- Design
- Randomised, open-label, 26 weeks, 22 primary care practices
- Population
- 283 adults with type 2 diabetes not treated with insulin
- Primary outcome
- HbA1c at 26 weeks
- Effect
- Adjusted difference -0.9 percentage points (95% CI -1.1 to -0.6)
The CONNECT trial randomised 283 adults aged 21 to 84 with type 2 diabetes not on insulin, in 22 US primary care practices, to continuous glucose monitoring (CGM) or routine care for 26 weeks. Published in the New England Journal of Medicine on 14 October 2026.
HbA1c fell from 8.8% to 7.2% with CGM and from 8.7% to 8.0% with routine care, an adjusted difference of -0.9 percentage points (95% CI -1.1 to -0.6). Time in range was 21 points higher with CGM. Time spent below 70 and below 54 mg/dL was slightly lower, and there was no severe hypoglycaemia.
This matters because most people with type 2 diabetes are not on insulin, and guidelines have been cautious about CGM for them. A 0.9-point fall is similar to adding a glucose-lowering drug, and it came from primary care, not specialist clinics.
The limits are an open-label design, funding by a sensor manufacturer, and 26 weeks of follow-up. Baseline HbA1c was high, so the effect may be smaller in better-controlled patients. Sensor cost is the main barrier in India; intermittent use may be the realistic way to apply this.
- Consider CGM for adults with type 2 diabetes not on insulin whose HbA1c is well above target
- Use the sensor data to show patients how meals and activity move their glucose
- Review the report with the patient at each visit; the device alone is unlikely to explain the benefit
- Where cost limits continuous use, a single 14-day sensor before a treatment decision is a pragmatic option
- Expect little hypoglycaemia in people on metformin, SGLT2 inhibitors or GLP-1 receptor agonists
Why it matters
It challenges the assumption that sensors only pay off for people who dose insulin.
Don't overread it
Open-label, manufacturer-funded and six months long; whether the benefit lasts or improves complications is not known.
The statistics, in plain English
The confidence interval for the HbA1c difference runs from -1.1 to -0.6, so even the least favourable estimate is a clinically meaningful fall. The hypoglycaemia results were tested for non-inferiority, meaning the trial set out to show CGM did not increase lows, and it did not.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for diabetes & endocrinology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free