What BMI actually measures — and what to use instead

The formula is the easy bit

BMI = weight (kg) / height (m)². A 1.75 m, 75 kg person has a BMI of 24.5. That is it. Two numbers in, one number out. The classification bands (underweight, normal, overweight, obese I/II/III) are statistical cut-offs from population studies, not biological thresholds.

What BMI actually predicts

At a population level, BMI is a useful proxy for population-level mortality risk. A BMI of 18.5–25 in large cohorts is associated with the lowest all-cause mortality. A BMI above 30 is associated with meaningfully higher risk of type 2 diabetes, hypertension, cardiovascular disease, and several cancers. That is the level at which the metric works. The doctor at the NHS, looking at 30,000 patients, can use BMI to spot the cohort that needs further investigation. It is a sieve, not a diagnosis.

What BMI is bad at

It is bad at almost everything that matters to a single person. Three specific cases:

1. Muscle versus fat

BMI does not distinguish muscle mass from fat mass. A lean, muscular 90 kg man at 1.83 m has a BMI of 26.9 — "overweight" by the chart. An inactive 90 kg man with 35% body fat at the same height has the same BMI. They are not at the same health risk.

2. Ethnicity

The cut-offs were derived from white European populations. The relationship between BMI and body-fat percentage, and between BMI and disease risk, differs across ethnic groups. The NICE guidelines in the UK recommend lower BMI cut-offs for South Asian populations, because disease risk starts at a lower BMI in those groups.

3. Age and sex

The "ideal" BMI band is similar across adults, but body composition changes with age. A 70-year-old with a BMI of 27 is not at the same risk as a 30-year-old with a BMI of 27. And women and men carry fat differently — women typically have a higher body-fat percentage at the same BMI.

What to use instead, or alongside it

For personal health decisions, BMI is the start of a conversation, not the end. Better metrics, in rough order of cost: waist circumference and waist-to-hip ratio (cheap, fast, and visceral fat — the dangerous kind — lives around the waist), body-fat percentage (via skinfold calipers, a bioelectrical impedance scale, or a DEXA scan), blood markers (fasting glucose, HbA1c, lipid panel, blood pressure), and cardiorespiratory fitness (VO₂ max is a stronger predictor of all-cause mortality than BMI).

When BMI is enough

If you fall squarely in the 18.5–25 range and have no other risk factors, you don't need a DEXA scan. The simple metric worked. If you fall outside the range, BMI tells you to look further, not to panic. The reason the doctor still uses it is exactly that: it is the cheapest tool in the drawer that catches most of the cases that need catching.