Continuous glucose monitoring improved glucose measures over 26 weeks in the CONNECT trial of adults with type 2 diabetes who were not using insulin. The study concerns people receiving diabetes care in US primary-care practices, rather than people without diabetes looking to track ordinary glucose fluctuations.
The distinction matters. A monitor can be used for different purposes in different populations. Evidence of a benefit in one setting should not be carried over to every person who can buy a sensor.
Who took part in CONNECT
The original published results summary describes 283 adults across 22 US primary-care practices: 145 assigned to continuous glucose monitoring, or CGM, and 138 to routine care.
The official trial registration independently records actual enrolment of 283. Eligibility included a type 2 diabetes diagnosis, no recent insulin use and a screening HbA1c of at least 7.5%. Participants also needed to be able to use the study devices and follow the protocol.
This is a selected clinical population. It does not represent every person with type 2 diabetes, including everyone whose glucose is already within their agreed treatment range. The registered design was open-label: people knew whether they were using real-time CGM.
The registry also distinguishes the 26-week randomised comparison from a planned extension in which both groups would use CGM. Those periods should not be combined into a claim that the published comparison already demonstrated a year of benefit.
Two ways of describing glucose
NIDDK explains that CGM uses a sensor to estimate glucose in the fluid between cells, with information sent to a display or app. It makes patterns across the day visible, rather than providing only one measurement at a particular moment.
HbA1c is different. The laboratory blood test reflects average glucose over approximately three months. It summarises a longer period without showing the timing of every rise or fall.
| Measure | What it describes |
|---|---|
| HbA1c | A longer-term average, reported as a percentage in this trial. |
| Time in range | The proportion of sensor time within a specified glucose interval. |
| A current sensor value | An estimate at that moment, rather than the whole pattern or long-term average. |
These measures can move together, but they do not have the same units or interpretation.
What changed during the trial
In the published summary, average HbA1c fell from 8.8% to 7.2% with CGM and from 8.7% to 8.0% with routine care. The adjusted difference between groups was 0.9 percentage points in favour of CGM, with a reported 95% confidence interval of 0.6 to 1.1 percentage points.
The CGM group also spent 21 percentage points more time in the trial’s range of 70–180 mg/dL. That is approximately five additional hours per day. The report states that no severe hypoglycaemia events occurred.
These are group results, not a forecast for an individual. “Percentage points” also matters: the HbA1c difference is not a 0.9% relative reduction, and the time-in-range difference is not twenty-one extra hours.
Dexcom funded the trial. Its 26-week glucose outcomes do not by themselves establish longer-term effects on complications, cost effectiveness or a particular insurance policy’s coverage.
A different question from monitoring without diabetes
Our guide to glucose monitoring without diabetes asks what tracking can establish outside diagnosed diabetes care. CONNECT answers a narrower clinical research question in people who already have type 2 diabetes and elevated HbA1c.
That makes the trial relevant without making CGM a universal recommendation. A sensor, the way its information is used and the care around it remain different parts of the intervention. Reading the population, comparator, duration and outcome together is the useful way to understand what this result adds.