Across 62,122 free-living adults wearing consumer continuous glucose monitors, average glucose was the steadiest metric from week to week (intraclass correlation 0.747). In 7,930 non-diabetic participants with lab data, it tracked insulin resistance, triglycerides, HDL cholesterol, uric acid and C-reactive protein more closely than HbA1c, explaining nearly three times the variance in fasting insulin (R2 0.131 vs 0.048). Among people with normal HbA1c, fasting insulin of 10 µIU/mL or higher appeared in 27.5% of those averaging 110 mg/dL or above, versus 12.6% of those below. At equal average glucose, spike width, not height or frequency, tracked insulin resistance, triglycerides and CRP.
The practical message is that a normal HbA1c can mask early metabolic dysfunction, and a sustained glucose plateau may carry more information than dramatic peaks. That challenges the popular habit of fixating on individual spikes, and the finding that carbohydrate sensitivity did not track insulin resistance cuts against a common wearable-marketing claim. Important caveats apply. The design is cross-sectional, so it shows association rather than prediction, and no hard outcomes such as diabetes or cardiovascular events were measured. Users of a paid wellness platform are self-selected, and even the best R2 of 0.131 leaves most variance unexplained, so this is no replacement for fasting insulin or standard screening. This is also a preprint that has not been peer reviewed, and its conclusions may shift. Overall, it is a useful incremental step toward interpreting CGM outside diabetes, not a paradigm shift, and it needs longitudinal confirmation.