Is BMI accurate?

What it was built for

BMI is weight in kilograms divided by height in metres squared. It was devised in the 1830s by Adolphe Quetelet, a Belgian mathematician and astronomer who was explicitly studying the statistical properties of populations, not assessing patients. He said so at the time.

It came into medical use in the 1970s largely because it was the best of the available options for a specific job: comparing large groups cheaply. It requires two measurements anyone can take, needs no equipment beyond scales and a tape, and correlates well enough with body fatness across a population to be informative about trends.

For that purpose it works. Population-level BMI tracks obesity-related disease rates well, and public health research would be considerably harder without it. The criticism is not that the number is meaningless; it is that a statistic designed to describe a distribution behaves poorly when applied to one point in it.

The category boundaries are also more arbitrary than they look. The cutoffs at 18.5, 25 and 30 are round numbers chosen for convenience, and the risk they describe rises continuously rather than stepping at those points. A BMI of 24.9 and one of 25.1 are not meaningfully different, despite falling in differently named categories.

Where it fails

The formula uses only weight and height, so it cannot distinguish what the weight is made of. Muscle is considerably denser than fat, which means a well-trained person routinely records a BMI in the overweight or obese range while carrying low body fat. This is the criticism most often raised and it is genuinely correct — though it applies to fewer people than the number of people who invoke it.

The opposite failure is less discussed and more common. Someone with little muscle and a high proportion of fat can record a perfectly normal BMI while carrying the metabolic risk associated with excess fat. This is sometimes described as normal-weight obesity, and because the number looks reassuring it tends not to be investigated.

The formula also scales badly with height. Dividing by height squared was a convenience rather than a derivation, and bodies do not scale as squares — very tall people tend to record higher BMIs and very short people lower ones for the same build. This is why the same threshold applied across a whole population misclassifies at both ends.

Population differences matter too. The relationship between BMI and health risk varies between ethnic groups, and several health bodies use a lower obesity threshold for people of South Asian, Chinese and other Asian descent because metabolic risk appears at a lower BMI. Applying one set of cutoffs everywhere builds in a systematic error.

What to measure instead, or as well

Waist circumference adds the single most useful piece of information BMI lacks: where fat is stored. Visceral fat around the organs carries substantially more metabolic risk than the same mass on hips and thighs, and BMI cannot see the difference at all. A tape measure at the navel costs nothing.

Waist-to-height ratio is the simplest way to use it, and the guidance is easy to remember: keep your waist under half your height. It applies across adult heights and sexes without separate tables, and several studies find it predicts cardiometabolic risk better than BMI does.

Body fat percentage is more directly relevant than either, and harder to measure well. Skinfold calipers and the Navy tape method are cheap and roughly repeatable; bioelectrical impedance scales are convenient and vary with hydration enough that the trend matters more than any single reading. DEXA is accurate and requires a clinic.

The pragmatic combination for most people is BMI plus waist. Two numbers, no equipment, and between them they catch both the athlete misclassified as overweight and the sedentary person with a reassuring BMI and a high waist. Either alone misses one of those cases.

How to read your own number

Treat BMI as a screening prompt rather than a diagnosis. A result outside the normal range is a reason to look at other measurements, not a conclusion in itself, and a result inside it is not evidence that nothing needs attention.

Ask what the weight is made of. If you carry visible muscle and train regularly, an overweight BMI probably reflects that, and body fat percentage or waist measurement will say so. If you do not, a normal BMI with a large waist is the more likely picture and the more actionable one.

Watch the direction rather than the value. A BMI moving steadily upward over years carries information regardless of which category it currently sits in, and it is easier to act on early. The same applies to waist measurement, which often changes before weight does.

One consequence of the smooth curve is worth stating plainly: the lower boundary matters as much as the upper one. A BMI below 18.5 carries its own risks — reduced bone density, impaired immune function and, in older adults, a markedly worse prognosis after illness or surgery. Public attention sits almost entirely on the high end, which means underweight results are more often left uninvestigated than overweight ones.

It is also worth knowing what BMI does when age enters the picture. Muscle mass falls steadily from around the fourth decade while fat tends to rise, so an unchanged BMI across twenty years usually conceals a substantially changed body composition. The number stays still precisely while the thing it is standing in for moves, which is the clearest illustration of why it is a proxy rather than a measurement.

And treat the boundaries with appropriate scepticism. Crossing from 24.9 to 25.1 changes the label and almost nothing else. The risk curve is smooth; the categories are administrative conveniences laid over it, and reacting to a category change rather than a trend is responding to the wrong signal.

Questions