What Changes Before Your Blood Sugar Does

A claim circulates widely in metabolic health content: insulin resistance is detectable ten or twenty years before type 2 diabetes is diagnosed. The underlying research is real, frequently cited, and does not say that.

The Whitehall II study followed 6,538 British civil servants who were free of diabetes at baseline, and 505 of them went on to be diagnosed. Because the cohort was assessed repeatedly, the researchers could do something unusual: work backwards from each diagnosis and reconstruct what the preceding years actually looked like.

Fasting glucose in the group who developed diabetes was already drifting upward relative to controls more than a decade out, but slowly — on the order of 0.028 mmol/L per year, a gradient far too shallow to notice on any individual result. The sharp inflection came late. Fasting glucose turned steeply upward starting about three years before diagnosis, and two-hour post-load glucose followed the same three-year pattern.

Insulin sensitivity, measured by HOMA, declined steeply earlier — over roughly the last five years. Beta-cell function did something more interesting still: it increased between four and three years before diagnosis, compensating for rising resistance, and then fell away as that compensation failed.

So the sequence is real. The body works progressively harder to hold glucose in range, and that effort is measurable before the number it is defending breaks. The lead time in this data is about two years, not twenty.

Two years is still worth having. Those are precisely the years in which the trajectory is most amenable to change, and the difference between catching a trend and catching a diagnosis.

The limitations are important. This cohort was 71% male, 91% white, and entirely British civil servants, which is not a general population. The curves are population averages with heavy overlap between groups — this is a description of how a disease develops, not a test that reliably flags an individual. And critically, HbA1c was not the comparator in this analysis, so this is not evidence that fasting insulin beats HbA1c as a screening tool.

That last point deserves emphasis, because fasting insulin and HOMA-IR are frequently promoted as superior early screening. They are not standardised across laboratories, have no agreed diagnostic cut-off, and are not recommended for screening by any major diabetes guideline. There is no validated "optimal" HOMA-IR figure to aim at.

Why this matters for tracking: the useful application here is trend, not threshold. Fasting insulin, fasting glucose and HbA1c drawn from the same laboratory over several years tell you which direction your metabolic control is moving, and roughly how fast. A single value in isolation — particularly a fasting insulin — carries far less information than most content implies, and is best interpreted alongside the rest of the panel rather than as a standalone verdict.