Thirty is the number. Below it you are not obese, at it or above it you are, and a great deal follows from which side of it you land on: what a doctor says to you, what a study counts you in, what an insurer prices, what a drug trial will enrol you for.
It is a strange number to carry that much weight, because it is derived from a formula that was never built to describe a person. It was built to describe a crowd.
What the thing actually is
In the 1830s a Belgian mathematician, Adolphe Quetelet, was working on a project of social statistics. He wanted to characterize l’homme moyen, the average man, and to do that he needed a way of comparing people of different heights on a single scale. Dividing weight by the square of height gave him one. It worked reasonably well for the population he had data on, which was white European men.
Quetelet was explicit about the level at which his methods applied. His interest was in what emerges from large numbers, not in individual peculiarities. The index was a way of describing a distribution, and a distribution is not a patient.
The formula sat in that role for well over a century. In 1972 Ancel Keys and colleagues published a comparison of the available weight-for-height indices and gave Quetelet’s the name it carries now: body mass index. The name stuck, and with it the number crossed from describing groups to being applied to the person in the room.
It is worth being fair about why. BMI is free. It requires a scale, a tape measure and arithmetic a clinician can do in their head. It can be extracted retrospectively from records collected for other reasons, which is why so much of what we know about weight and health rests on it. A measure that costs nothing and exists in every dataset has an enormous practical advantage over a better one that does not.
What it misses
The objection is not that BMI is meaningless. Across populations it tracks health outcomes well enough to have supported decades of useful research. The objection is that it does not distinguish between things that matter clinically.
It does not separate muscle from fat, so it misclassifies the very muscular. It says nothing about where fat sits, and visceral fat around the organs behaves differently from fat elsewhere. It does not vary by ancestry, though the relationship between the ratio and metabolic risk does. Most importantly for what follows, it says nothing at all about whether the person in front of you is unwell.
What is replacing it
In 2025 a Lancet commission of 58 experts proposed a different structure, and the change is conceptual rather than a tweak to the cutoff.
The first part is measurement. Rather than a single ratio, excess adiposity is established using BMI alongside direct measures of body shape and composition: waist circumference, waist to hip and waist to height ratios, body fat percentage. BMI is not discarded. It is demoted to one input among several.
The second part is the one that matters. Having established excess adiposity, the framework then splits it in two. Clinical obesity is defined as ongoing organ dysfunction or limitation of daily activities attributable to the adiposity itself. Pre-clinical obesity is excess adiposity with variable risk but no current illness. The first is a disease. The second is a risk state.
That distinction is doing something a threshold cannot. It separates being at risk of becoming unwell from being unwell, which is a line medicine draws carefully in almost every other chronic condition and has not drawn here.
What happened when someone applied it
A 2026 paper in the journal Obesity applied the commission’s excess adiposity criteria to a United States adult population. Under those criteria close to three in five adults qualify.
That number invites a wrong reading, so it is worth slowing down on. It does not mean three in five American adults are newly sick. It does not mean anybody’s health changed. It is a prevalence estimate produced by a broader definition of excess adiposity, and the broader definition catches people that BMI alone misses, particularly those whose ratio is unremarkable but whose waist measurement is not.
The clinically ill subset is smaller, because clinical obesity requires demonstrated organ dysfunction, and that determination has to be made person by person rather than calculated. One paper is one paper: this is a single prevalence analysis in one population, and prevalence estimates move with the definition applied to them. What it establishes is that the definition you choose changes the size of the group by a great deal, which is a finding about measurement rather than about bodies.
A separate application of the criteria to a clinical trial population reported something more awkward. Applying the framework was complicated even in a well-resourced trial setting, with staff, time and protocol on its side. That is a warning about implementation rather than about validity, and it is the more practically important of the two results.
The part that does not resolve
A better definition that is harder to apply is not automatically an improvement. Whether it is depends entirely on where it has to be applied.
In a research setting, or a specialist clinic with time and equipment, the commission’s framework is plainly better. It measures more of the right things and it distinguishes illness from risk.
In a fifteen-minute primary care appointment, with a queue outside and a tape measure that may or may not be in the drawer, a framework requiring multiple measurements and a judgment about organ dysfunction competes against a number the computer has already calculated. Clinical guidance that is too slow to follow does not produce better care. It produces inconsistent care, applied properly by clinicians who have time and skipped by those who do not, which is a different failure from the one it was designed to fix.
None of that is an argument for keeping a nineteenth-century population statistic as a diagnostic threshold. The case against BMI in that role is strong and the commission has made it carefully.
It is an argument that the interesting question has moved. It is no longer whether BMI is a good measure, because it is not, and everyone who works with it already knows. It is whether a health system can absorb a measure that is better and more expensive, and what happens to the patients seen in the years while it tries. That question will not be answered by another prevalence paper. It will be answered slowly, unevenly, in appointments, by people deciding whether they have time to get the tape measure out.
