Body mass index is criticised constantly and used constantly. Both are reasonable, because the measure performs adequately at the task it was designed for and poorly at the task it is usually applied to.
The formula was built for populations
The ratio of weight to the square of height was developed in the nineteenth century as a way of describing the distribution of body size across a population.
It was never intended as a diagnostic tool for individuals, and its originator noted explicitly that it did not apply well to any single person.
Its adoption in clinical settings came much later and was driven largely by the practical need for a measure requiring nothing but scales and a tape.
It cannot distinguish between tissues
The formula uses total mass, which includes muscle, bone, fat and fluid without separating them.
A muscular individual and a sedentary one of the same height and weight receive the same value despite having very different body composition.
The error runs in the other direction too, since someone with low muscle mass and higher fat mass can sit within the normal range while carrying the associated risks.
The squared term distorts at the extremes
Dividing by height squared was an empirical choice that fits average adult proportions reasonably well and fits unusual ones poorly.
Very tall people are systematically assigned higher values and very short people lower ones than their actual composition warrants.
The thresholds also perform differently across populations, which is why several countries apply lower cut-offs than the standard set.
Distribution matters more than total
Fat stored around the abdominal organs is more strongly associated with metabolic and cardiovascular risk than fat stored under the skin.
Two people with identical values can differ substantially in that distribution, which is why waist measurement adds information that the index cannot.
Waist-to-height ratio is increasingly used alongside it for this reason, being nearly as simple to obtain.
Why it has not been replaced
Methods that measure composition directly require equipment, trained operators and cost, which makes them impractical for routine screening at scale.
Because the index has been recorded consistently for decades, it also underpins a large body of research that cannot easily be re-derived from another measure.
Its appropriate use is as one input among several in a clinical assessment, which is how it is treated in current guidance rather than as a verdict on an individual.