Is BMI accurate?

BMI is the most widely used health metric in the world and one of the most widely misunderstood. It is not useless — but it answers a much narrower question than people think it does.

This article is general information, not medical advice. Any decision about your health belongs with a qualified clinician who can see the whole picture.

What BMI actually measures

Body Mass Index is weight in kilograms divided by height in metres squared. That is all it is: a ratio of mass to height.

The conventional adult categories are under 18.5 underweight, 18.5 to 24.9 healthy weight, 25 to 29.9 overweight, and 30 and above obese. The BMI calculator works it out in metric or imperial.

Notice what is not in that formula. It does not know what your weight is made of. Muscle, fat, bone and water all count the same. It does not know where the weight sits, and it does not know your age, sex, fitness or medical history.

The squaring is also an approximation. Bodies do not scale as neatly as the formula assumes, which is why BMI tends to overestimate for tall people and underestimate for short people, all else being equal.

It was never designed for individuals

The formula was devised in the 1830s by Adolphe Quetelet, a Belgian mathematician and astronomer. He was not a physician and he was not studying health. He was developing statistics about the distribution of characteristics across populations, and he said explicitly that the ratio should not be used to judge an individual.

It acquired its current name and role in 1972, when the physiologist Ancel Keys evaluated it against other measures and found it the best of the simple options for population studies. It spread because it is cheap: it needs a set of scales and a tape measure, which is what makes it practical for a survey of a hundred thousand people.

The point is not that BMI is a fraud. It is that a tool built to describe groups is being used to categorise individuals, and it was known from the beginning that it does that badly.

The people it gets wrong

Muscular people. Muscle is considerably denser than fat, so athletes routinely land in the overweight or obese categories with very low body fat. Most professional rugby players and many sprinters are technically obese by BMI. This is the most cited failure, and it is genuine — though it applies to fewer people than the internet suggests.

Older adults. People lose muscle and bone density with age, so someone can hold a steady BMI for decades while their body composition changes substantially. A BMI in the healthy range can mask sarcopenia — dangerous muscle loss.

People with "normal weight obesity". This is the failure that matters most and gets discussed least. Someone can sit squarely in the healthy BMI range while carrying a high proportion of body fat and little muscle, with the associated metabolic risk. BMI marks them as fine and nobody looks further.

Very short and very tall people. The squared term skews both directions, as above.

Pregnant people, children and anyone with significant fluid retention. Standard adult BMI simply does not apply. Children are assessed on age-and-sex percentile charts instead, and comparing a child to adult categories is meaningless.

Ethnicity changes the thresholds

This is well established and still not widely known.

The standard categories were derived largely from European populations. People of South Asian, Chinese and other Asian descent tend to carry more visceral fat and develop cardiometabolic risk at lower BMIs — so the same number represents a different level of risk.

Health bodies have responded. The UK's NICE guidance recommends lower thresholds for people of South Asian, Chinese, Black African and African-Caribbean family background, treating 23 as the point of increased risk rather than 25. The World Health Organization has published similar adjusted cut-offs for Asian populations.

Conversely, some research suggests the standard thresholds may overstate risk for Black populations, who on average have greater bone density and muscle mass.

A single set of numbers applied to everyone is the problem here, not the arithmetic.

Measurements that tell you more

Waist circumference is the most useful cheap addition, because where fat sits matters more than how much there is. Visceral fat around the organs carries far more metabolic risk than fat on hips and thighs. General guidance puts increased risk above roughly 94 cm for men and 80 cm for women, with lower thresholds for South Asian populations.

Waist-to-height ratio may be better still, and has the advantage of a single memorable rule: keep your waist under half your height. It needs no chart and no adjustment for units. The unit converter handles the arithmetic if your tape measure and your height are in different systems.

Body composition — DEXA scanning, or bioelectrical impedance with a large pinch of salt — measures the thing BMI is proxying for. DEXA is accurate and not cheap; consumer scales that claim body-fat percentage are frequently unreliable.

Blood markers and blood pressure describe metabolic health directly, and are what a clinician will actually act on.

Where it is still useful

Having said all that, dismissing BMI entirely is an overcorrection.

It remains genuinely good at what it was built for: comparing populations, tracking trends over time, and epidemiological research where measuring body composition for a hundred thousand people is impossible. Very large studies do show meaningful associations between BMI ranges and health outcomes at the population level.

For an individual it is best treated as a screening flag, not a diagnosis — a cheap first look that says whether a conversation is worth having. A BMI of 34 in someone who does not lift weights is worth discussing. A BMI of 26 tells you almost nothing on its own.

It is also reasonable for tracking yourself over time, because your height is fixed and most of its weaknesses are about comparing different people rather than the same person across months.

Use it as one number among several, and take any conclusion to someone qualified to interpret it. Pikkit has the rest of the calculators, and Calculators & Time collects them.