Where BMI Came From, and Why 25 Became the Line

Two measurements go in and one number comes out, and that arithmetic hasn't changed in nearly two centuries. What has changed, more than once, is where the boundaries between the categories sit and who is expected to do something about them. It helps to know that history before you read your own figure as a verdict.

This is general information put together by a software team. Nobody with medical training has reviewed this page. None of it is medical advice, and none of it can tell you whether your own result means anything.

An index built to describe crowds

Adolphe Quetelet was a Belgian astronomer who spent much of his working life applying observational statistics to people. The ratio of weight to the square of height appears in his 1835 treatise on what he called l'homme moyen, the average man, as a way of describing how build is distributed across a population.

He wasn't proposing a screening test. There was no clinical data behind the ratio, no cohort followed for outcomes, and no suggestion that one person's value meant anything on its own. It was a descriptive statistic for groups, which is a different kind of thing altogether.

The modern name arrived 137 years later. Ancel Keys and four co-authors, writing in the Journal of Chronic Diseases in 1972, compared several simple height-and-weight ratios against measured body fat in cohorts drawn from five countries. Weight over height squared tracked measured fatness better than the alternatives and depended least on how tall the subject was, so they recommended it and gave it the name body mass index. The same paper noted that the index suited population studies and wasn't meant for diagnosing individuals. That qualification did not travel with the formula.

Why the height is squared

The exponent is an empirical fit, nothing more. If people scaled like geometrically similar solids, mass would rise with the cube of height and the sensible denominator would be height cubed. Real people don't scale that way (taller people tend to be proportionally slimmer), and Quetelet's measurements put adult weight nearer the square.

The square fits the middle of the distribution well and drifts at the ends, and the drift is systematic, not random. Nick Trefethen, a numerical analyst at Oxford, argued in a 2013 letter to The Economist that weight scales closer to height to the power 2.5, and proposed 1.3 times weight divided by height to the 2.5 as a replacement.

Take two people who share a BMI of exactly 25.0 and run them through both. Someone 1.60 m tall weighing 64 kg comes out at 25.7 on the alternative index; someone 1.90 m weighing 90.25 kg comes out at 23.6. That's two points between people the standard index calls identical, purely because of height. The proposal hasn't been adopted, and that isn't because it's wrong. Most of an index's value comes from everyone using the same one.

The week that millions of people changed category

Through the 1980s and early 1990s, United States federal guidance put the overweight boundary at roughly 27.8 for men and 27.3 for women. Those values came from the 85th percentile of a national survey, with no biological threshold behind them. The National Heart, Lung, and Blood Institute's clinical guidelines, released in 1998, replaced both figures with a single cut-off of 25 for everyone, bringing United States practice into line with the classification the World Health Organization had adopted.

Nobody's body changed. Tens of millions of American adults were reclassified overnight by an edit to a document. Whatever risk each of them carried on the Monday, they carried exactly the same risk on the Friday.

That's all a cut-off is: an administrative line ruled through a smooth, continuous distribution, because guidance has to say something and continuous variables are awkward to write policy about. Someone at 24.9 and someone at 25.1 are, physiologically, the same person twice. Several health authorities also use lower action points for South and East Asian populations, which amounts to admitting that a single global boundary was always a compromise.

The mortality evidence is messier than the table suggests

The bands are usually shown as a ladder of increasing danger. The largest attempt to check that against death records made the picture a lot more complicated.

Katherine Flegal and colleagues pooled 97 studies covering roughly 2.88 million people and published the result in JAMA in 2013. Grade 1 obesity, 30 to 35, showed no statistically significant excess of all-cause mortality. The overweight band, 25 to 30, was associated with slightly lower mortality than the healthy-weight band. Only grades 2 and 3, above 35, showed the clear excess the ladder predicts.

The paper was attacked hard, and the criticisms are serious. Smokers tend to be lighter and die earlier, which drags the reference group down. People who are already ill often lose weight before they die, so illness can look like a result of thinness when it's really the cause of it. The debate is genuinely unsettled, and anyone presenting either side as established fact is overselling it. What it does show is that the mortality data doesn't simply confirm the neat progression printed in the table.

In 2023 the American Medical Association adopted a policy that explicitly acknowledges the index's limitations and recommends using it alongside other measures, not on its own.

What to do with your own number

Our BMI calculator prints the same standard bands, and it inherits every problem described above. It reports a population statistic. It doesn't know your muscle mass, waist circumference, blood pressure, blood glucose, family history or age, and those are the things that carry the clinical information.

Used as one input among several, it's a reasonable one. It's cheap, it's reproducible, and it's comparable over time if you weigh yourself the same way. Used as a diagnosis, it's being asked to do a job it was never built for, and one its own author declined to claim for it.

If you want a measurement that responds to shape instead of mass, a tape measure and the body fat calculator will get you closer. If you want to know what your result means for you, talk to a clinician who can see the rest of the picture.