BMI: what body mass index measures, what it predicts about health and what it misses

Body mass index is one of the most widely used numbers in medicine. It decides which box a patient falls into on a chart, it appears in the entry criteria for trials and treatments, and anyone with a bathroom scale can work it out. It is also one of the most criticized, on the grounds that a single ratio of weight to height cannot tell a bodybuilder from someone carrying a dangerous amount of fat.

Both views have evidence behind them. Across very large populations, BMI tracks closely with the risk of dying early, as the largest analysis of the question showed in 2016. For any one person, it is a rough guide that misses a good deal, and the organizations that set the categories have been saying so for twenty years.

What BMI is and where the formula came from

BMI is weight in kilograms divided by the square of height in meters. A person 1.75 meters tall who weighs 70 kilograms has a BMI of just under 23. Dividing by height squared is meant to cancel out the fact that taller people weigh more, leaving a number that reflects build.

The arithmetic is old. The physiologist Ancel Keys, who gave the index its modern name in a 1972 paper, traced it to Adolphe Quetelet, described there as the great pioneer in anthropometry and statistics, who had explored the ratio almost a century and a half earlier. Keys and colleagues compared several rival formulas against measurements of height, weight and body fat in groups of men from Europe, Japan, South Africa and the United States, and concluded that weight over height squared did the job best. “It is proposed that this ratio, W/H2, be termed the body mass index,” they wrote in 1972.

Keys was after a tool for comparing populations, as a replacement for the height and weight tables that life insurers had published since 1912. The use of BMI to sort individual patients came later.

What the BMI categories mean

The bands in use today date from the 1990s, when, as one account puts it, a WHO committee of experts proposed BMI cutoffs for overweight and obesity.

BMI	World Health Organization category
Below 18.5	Underweight
18.5 to under 25	Normal weight
25 to under 30	Overweight
30 to under 35	Obesity, grade 1
35 to under 40	Obesity, grade 2
40 or above	Obesity, grade 3

The lines were drawn from mortality statistics. The threshold of 30 for obesity was developed from observational studies in Europe and the USA of exclusively White populations and based on the association between BMI and mortality. A category is therefore a statement about averages in the populations that were studied. It describes a range of weight for height and does not diagnose anything.

What BMI predicts: mortality in 10.6 million adults

The fullest test of those averages came from the Global BMI Mortality Collaboration, coordinated by researchers at the University of Cambridge and Harvard and published in The Lancet in 2016. It combined individual records from 239 long-term studies and 10.6 million adults on four continents.

The collaboration’s central problem was that illness and smoking both cause weight loss and both shorten life. As the authors put it, chronic disease and smoking can themselves affect BMI. If thin smokers and people losing weight to undiagnosed cancer are left in, low BMI looks deadly and higher BMI looks safe by comparison. The team dealt with this by restricting analyses to never-smokers and excluding pre-existing disease and the first 5 years of follow-up. That left about 4 million people and 385,879 deaths.

Among them, death rates were lowest between a BMI of 20 and 25. They rose steadily above that range: about 7% higher at 25 to 27.5, 20% higher at 27.5 to 30, 45% higher in grade 1 obesity, nearly double in grade 2 and close to three times as high in grade 3. Above 25, each additional five units of BMI went with roughly a 30% to 40% higher death rate, depending on the region.

Rates rose on the other side too. Adults with a BMI between 18.5 and 20, still inside the normal category, had a 13% higher death rate, and those below 18.5 had a rate about half as high again. The authors remarked that the WHO’s normal range “is excessively wide” and noted that in excessively lean adult populations underweight remains a cause for concern.

That pattern varied between groups. The relative increase with BMI was larger in men than in women and larger in younger adults than in older ones, although the absolute excess mortality was higher in older people, because more of them die in any year. Overall, the associations of both overweight and obesity with higher death rates from all causes were broadly consistent in four continents.

Why some BMI studies found that overweight people live longer

Three years earlier a widely reported analysis had reached a different verdict. Katherine Flegal and colleagues at the US Centers for Disease Control and Prevention pooled the published literature, where 97 studies were retained for analysis, providing a combined sample size of more than 2.88 million individuals and more than 270,000 deaths. Those in the overweight category had a death rate about 6% lower than those in the normal category, and grade 1 obesity carried no clear excess. The finding became known as the obesity paradox.

The Lancet collaboration argued that the paradox was an artifact of the same biases it had worked to remove. Its authors wrote that their study “was able to reproduce such findings when conducting crude analyses with inadequate control of reverse causality, but not when we conducted appropriately strict analyses”. Reverse causality is the situation in which the outcome shapes the exposure, here illness causing weight loss.

The dispute has not entirely gone away. Removing smokers and people with existing disease leaves a healthier-than-average group, so the strict analysis may not describe everyone. What the two analyses agree on is that death rates are clearly higher from a BMI of 35.

What BMI misses: muscle, where fat sits, ethnicity and age

The limits of BMI are easiest to see when it is checked against instruments that measure fat directly. A team led from the University of Missouri pooled such comparisons in 2010. Twenty-five articles qualified, and these studies evaluated 32 different samples totaling almost 32,000 people. A BMI of 30 or more was a trustworthy sign of high body fat, in that almost nobody above the line had a normal amount. The reverse did not hold. The standard cutoffs, the authors concluded, fail to identify half of the people with excess body fat. Some people with a BMI in the twenties carry little muscle and a great deal of fat.

BMI is also silent on where fat is stored, which matters at least as much as how much there is. Fat packed around the abdominal organs is more strongly tied to diabetes and heart disease than fat under the skin of the hips and thighs. An international group of specialists argued in 2020 that waist circumference provides both independent and additive information to BMI for predicting morbidity and risk of death, morbidity meaning illness. The group stated flatly that “BMI alone is not sufficient to properly assess or manage the cardiometabolic risk associated with increased adiposity in adults”. Cardiometabolic risk is the risk of heart disease, stroke and diabetes taken together.

Ancestry changes the meaning of the number. A WHO expert consultation concluded in 2004 that Asian populations have different associations between BMI, percentage of body fat, and health risks than do European populations, with diabetes and heart disease common at BMIs below 25. A study of 1.47 million adults in England, led by Rishi Caleyachetty, later put figures on this. The diabetes risk that white adults reached at a BMI of 30 was reached by south Asian adults at about 24, by Arab and Chinese adults at about 27, and by Black adults at about 28. For south Asian adults that is a cutoff much lower than the recommended ethnicity-specific cutoff of 27.5 then in use.

Age shifts it again. Older adults lose muscle and height, both of which distort a ratio built on weight and height, and the Lancet analysis found that a five-unit rise in BMI carried a smaller relative penalty at 70 to 89 than at 35 to 49.

What guidelines say about using BMI

The WHO consultation decided against separate scales for different populations. It agreed that the WHO BMI cut-off points should be retained as international classifications, and added further “action points” at 23, 27.5, 32.5 and 37.5 that countries could use to define raised risk in their own populations. The WHO and, in England, the National Institute for Health and Care Excellence both recommend 27.5 as the point that triggers action for south Asian and Chinese adults.

The 2020 consensus statement, from the International Atherosclerosis Society and the International Chair on Cardiometabolic Risk, recommends that waist circumference be measured routinely alongside BMI and treated as a vital sign. Its authors acknowledge that the waist thresholds for each BMI category still need refining by age, sex and ethnicity.

How a BMI figure bears on any one person’s health depends on the rest of their risk picture, which is assessed by their doctor.

Open questions about BMI and health

The Lancet collaboration listed its main weakness plainly: “our only measure of adiposity was BMI, so we could not directly address aspects of body composition such as visceral fat or fat distribution”. Visceral fat is the fat around the abdominal organs. The largest study of BMI, in other words, cannot say whether BMI or something it stands in for is what matters.

Geography narrows the evidence as well. In that analysis south Asia, Africa, and Latin America were either unrepresented or poorly represented, which are the regions where the relationship between weight, body fat and disease is most likely to differ. And nearly all of it is observational, built on a single weighing at the start of each study, with no information on whether a person’s weight had been stable, rising or falling.

BMI remains a cheap and consistent way to compare populations, which is the job it was named for, and a blunt instrument for judging one body.