BMI by Age and Sex: Why One Table Doesn't Fit Everyone

The World Health Organization's BMI classification is a single table. A BMI of 18.5 to 25 is normal, 25 to 30 is overweight, 30 and above is obese — and those numbers are applied identically to a 22-year-old male powerlifter, a 45-year-old woman of average build, and a 78-year-old man who has been losing muscle for two decades.

That universality is the reason BMI is useful. It needs no equipment, no reference charts, and no judgement, which is why it can be applied to entire populations cheaply. It is also the reason BMI attracts more criticism than any other measure in health, because those three people are not comparable and the table pretends they are.

This piece is about where the single table genuinely breaks down, and what to do about it in each case. It is not an argument that BMI is worthless — the relationship between BMI and mortality across large populations is real and well replicated. It is an argument for knowing which of the several ways your own reading might be misleading.

Body composition changes throughout adult life even when weight does not. From roughly the fourth decade onward, most people lose skeletal muscle and gain fat mass at a fairly steady rate. Someone who weighs the same at 70 as they did at 30 is carrying substantially more fat and substantially less muscle, and their BMI has not moved a point.

This means a 'normal' BMI in an older adult can coexist with meaningful sarcopenia — the age-related loss of muscle mass and strength that predicts falls, fractures, hospital stays and loss of independence. The number that looks reassuring is precisely the one hiding the problem.

The evidence on where the optimal BMI sits in older age is also less tidy than most people assume. Several large analyses have found the BMI associated with lowest mortality drifts upward with age, and that in adults over about 65 a modest amount of extra weight appears protective rather than harmful — probably because it provides reserve during illness. The finding is contested and confounded by reverse causation, since illness causes weight loss as well as the other way round. But it is enough to say confidently that applying the 18.5–25 band to a 78-year-old with the same firmness as to a 28-year-old is not supported.

Men and women get the same BMI thresholds despite having reliably different body composition. At any given BMI, women carry more body fat than men — typically around ten percentage points more — because essential fat requirements differ for reproductive reasons.

This is not a flaw so much as a deliberate simplification. The WHO thresholds were set against health outcomes rather than against body fat, and the outcome-based thresholds happen to land in a similar place for both sexes even though the underlying composition differs. It works, in the sense that the categories predict risk in both groups.

Where it stops working is when people use BMI as a proxy for body fat and compare across sexes. A man and a woman at BMI 24 are not in the same physical condition, and comparing them as though they were produces confusion — particularly for women, who can be at a perfectly healthy body fat percentage while a partner at the same BMI is not.

This is the criticism that gets raised first and it is entirely valid, just narrower than people think. BMI is a measure of weight relative to height, and muscle is denser than fat, so a trained lifter can sit in the overweight or obese band with single-digit body fat.

The narrowness is that this affects far fewer people than the frequency of the objection suggests. Carrying enough muscle to be misclassified requires years of serious resistance training. For most of the population, a high BMI reflects a high fat mass, and the study evidence bears this out: BMI-defined obesity has good specificity — it rarely flags someone who is not carrying excess fat — while its sensitivity is poor, meaning it misses a great many people who are.

So the muscle objection cuts in an unexpected direction. BMI's real weakness is not that it over-diagnoses athletes; it is that it under-diagnoses everyone whose excess fat sits on a frame that reads as normal weight. That group is much larger, and much less likely to be told anything is worth looking at.

Cardiometabolic risk does not begin at the same BMI in every population. There is consistent evidence that people of South and East Asian ancestry develop type 2 diabetes and cardiovascular disease at lower BMI values than European-ancestry populations, largely because of differences in where fat is stored and how much of it is visceral at a given weight.

A WHO expert consultation reviewed this in 2004 and considered redrawing the categories for Asian populations. It declined, on the grounds that the risk gradient is continuous and varies substantially between Asian populations too, but it recommended additional public-health action points at 23 and 27.5 kg/m² alongside the international cut-offs.

Several countries have since adopted lower national thresholds. If your ancestry falls into these groups, read a BMI in the 23 to 25 range as a prompt to check waist circumference and metabolic markers rather than as reassurance.

In June 2023 the American Medical Association adopted a policy formally recognising BMI as an imperfect measure that should not be used in isolation. The policy states that BMI is problematic when used alone as a clinical measure, notes its historical basis in data drawn overwhelmingly from white populations, and recommends it be used alongside other valid measures including visceral fat, body composition, waist circumference and metabolic factors.

It is worth being clear about what this did and did not mean. The AMA did not say BMI is useless or should be abandoned; it said BMI on its own is not sufficient for clinical decisions about an individual. That is roughly what a careful clinician was already doing, and it is exactly what this site has always recommended: read the number alongside something that describes where the weight sits and what it is made of.

The practical response to all of the above is not to distrust BMI, but to stop asking it to answer questions it cannot.