Healthy individuals affix glucose sensors to their arms and use interstitial fluid glucose fluctuations to score their three daily meals—this usage is growing rapidly, prompting even the field of nutrition science to convene a dedicated expert workshop. Meanwhile, in the second half of 2026, this device—having moved beyond diabetes wards—is facing its true evidentiary moment: The New England Journal of Medicine (NEJM) published a definitive randomized trial of continuous glucose monitoring (CGM) in people with type 2 diabetes not using insulin; a related substudy, an expert position statement, and a systematic review addressed two subtler questions—how much prognostic information do CGM’s novel metrics actually carry? And does wearing CGM benefit healthy individuals?
Together, these three layers of evidence form a clear gradient: the further one moves from established clinical indications, the thinner the evidence becomes. The benefits of CGM in people with type 1 diabetes and in those with type 2 diabetes using insulin are already well-established consensus; this article does not address them. What is new occurs outside current indications.
## Layer One: The Definitive Trial Has Arrived
The CONNECT trial, published in NEJM on October 29, 2026, enrolled 283 adults (aged 21–84 years) with type 2 diabetes not using insulin across 22 primary care clinics in the United States. Participants were randomly assigned to either the CGM group (n = 145) or usual care group (n = 138) for 26 weeks. Results were uniformly positive: HbA1c in the CGM group decreased from a baseline of 8.8 ± 1.3% to 7.2 ± 1.1%, while in the usual care group it decreased from 8.7 ± 1.2% to 8.0 ± 1.4%; the adjusted between-group difference was −0.9 percentage points (95% CI −1.1 to −0.6; P < 0.001). Time-in-range (70–180 mg/dL) was 21 percentage points higher in the CGM group versus controls (CI 16–26). Safety not only met but exceeded expectations: time below 70 mg/dL was 0.20 percentage points lower in the CGM group (non-inferiority P < 0.001; superiority P = 0.004), and no severe hypoglycemic events occurred throughout the trial.
Funding sources need not be concealed: this trial was funded by sensor manufacturer Dexcom, and five authors’ affiliations are listed as Dexcom. This does not detract from its status as the highest-quality evidence currently available for this population—multicenter, randomized, with HbA1c as the primary endpoint and positive results—but readers must bear two things in mind: who paid for it, and that follow-up lasted only 26 weeks; long-term outcomes and cost-effectiveness remain unassessed.
## Layer Two: What Is Novel About CGM’s New Metrics?
One selling point of CGM is a suite of new metrics: Glucose Management Indicator (GMI) and Time-in-Range (TIR). If these are merely conversions of HbA1c, their prognostic value is limited. The GRADE-CGM substudy, published in the December 2026 issue of the Journal of Diabetes and Its Complications, directly compared them: among 1,363 participants with type 2 diabetes who wore blinded CGM for at least 10 days, median follow-up was 2.3 years and 74 cardiovascular events or deaths occurred. GMI (HR 1.44 per 1% increase [CI 1.21–1.71]), HbA1c (HR 1.37 per 1% increase [1.16–1.61]), and TIR (HR 0.73 per 20% increase [0.61–0.87]) were all associated with cardiovascular events or death. For kidney outcomes (annual eGFR decline of 2.37 mL/min/1.73m²), both GMI and HbA1c correlated with rate of decline, whereas TIR did not.
The critical final step: after mutual adjustment among metrics, most associations attenuated—indicating highly overlapping prognostic information. The authors concluded that GMI and HbA1c show comparable associations with cardiovascular and kidney outcomes, while TIR associates only with cardiovascular outcomes. This is an observational substudy; attenuation after mutual adjustment does not prove which metric is more fundamental, but the direction is clear: CGM’s clinical value likely stems from behavioral feedback and tighter glycemic control itself—the first-layer trial showed CGM users spent 21 percentage points more time in target range and achieved greater HbA1c reduction—not from some novel prognostic information unavailable from HbA1c. Among the substudy’s authors, one had previously received Dexcom-donated devices for an unrelated study, and another served on Amgen’s Data Safety Monitoring Board and received royalties from UpToDate; both disclosures appear in the paper.
## Layer Three: Does CGM Benefit Healthy Individuals?
This is the segment where the consumer market has expanded most rapidly—and where evidence remains thinnest. A multidisciplinary expert workshop convened by the International Alliance for Human Nutrition Science (IAFNS) in May 2025 issued a position statement, published online ahead of print in Advances in Nutrition in November 2026. It first acknowledges the rationale: postprandial glucose responses vary enormously between and within healthy individuals, driven by complex interactions among diet composition, lifestyle, and physiological traits; CGM can capture these dynamics under real-world conditions and may help identify meaningful individual differences. Yet experts list four major interpretive limitations—analytic variability, insufficient precision for single-meal or acute assessments, lack of standardized study protocols and consensus guidelines, and, most critically, insufficient evidence linking CGM metrics to long-term clinical outcomes. The conclusion is notably restrained: until high-quality randomized trials and consensus guidelines emerge, CGM use outside diabetes care should be approached cautiously—and ideally embedded within structured research or clinical settings offering appropriate education and professional supervision.
A systematic review published in Medicina in September 2026 quantifies the current state of evidence: screening 6,265 records from seven databases plus two trial registries yielded only 12 studies examining the behavioral and psychosocial impact of "participant-visible CGM feedback" on adults without diabetes. Heterogeneity was so great that meta-analysis was impossible. Positive signals centered on dietary behavior: in one 8-week trial, CGM plus low-glycemic-index/load education versus education alone reduced carbohydrate intake by 15.2 g/day (CI −20.9 to −9.5), lowered glycemic index by 10.0, and reduced glycemic load by 19.9; no clear difference emerged for physical activity (−28.5 MET-minutes/week, CI −171.9 to 114.8); evidence for psychosocial outcomes was insufficient to confirm benefit or rule out harm. Certainty of evidence: low for dietary outcomes, very low for all other domains. The review authors conclude succinctly: participant-visible CGM may serve as a biofeedback aid supporting certain dietary changes within structured education or lifestyle interventions—but substantial, durable benefits and independent efficacy remain unestablished. The review authors declared no conflicts of interest.
## My Three Judgments
Assembled across these three layers, the following three judgments constitute this article’s synthesis—not conclusions drawn from any single source.
First, indication expansion has now attained its appropriate evidentiary form. For non-insulin-treated type 2 diabetes—the largest incremental population—there now exists a multicenter randomized trial published in NEJM with HbA1c as the primary endpoint. Evidence for payers and primary care systems to make coverage decisions has fundamentally shifted; however, expected benefits must be calibrated to the trial’s observed magnitude—approximately 0.9 percentage-point HbA1c reduction and 21-percentage-point increase in time-in-range—not to promotional claims.
Second, the metrics layer cools the narrative that “more data is always better.” GMI and TIR convey prognostic information highly overlapping with HbA1c; CGM’s value anchor lies in behavior change and improved control itself—not in mysterious new metrics.
Third, the gap between the consumer market for healthy individuals and the evidence base is real. Manufacturers’ and influencers’ narratives position CGM as a universal metabolic optimization tool, whereas expert position statements demand caution and professional supervision, and systematic reviews support only biofeedback assistance within structured education—no independent efficacy has been established. What is sold to healthy individuals is closer to a self-quantification experience; it may be a legitimate consumer choice, but it should not masquerade as clinically proven benefit.
## What You Can Take Away
If you are a person with type 2 diabetes not using insulin: this trial supports discussing CGM as one option for primary care management—with awareness of both industry funding and the 26-week follow-up limitation.
If you are a payer or hospital: the evidentiary basis for coverage decisions has updated, but benefit expectations must align with the trial’s observed magnitude.
If you are a healthy individual intrigued by glucose curves: current evidence supports CGM use only alongside structured education; durable benefit from standalone use remains unproven, and expert guidance emphasizes caution.
If you are a researcher or manufacturer: the next step for CGM in healthy populations is not marketing—it is long-term outcome trials and standardized protocols. The IAFNS position statement has already laid out the requirements list.
The truly debatable question is this: does the glucose monitor sold directly to consumers include, inside its packaging, the structured education and professional supervision that experts require? Join the discussion in the comments.
Source: CONNECT randomized trial (N Engl J Med 2026;395:1717–1727). https://doi.org/10.1056/NEJMoa2608571