# CGM for type 2 diabetes not on insulin cut A1c 0.9 points URL: https://sciencebased.news/cgm-type-2-diabetes/ Published: 2026-10-17 Updated: 2026-10-19 Claim: If you have type 2 diabetes and aren't on insulin, a glucose monitor can drop your A1c almost a full point. Grade: Promising What to do: - If your HbA1c is 7.5% or higher and you are not on insulin, ask your doctor about a real-time CGM worn all the time. The trial used a Dexcom G7 for 26 weeks. - Go over the readings with your clinician. The trial did this at six visits in 26 weeks, and the sensor group added a diabetes medicine more often (31% against 17%). - Expect the biggest change in the first month. HbA1c fell 1.6 points with the sensor against 0.7 with routine care, and 46% against 24% got under 7%. - Safety: no one had a severe low. 3 of 145 people had minor bleeding where the sensor went in. - Not proven: that the drop lasts past 6 months, or that it prevents heart, kidney or eye disease. Dexcom, the sensor's maker, paid for the trial. # CGM for type 2 diabetes not on insulin cut A1c 0.9 points *Evidence check, October 17, 2026. Updated October 19, 2026.* > **The claim:** "If you have type 2 diabetes and aren't on insulin, a glucose monitor can drop your A1c almost a full point." > > **Our grade: Promising.** In a randomized trial of 283 adults not on insulin, a continuous glucose monitor lowered HbA1c 0.9 points more than routine care in 26 weeks, with the same doctor visits in both groups. It is one trial, paid for by the sensor's maker, and it lasted 6 months. ![A woman in her 60s with short gray hair smiles at a glucose graph on her phone, a small round glucose sensor on the back of her upper arm](https://sciencebased.news/images/cgm-type-2-diabetes-hero.jpg) Ruth is 64, and her last blood test put her HbA1c at 8.8%. She takes pills for type 2 diabetes, not insulin. A sensor on her arm would only show her the numbers. It treats nothing. Could just seeing the numbers bring hers down? In a new trial, it did. CONNECT, published in the *New England Journal of Medicine* in October 2026, gave 145 adults with type 2 diabetes a continuous glucose monitor (CGM), a coin-sized sensor that reads sugar under the skin every few minutes and sends it to a phone. Another 138 got routine care. After 26 weeks, HbA1c, a three-month average of blood sugar, had fallen from 8.8% to 7.2% with the sensor and from 8.7% to 8.0% without it. The gap was 0.9 points [1]. That is a puzzle. Eight earlier trials in the same kind of patient found gaps of about 0.4 points on average [3]. Why did this one find more than twice as much, from a device that treats nothing? ## Where the claim came from The result was first shown at the American Diabetes Association's meeting in June 2026, under the headline that 82% of adults with type 2 diabetes not on insulin improved their blood sugar with a CGM [2]. The full paper followed in October [1]. CGMs have long been standard for people who take insulin, who need the readings to dose it and to catch dangerous lows. For people who don't take insulin, the case was thinner, even after the FDA cleared the first CGM sold without a prescription, Dexcom's Stelo, for them in 2024 [7]. ## The only difference was seeing the numbers CONNECT ran at 22 US primary care practices, the ordinary doctor's offices where most type 2 diabetes is treated. The 283 adults were 21 to 84 years old, 60 on average, and none had used insulin in the past year. Their HbA1c was 7.5% or higher; 40% were already on a GLP-1 drug and 37% on an SGLT2 drug [1]. Both groups kept their medicines, got the same diet and exercise advice, and saw a clinician at weeks 1, 4, 8, 13, 19 and 26, who could change their drugs. The routine-care group even wore a sensor, but a blinded one that showed nothing until the end. So the gap is hard to blame on extra attention. With all those visits, routine care alone cut HbA1c 0.7 points. The sensor group's fell 1.6 [1]. ## What the sensor changed The clearest picture is time in range: the share of the day blood sugar stays between 70 and 180 mg/dL, the target band. At the start, both groups spent about 30% of the day there. The chart is redrawn from the paper's own figure. ![Line chart of time in range over 26 weeks. With a CGM, the share of the day with blood sugar between 70 and 180 mg/dL rose from about 29% at the start to 53% at week 4, 59% at week 13 and 65% at week 26. With routine care it went from about 30% to 33%, 39% and 44%. Redrawn from Figure 2 of Oser et al., NEJM 2026.](https://sciencebased.news/images/cgm-type-2-diabetes-chart.png) *Redrawn from Figure 2 of the CONNECT trial [1].* Most of the jump came in the first four weeks. Over weeks 13 and 26 together, the sensor group spent 62% of the day in range against 42%: a gap of 21 points, about 5 more hours a day [1]. Very high readings, over 250 mg/dL, fell from 33% of the day to 7%, against 31% to 22% with routine care. By the end, 82% of the sensor group had dropped at least half a point, against 56%, and 46% had an HbA1c under 7%, the usual goal, against 24% [1]. Tighter control can bring lows. Here they got rarer: time under 70 mg/dL went from 0.30% of the day to 0.12% with the sensor, against 0.28% to 0.31%. No one had a severe low [1]. ## So why so much? The first suspect is medicine. Seeing high numbers, clinicians added a diabetes drug for 31% of the sensor group and 17% of the routine-care group [1]. That is part of it, but not most. Among people who started no new drug, the sensor group still ended 0.9 points lower than routine care, and among those who did, 0.5 lower [1]. The second suspect is weight. The sensor group lost 2.5 kg, against 1.8 kg, a difference the trial could not tell from chance [1]. That leaves behavior: what people ate, when they walked, how they reacted to a high reading. Here the trial is silent. It did not record diet or exercise, and the authors say they cannot tell how the sensor worked [1]. The best clue is in what the earlier trials tested. A 2025 review pooled the 8 randomized trials in people not on insulin, 541 people in all, and found HbA1c 0.37 points lower with a sensor and time in range up about 9 points [3]. Four used scanning sensors, which show a reading only when you hold a phone to them. Three used real-time sensors for a week or two at a time. None used a real-time sensor all the time [3]. CONNECT did: the sensor showed sugar on the phone around the clock, warned of highs, and people wore it 97% of the time [1]. Constant feedback is a plausible reason the effect more than doubled. No trial has tested that head to head. In the Canadian IMMEDIATE trial, 116 people got a FreeStyle Libre scanning sensor and diabetes education, or education alone, for 16 weeks. Time in range rose about 10 points more with the sensor, and HbA1c fell 0.3 points more [4]. In a Japanese trial of 100 people, a scanning sensor lowered HbA1c 0.29 points more than fingersticks over 24 weeks [5]. The 2025 review's authors report fees from Abbott and Dexcom, the two main sensor makers [3]. In CONNECT, 97% of the sensor group finished, against 90% with routine care. Fingerstick tests went from 3 a week to none with the sensor, and from 2 to 3 with routine care [1]. Serious adverse events happened to 6 people in each group, none linked to the trial; one person in the routine-care group died of a heart attack. Three sensor problems were reported, all minor bleeding where it went in [1]. So back to Ruth. The sensor can't lower her sugar by itself. What CONNECT shows is that seeing it all day, with a clinician who sees it too, lowered HbA1c almost a full point more than the same visits without it. The sensor works through the choices it changes. ## The limits It is one trial. The 0.9 points comes from 283 people in a single study, and no other trial has found a gap that large. Dexcom paid for it, provided the sensors and helped design it, though the paper says the company had no role in collecting or analyzing the data [1]. It lasted 26 weeks. A 6-month extension is under way [2]. The best longer look is an observational study of 25,648 people at Kaiser Permanente: HbA1c was 1.1 points lower after one year in those who started a real-time CGM and about 0.8 lower after three. But nearly 9 in 10 of those users took insulin, and 44% stopped wearing the sensor [6]. HbA1c is a stand-in. It predicts heart, kidney and eye damage, but no trial has shown that a CGM prevents them in people not on insulin. The setup was generous. Sensors were free, and both groups saw a clinician six times in six months, more than most people get. What the sensor does without those visits is unknown. ## Our grade So which parts of the claim hold, and what would make it Solid? | Question | Evidence | Grade | |---|---|---| | A CGM lowers HbA1c more than fingersticks in type 2 diabetes not on insulin [1, 3, 4, 5] | 8 earlier randomized trials, 541 people, 0.37 points; CONNECT, 283 people, 0.9 points | Solid | | Almost a full point [1] | One randomized trial of 283 people, 26 weeks, open-label, funded by Dexcom | Promising | | The sensor works by itself, without a clinician [1] | Not tested: both groups had six visits, and the sensor group got more new drugs | Unproven | | Overall: a CGM can drop A1c almost a full point in type 2 diabetes not on insulin | One large, well-built trial; smaller trials agree on direction at less than half the size | Promising | The case for the other side: CONNECT was large, randomized and carefully controlled, down to a hidden sensor in the comparison group. Every smaller trial points the same way. That is close to Solid. What holds it back is that the full point rests on one trial, paid for by the maker, over six months. A second independent trial finding a gap near 0.9, or the extension showing it holding at a year, would move it up. ## What this means for you - If you have type 2 diabetes, aren't on insulin, and your HbA1c is 7.5% or higher, ask your doctor about a real-time CGM. CONNECT used a Dexcom G7, worn all the time for 26 weeks [1]. - Use it with your clinician. In the trial, readings were reviewed at six visits, and 31% of the sensor group added a drug [1]. Watch time in range, the share of the day between 70 and 180 mg/dL, not single spikes. - Without a prescription, Stelo is cleared for adults not on insulin [7]. It was not the sensor CONNECT tested. - For comparison, [berberine](/berberine/) lowered HbA1c about 0.4 points beyond placebo in 12-week trials. The sensor's effect in CONNECT was more than twice that. ## What's next The CONNECT extension will show whether the gap holds past 6 months [2]. The next thing we want to see is a trial that tests the sensor without the extra clinic visits, the way most people would actually use one. For people without diabetes, a CGM tells a different story: see [what a glucose spike after oatmeal really means](/glucose-monitors/). If your numbers are in the diabetes range, read our check of [berberine](/berberine/), and for the strongest drugs, [keeping muscle on GLP-1 drugs](/glp1-muscle-loss/). ## Sources 1. Oser TK, Martens T, Ruedy KJ, et al., for the CONNECT Study Group. Continuous glucose monitoring in non-insulin-treated type 2 diabetes. *New England Journal of Medicine*, 2026. 283 adults at 22 US primary care practices, randomized, open-label, 26 weeks; funded by Dexcom, which provided the sensors. [doi:10.1056/NEJMoa2608571](https://doi.org/10.1056/NEJMoa2608571) 2. American Diabetes Association. 82% of adults with type 2 diabetes not on insulin improve blood glucose levels with CGM. Press release, June 6, 2026. [professional.diabetes.org](https://professional.diabetes.org/sites/dpro/files/2026-06/final_ada.26_connect_trial_press_release_1.pdf) 3. Aronson R, Abitbol A, Bajaj HS, Cheng AYY, et al. Continuous glucose monitoring in noninsulin-treated type 2 diabetes: a critical review of reported trials with an updated systematic review and meta-analysis of randomised controlled trials. *Diabetes, Obesity and Metabolism*, 2025. 8 randomized trials, 541 participants; authors report fees from Abbott and Dexcom. [doi:10.1111/dom.70008](https://doi.org/10.1111/dom.70008) 4. Aronson R, Brown RE, Chu L, Bajaj HS, et al. IMpact of flash glucose Monitoring in pEople with type 2 Diabetes Inadequately controlled with non-insulin Antihyperglycaemic ThErapy (IMMEDIATE): a randomized controlled trial. *Diabetes, Obesity and Metabolism*, 2023. 116 adults in Canada, randomized, 16 weeks. [doi:10.1111/dom.14949](https://doi.org/10.1111/dom.14949) 5. Wada E, Onoue T, Kobayashi T, Handa T, et al. Flash glucose monitoring helps achieve better glycemic control than conventional self-monitoring of blood glucose in non-insulin-treated type 2 diabetes: a randomized controlled trial. *BMJ Open Diabetes Research & Care*, 2020. 100 adults in Japan, randomized, 24 weeks. [doi:10.1136/bmjdrc-2019-001115](https://doi.org/10.1136/bmjdrc-2019-001115) 6. Karter AJ, Parker MM, Finertie H, et al. Continuous glucose monitoring and long-term glycemic and kidney outcomes. *JAMA Network Open*, 2026. 25,648 adults at Kaiser Permanente, observational (emulated trial), 3 years; most CGM users took insulin. [doi:10.1001/jamanetworkopen.2026.37698](https://doi.org/10.1001/jamanetworkopen.2026.37698) 7. US Food and Drug Administration. FDA clears first over-the-counter continuous glucose monitor. Press release, March 5, 2024. [fda.gov](https://www.fda.gov/news-events/press-announcements/fda-clears-first-over-counter-continuous-glucose-monitor)