Continuous glucose monitors improve health outcomes in people without diabetes

Cross-sectional analysis of 972 adults aged 40+ (421 with type 2 diabetes, 319 with prediabetes, 232 normoglycaemic) who wore a Dexcom G6 for up to 10 days, comparing eight CGM metrics against laboratory HbA1c. The association between CGM metrics and HbA1c was strong in type 2 diabetes (mean glucose standardised β = 0.79, p < 0.001), substantially attenuated in prediabetes (β = 0.22), and essentially absent in normoglycaemic participants — CGM metrics were largely unrelated to HbA1c in people without diabetes. The practical implication is that the numbers an OTC CGM shows a healthy user do not track the one glycaemic measure with established prognostic value, so "improving your CGM numbers" is not the same thing as improving glycaemic health.

The ADA's annually updated Standards of Care is the most widely used professional guideline on glucose monitoring. Its 2026 diabetes-technology section broadens CGM recommendations considerably — CGM is now recommended from diagnosis onward for adults on insulin, on non-insulin therapies that can cause hypoglycaemia, and on any therapy where CGM aids management — but every graded recommendation is confined to people who have diabetes. On over-the-counter devices it says only, descriptively, that "anyone can purchase the OTC-CGM devices, including those without diabetes or with prediabetes who wish to assess their glycemic responses to their lifestyles, including the effects of food choices and exercise." It issues no recommendation, no interpretive framework and no target ranges for people without diabetes.

Randomised crossover trial in 15 healthy adults (9 female, 6 male) without diabetes who each completed 7 laboratory visits, with glycaemic responses to test foods and drinks measured simultaneously by CGM (Abbott FreeStyle Libre 2) and by capillary fingerprick sampling as the criterion method. CGM overestimated glucose by 0.9 ± 0.6 mmol/L when fasting and 0.9 ± 0.5 mmol/L postprandially, and overestimated time spent above the 7.8 mmol/L threshold by roughly 4-fold (~2-fold after adjusting for baseline offset). The bias was not a fixed offset: between-participant SD of the bias was 0.6 mmol/L, and CGM reclassified a commercial fruit smoothie from a low glycaemic index of 53 (capillary) to a medium/high GI of 69. In healthy people, the device therefore manufactures "spikes" and out-of-range time that do not exist in blood.

Systematic review and meta-analysis of 25 randomised controlled trials (n = 2,996) testing CGM feedback as a behaviour-change tool in populations with and without diabetes. Pooled across all populations, CGM arms achieved a modest HbA1c reduction of 0.28% (95% CI 0.15–0.42) and +7.4% time in range, but there were no significant effects on body weight (−0.7 kg) or BMI (−0.4 kg/m²), and only 5 of 25 trials measured physical activity (results mixed, no consistent improvement). Crucially, the population was overwhelmingly diabetic (~68% type 2 diabetes) and only 3 of the 25 trials enrolled people without diabetes (all in overweight/obesity); the authors state that more studies are needed "particularly in subgroups that have been minimally investigated (e.g., participants without diabetes)" and that different outcome measures may be more appropriate in that group.

Method-comparison study in 34 healthy, normoglycaemic adults (mean age 65.7 y) in whom glucose was measured simultaneously every 10 minutes for 24 hours by an Enlite CGM sensor and by venous blood sampling. Agreement was mediocre: the mean absolute relative difference (MARD) was 17.6% (SD 17%) and the median within-person correlation between CGM and venous glucose was only 0.68 (IQR 0.40–0.78) — far worse than the ~9–10% MARD that CGMs achieve in the diabetic range they are validated against. CGM also systematically read 0.22 mmol/L higher than venous blood during the day and understated glycaemic variability (24-h SD 1.07 by CGM vs 1.26 by venous sampling, p = 0.004). The authors' conclusion of "good agreement" is generous given the error magnitude relative to the narrow glucose range healthy people actually occupy.

Mixed-methods cross-sectional survey of 56 adults (with and without diabetes, none on insulin) who had used a CGM in the previous year, measuring CGM-related distress alongside psychological traits and thematic analysis of open-ended responses. More than two-thirds (68%) reported fear of developing type 2 diabetes when they saw an elevated glucose reading — regardless of whether they actually had diabetes — and greater distress was associated with younger age and obesity (p < 0.01), while higher eating-disorder symptom scores correlated with distress about the device (r > 0.3, p < 0.01). At the same time 89% reported making positive dietary and/or physical-activity changes, and the qualitative data showed distress often coexisting with behaviour change. Small, self-selected, retrospective sample, so it documents that the harm signal is real and common among users rather than quantifying its incidence.