Understanding BMI: What the Number Measures, and What It Misses

Body Mass Index is weight divided by height squared. That is the whole formula, and its simplicity is both why it survived two centuries and why it fails in predictable ways.

The 703 is purely a unit conversion. It turns pounds per square inch into kilograms per square metre. There is no biology in it.

Where it came from

Adolphe Quetelet, a Belgian astronomer and statistician, published the ratio in 1832 as part of his work on l’homme moyen: the average man. He was building a statistical description of populations. He said explicitly that it was not a measure of individual fatness.

It stayed obscure until 1972, when Ancel Keys tested several height-weight ratios against body fat measurements across thousands of men and found Quetelet’s the best of a mediocre set. He renamed it the Body Mass Index and recommended it (again explicitly) for population studies rather than individual diagnosis.

Insurers and clinicians adopted it anyway, because it is free, instant, and requires no equipment beyond scales and a wall.

The standard categories

BMI	WHO category	Asian-Pacific cutoff
Under 18.5	Underweight	Under 18.5
18.5 to 24.9	Normal weight	18.5 to 22.9
25.0 to 29.9	Overweight	23.0 to 27.4
30.0 to 34.9	Obese class I	27.5 to 32.4
35.0 to 39.9	Obese class II	32.5 to 37.4
40.0 and above	Obese class III	37.5 and above

The boundaries are round numbers, and that should tell you something. There is no biological discontinuity at 25.0. Risk rises as a smooth curve; the categories are administrative conveniences laid over it. A BMI of 24.9 and one of 25.1 are the same person on two different days.

The problem with squaring height

Human bodies are three-dimensional. If people scaled up perfectly, mass would rise with height cubed, and the correct denominator would be m³. But taller people are not simply scaled-up shorter people. They are proportionally more slender. Empirically the best exponent for adult populations falls somewhere between 2 and 3, usually around 2.3 to 2.7.

Quetelet chose 2 because it fit his Belgian data adequately and was easy to compute by hand. The consequence is a systematic bias:

• Tall people get BMI values that overstate their adiposity.
• Short people get values that understate it.

Nick Trefethen at Oxford proposed a corrected version in 2013 using an exponent of 2.5 with a scaling constant. It has better statistical properties and essentially no clinical adoption, because changing a universally-used index is harder than living with a known bias.

The four things it cannot see

Muscle versus fat

The scale measures mass. It cannot distinguish tissue types. A 100 kg bodybuilder at 8% body fat and a 100 kg sedentary person at 35% body fat, both 1.8 m tall, have identical BMIs of 30.9. Their health risks are not remotely comparable.

Where the fat is

Visceral fat (around the organs) is metabolically active and strongly associated with cardiovascular and metabolic disease. Subcutaneous fat on the hips and thighs is far less so. Two people with identical BMIs can carry the same fat mass in places with quite different risk profiles. BMI is blind to this entirely.

Age

Adults lose muscle and bone with age while often gaining fat. Someone at 70 with the same BMI they had at 30 typically has considerably more body fat. Several large studies have also found the lowest all-cause mortality in older adults sits at a higher BMI than in younger ones: a finding that remains debated, but which at minimum argues against applying one set of thresholds across the whole adult lifespan.

Ancestry

Body composition at a given BMI varies systematically between populations, which is the entire reason the Asian-Pacific cutoffs exist. Applying a single global threshold misclassifies risk in both directions.

What to measure alongside it

Waist-to-height ratio is the highest-value addition, and it costs one tape measure. Measure your waist at the midpoint between the lowest rib and the top of the hip bone, then divide by your height in the same units.

The guidance is memorable: keep your waist under half your height. Under 0.5 is generally healthy; 0.5 to 0.6 warrants attention; above 0.6 is associated with substantially increased risk. It captures fat distribution, which is precisely what BMI cannot, and several meta-analyses have found it a better predictor of cardiometabolic outcomes than BMI alone.

Waist circumference on its own is also used clinically, with thresholds around 94 cm for men and 80 cm for women in European populations, adjusted downward for South Asian populations.

Body fat percentage from a circumference method is a rough estimate but a useful one, especially for tracking direction. The body fat calculator implements the US Navy formula, which needs a tape measure and nothing else.

A reasonable way to use BMI

Treat it as a first screen with a wide error bar, not a verdict.

1. Track your own trend rather than comparing yourself to a threshold. A BMI moving from 27 to 25 over a year is meaningful information about you. A BMI of 25 versus someone else’s 24 is noise.
2. Pair it with a waist measurement. The combination catches the case where BMI is normal but visceral fat is high, sometimes called normal-weight obesity, and easy to miss otherwise.
3. Discount it if you are heavily muscled, very tall, very short, or over 65. Not to zero, but with a known direction of bias.
4. Do not use adult BMI for children. Paediatric assessment uses age-and-sex-specific percentile charts, because the relationship between height, weight and body composition changes throughout growth.
5. Take it to a clinician rather than acting on it. A single ratio is a prompt for a conversation, not a substitute for one.

The BMI calculator handles both unit systems and shows which category the result falls into. What it cannot do (what no calculator can) is tell you whether that category means anything for you specifically.