What the research found

Researchers analysed CGM data from over 62,000 adults without diabetes using a commercial app, comparing what their glucose patterns actually predicted against conventional markers like HbA1c. They found that average glucose—a simple rolling mean—was far more stable week-to-week than other CGM metrics and correlated meaningfully with markers of metabolic health including fasting insulin, triglycerides, HDL, uric acid, and inflammation. Notably, average glucose explained roughly three times more of the variation in fasting insulin than HbA1c did.

A striking finding emerged when they looked at people with normal HbA1c but elevated average glucose (≥110 mg/dL): these individuals showed hidden hyperinsulinemia (fasting insulin ≥10 µIU/mL) at roughly double the rate of those with lower average glucose. This suggests conventional screening misses a subset of metabolically stressed people. Unexpectedly, the width of glucose spikes—not their height or how often they occurred—was what actually tracked insulin resistance and inflammation. Individual carbohydrate sensitivity, though a repeatable measurement, didn't correlate with insulin resistance.

Why it matters for you

If you're tracking CGM to optimise metabolic health, this gives you a clearer lens on what to actually watch. Your average glucose level appears to be the most reliable signal—more useful than obsessing over individual spike heights. If yours trends above 110 mg/dL even with a normal HbA1c, this work suggests you may already have subclinical hyperinsulinemia worth addressing, even if your doctor isn't concerned. That's actionable: it points to insulin sensitivity as a target, not just carbohydrate avoidance.

For protocol design, the finding on spike width is interesting. It hints that how long your glucose stays elevated matters more for metabolic stress than how high it spikes. This could shift focus toward sustained postprandial clearance (perhaps via movement timing, fibre, or sequencing) rather than simply minimising peak magnitude. If you're also tracking fasting insulin, triglycerides, or hs-CRP, correlating those against your average glucose and spike kinetics rather than HbA1c may give you earlier warning of drift.

Caveats

• Observational, cross-sectional design: no causal claims; correlation with biomarkers doesn't prove metabolic harm from elevated glucose
• Self-selected cohort: Levels users are health-conscious and tech-adopting, not representative of the general population
• Laboratory data not all participants: only ~7,930 of the 62,122 had matched blood work, reducing generalisability of biomarker associations
• Commercial platform: industry-funded study; potential bias toward framing CGM as useful in non-diabetics
• Early framework: the authors position this as a proposed interpretation scheme, not a validated clinical tool; thresholds (e.g. 110 mg/dL) should not be treated as hard cutoffs yet