Metabolic health beyond BMI: insulin resistance and what to test

The BMI guide made the case that BMI predicts population outcomes well despite being a crude measure. This guide covers the most important thing it misses: two people at the same BMI can be in substantially different metabolic condition, and metabolic condition is closer to what actually drives disease.

Why BMI and metabolic health come apart

Research on metabolic health has identified two populations that complicate the simple weight-equals-risk story.

Metabolically healthy obesity. A minority of people with obesity show normal insulin sensitivity, normal blood pressure, normal lipids, and no evidence of fatty liver. Their near-term cardiometabolic risk is considerably lower than others at the same BMI. The important caveat, from longer follow-up studies, is that this state is often transitional rather than stable — a substantial share convert to metabolically unhealthy over time — so it's better read as a favourable current position than a permanent exemption.

Metabolically unhealthy normal weight. The converse, sometimes described with the unhelpful phrase "thin outside, fat inside." People with BMI in the normal range but significant visceral and hepatic fat, insulin resistance, and adverse lipids. Their risk is elevated despite a reassuring BMI, and because nothing about their weight prompts investigation, they're frequently missed.

The variable doing most of the work in both cases is where fat is stored, not how much there is. Subcutaneous fat on hips and thighs is comparatively benign metabolically. Visceral fat around the organs and fat accumulating inside the liver and muscle are strongly associated with insulin resistance and inflammation. BMI cannot distinguish these, which is the single largest reason it misclassifies individuals.

What insulin resistance actually is

Insulin's job is to signal cells to take glucose out of the blood. Insulin resistance means that signal works less well, so the pancreas compensates by producing more insulin to achieve the same effect.

The compensation is why the condition stays hidden for so long. For years — often many years — blood glucose remains normal because rising insulin output offsets the falling response. Measuring glucose alone during this phase shows nothing wrong. What has actually changed is that the system is working much harder to produce the same result.

Eventually compensation becomes insufficient. Fasting glucose drifts up, then post-meal glucose, then diagnostic thresholds are crossed. By the time type 2 diabetes is diagnosed, the underlying process has typically been developing for a long time and beta-cell function has often already declined substantially.

Chronically elevated insulin isn't metabolically neutral either. It promotes fat storage, contributes to sodium retention and raised blood pressure, and is associated with adverse lipid patterns — which is why insulin resistance tends to arrive with a cluster of other findings rather than alone.

The long silent progression

The practical significance of that timeline is that the window for intervention is long and mostly unmonitored:

Stage	What's happening	What routine testing shows
Early insulin resistance	Insulin output rising to compensate	Normal glucose, normal HbA1c
Compensated	High insulin holding glucose in range	Still often normal; fasting insulin raised if measured
Prediabetes	Compensation failing	Raised fasting glucose and/or HbA1c
Type 2 diabetes	Beta-cell function substantially reduced	Clearly abnormal

Two implications follow. Progression is not inevitable — trial evidence, notably from the Diabetes Prevention Program, shows that intensive lifestyle intervention in people with prediabetes substantially reduces conversion to diabetes, and in that trial outperformed metformin. And earlier detection is more useful than later, because reversal gets harder as beta-cell function declines.

The markers worth knowing

If you're getting blood work done, these are the ones relevant to this picture:

• HbA1c. Glycated haemoglobin — reflects average blood glucose over roughly the preceding three months. The most practical single marker; no fasting required.
• Fasting glucose. Standard, widely available, but a late signal for the reasons above.
• Fasting insulin. Rises earlier than glucose, so potentially a much earlier warning. Not routinely ordered, and reference ranges are less well standardised.
• Triglyceride-to-HDL ratio. Computable from a standard lipid panel you may already have, and correlates reasonably with insulin resistance.
• ALT and liver markers. Elevated ALT without another explanation may indicate fatty liver, which is closely tied to insulin resistance.
• ApoB or non-HDL cholesterol. For cardiovascular rather than glycaemic risk, these are generally considered better measures of atherogenic particle burden than LDL cholesterol alone.

We're describing what these markers mean, not recommending a testing protocol — what's appropriate depends on your age, history, and risk factors, and that's a conversation for a clinician rather than a calculator.

Why HbA1c is the practical one

Of the list above, HbA1c is the one most worth knowing, for reasons of practicality rather than superiority. It doesn't require fasting, it isn't distorted by what you ate yesterday, it's cheap and near-universally available, and its diagnostic thresholds are well standardised internationally.

Its limitations are worth knowing too: it can be misleading in the presence of anaemia, haemoglobin variants, recent blood loss, or altered red-cell turnover, since it depends on red blood cells having a typical lifespan. And as an average, it can conceal considerable post-meal variability.

What actually improves it

The encouraging part: insulin sensitivity responds to intervention faster than almost anything else discussed on this site.

Exercise has both an acute and a chronic effect. A single session of moderate exercise improves insulin sensitivity measurably for a day or more, partly through glucose uptake pathways that don't require insulin at all. Regular training raises the baseline. Resistance training contributes independently by increasing the muscle mass available to dispose of glucose.

Visceral fat loss improves it, and visceral fat is preferentially mobilised early in weight loss — which is why metabolic markers often improve well before body weight changes much. Liver fat in particular can fall substantially within weeks of modest energy restriction.

Sleep matters more than people expect. Experimental sleep restriction reduces insulin sensitivity in healthy young adults within days, making this one of the more direct links between two of this site's factors.

Dietary pattern matters, though the specifics are contested. Reducing refined carbohydrate and ultra-processed food is well supported; beyond that, claims about optimal macronutrient ratios outrun the evidence.

Why the calculator asks about BMI instead

The same trade as elsewhere. BMI needs two numbers everyone knows. Metabolic markers require blood work most users don't have to hand, and a model input that most people leave blank or guess at is worse than a rough one they can answer honestly.

There's also the double-counting problem: exercise, sleep, diet, and BMI all feed into metabolic health, so adding it as a separate additive modifier in a model that sums its factors independently would count the same effect twice.

What this means for reading your result: your BMI band is a proxy for metabolic risk, and proxies fail in individual cases. If you have recent blood work showing good metabolic markers despite a BMI above 25, your real risk is likely better than the model implies. If your BMI is normal but your markers are poor — or you have a strong family history of type 2 diabetes and have never been tested — it's likely worse. The blood work is the better information, and where they disagree, believe it over us.