yet know whether the benefit lasts.
- Everyone started with high blood sugar. For people whose diabetes is already near target, the benefit looked small.
- It doesn’t show effects on complications like heart disease, kidney damage or eye problems, which would take much longer to measure.
- It says nothing about people without diabetes. Wearing a sensor out of curiosity is a different situation, and this trial didn’t test it. People with eating disorders were excluded.
Adults above their target can reasonably ask about a sensor
The evidence applies to adults with type 2 diabetes who don’t use insulin and whose blood sugar is above target.
As a reasonable inference:
- If that describes you, it’s reasonable to ask your doctor or diabetes nurse about trying a continuous glucose monitor. Insurance coverage varies by country and plan.
- A sensor works best with support: someone to help you read the patterns and decide what to change.
- Use the data to experiment: try the same meal with and without a walk afterwards, or swap one evening snack, and see what changes.
If you don’t have diabetes, this trial isn’t evidence that a sensor will help you.
Information can work like a treatment
Millions of people with type 2 diabetes never use insulin, and many struggle to reach their blood sugar targets even with good medicines.
This trial suggests that one of the most powerful tools may not be a new drug, but information they can see and act on in real time.
Sometimes the most effective treatment is simply letting people see what their choices are doing.