BMI Definition: What It Measures and What It Misses

Body mass index, or BMI, is a single number derived from your weight and height, calculated by dividing your weight in kilograms by the square of your height in meters. A person who weighs 70 kg and stands 1.75 m tall, for example, has a BMI of about 22.9. The number was never designed to diagnose obesity in individuals, and its limitations have become increasingly well documented, but it remains the most widely used screening tool for weight status in clinical practice and public health research worldwide.

Where BMI Comes From

The formula traces back to the nineteenth-century Belgian mathematician Adolphe Quetelet, who observed that in adults, weight tends to increase roughly in proportion to the square of height. His “Quetelet Index” sat largely unused in medicine for over a century until the American physiologist Ancel Keys formally coined the term “Body Mass Index” in 1972, proposing it as a practical population-level tool for epidemiological studies of obesity.¹ Keys was explicit that BMI was better suited for comparing groups than for diagnosing any one person. That caveat has largely been lost as the number migrated from research papers into doctor’s offices and insurance forms.

The Standard Categories

The World Health Organization and most Western health agencies classify adult BMI into four broad bands: underweight (below 18.5), normal weight (18.5 to 24.9), overweight (25.0 to 29.9), and obese (30.0 and above), with obesity further divided into classes. These thresholds were chosen because, at the population level, the risk of conditions like type 2 diabetes, heart disease, and certain cancers tends to rise as BMI climbs past the mid-twenties. The boundaries are round numbers, not biological cliff edges. Someone with a BMI of 24.8 is not meaningfully healthier than someone at 25.2, even though the two fall on different sides of a cutoff.

What BMI Actually Measures and What It Misses

The formula captures total mass relative to height. It cannot tell how much of that mass is fat, how much is muscle, or where the fat sits on your body. When researchers compared BMI classifications with body-fat measurements from DXA scans (a type of detailed X-ray used in research settings), BMI misclassified roughly a third of both women and men.² Some people the scale calls “normal weight” are carrying excess fat, and some people the scale calls “overweight” are carrying a lot of lean mass.

The location of fat matters at least as much as the total amount. Visceral fat, the kind packed around internal organs in the abdomen, is far more metabolically dangerous than fat stored under the skin on your hips or thighs. In one large study, people in the highest quartile of visceral fat had nearly triple the risk of dying from cardiometabolic causes compared with those in the lowest quartile, an association substantially stronger than what BMI alone captured.³ Even among people whose BMI falls squarely in the “normal” range, high visceral fat is associated with substantially elevated metabolic risk.⁴ People with normal-weight central obesity, sometimes called “thin on the outside, fat on the inside,” face a mortality risk similar to or possibly higher than people who are both centrally obese and overweight by BMI.⁵

The Height Problem

Because the formula divides weight by height squared, it implicitly assumes that all body dimensions scale the same way as you get taller. They don’t, quite. In practice, BMI systematically runs a bit lower for very tall people and a bit higher for very short people, relative to their actual body fat. The bias is more pronounced in women than in men. Across English population surveys spanning two decades, the tallest quarter of women had a mean BMI almost two full points lower than the shortest quarter, even after accounting for other factors.⁶

The height bias is especially problematic in children. In eight-year-olds, body mass does not scale neatly with height squared, and BMI overestimates adiposity in taller children and underestimates it in shorter ones. Researchers have proposed alternative formulas using height cubed to correct for this, finding that around 6 to 7 percent of children get classified differently when the height bias is removed.⁷

BMI and Mortality: The J-Shaped Curve

One of the most consistent findings across large studies is that the relationship between BMI and the risk of dying follows a J-shaped curve rather than a straight line. Risk is elevated at the low end of the BMI spectrum, drops to its lowest in a range around the lower to middle twenties, then rises again as BMI climbs. A major meta-analysis pooling data from 230 cohort studies and over 30 million participants found that the lowest mortality among never-smokers occurred at a BMI of about 23 to 24.⁸ A UK study of 3.6 million adults confirmed the J-shape, estimating that below a BMI of 25 each five-point drop in BMI was associated with increased mortality, while above 25 each five-point rise was also associated with increased mortality.⁹

The curve’s shape explains a persistent puzzle in the medical literature called the “obesity paradox.” In several chronic diseases, including advanced kidney disease, heart failure, and chronic lung disease, patients who are mildly overweight by BMI sometimes survive longer than patients at so-called normal weight.¹⁰ Whether the extra weight is truly protective or whether BMI is simply a poor gauge of health status in people who are already sick remains debated, but the pattern is robust enough that clinicians treating these conditions have learned not to assume that a lower BMI is always better.

Why the Same BMI Means Different Things in Different Populations

The standard cutoffs were developed primarily from data on White European populations. For people of other backgrounds, the same BMI number often corresponds to a different amount of body fat and a different level of metabolic risk. A WHO expert consultation concluded that Asian populations develop type 2 diabetes and cardiovascular disease at BMIs well below the conventional overweight threshold of 25.¹¹

A large population-based study in England quantified how far apart these thresholds really are. To match the type 2 diabetes risk that White populations face at a BMI of 30, the equivalent BMI was about 24 in South Asian populations, about 27 in Chinese and Arab populations, and about 28 in Black populations.¹² That six-point gap between White and South Asian populations is enormous in practical terms. A South Asian person at a BMI of 24, well within the “normal” range by conventional standards, already faces the diabetes risk that a White person would not reach until a BMI of 30.

Part of the explanation lies in how fat distributes itself differently across populations. At the same BMI, White women tend to carry more total fat mass than Black or Hispanic women, but Hispanic women tend to concentrate a higher proportion of their fat in the trunk, which is the more metabolically risky location.¹³ Among men and women of the same BMI, non-Hispanic Black adults tend to have the lowest percentage of body fat, followed by non-Hispanic White adults, with Mexican American adults the highest, though the differences are relatively modest at around three percentage points.¹⁴ These patterns mean that a single BMI number carries genuinely different health implications depending on your ancestry and sex.

BMI in Older Adults

Aging changes the equation in ways that make BMI even less reliable. Muscle mass tends to decline with age while fat mass, especially visceral fat, tends to increase. Two people of the same height and weight at ages 30 and 70 may have very different body compositions, yet their BMI would be identical. On top of that, the “optimal” BMI for survival appears to shift upward as people age. In a large Polish study of adults aged 80 and older, the lowest mortality was observed in those with BMIs between 25 and about 40, while those with BMIs below 25, the range traditionally labeled “normal,” had the highest mortality.¹⁵

A review that looked specifically at functional outcomes in older people, things like balance, gait speed, fall risk, and muscle strength, found that both low BMI (below 25) and very high BMI (above 35) were associated with worse function. The optimal BMI range for maintaining physical capability appeared to be around 27 to 28 for older men and 31 to 32 for older women.¹⁶ A large Chinese prospective study similarly found that BMIs in the 24 to 30 range were protective against mortality at all ages, but the benefit of carrying extra weight was especially clear in older groups.¹⁷ The upshot is that advising an 80-year-old to aim for a BMI under 25 may do more harm than good, a point that many clinical guidelines are still catching up to.

BMI in Children and Adolescents

For children and teenagers, a raw BMI number means little by itself because body composition changes dramatically with normal growth. Instead, pediatric BMI is expressed as a percentile relative to children of the same age and sex, using growth charts from agencies like the Centers for Disease Control and Prevention. A child at the 85th percentile is considered overweight, and one at or above the 95th percentile is considered obese.

This percentile system has a known blind spot at the extremes. Standard BMI-for-age z-scores become unreliable above the 97th percentile, compressing children with very different body sizes into a narrow statistical range. Researchers have developed alternative metrics, such as expressing BMI as a percentage of the 95th percentile, to better track changes over time in children with severe obesity.¹⁸ For clinicians tracking a child’s progress, this matters: a child could lose meaningful amounts of fat without any change showing up in the standard z-score.

Alternatives That Add What BMI Lacks

Because BMI ignores fat distribution, several complementary measures have gained ground. The simplest is waist-to-height ratio: divide your waist circumference by your height, and a value above 0.5 suggests elevated central fat. Meta-analyses in adults have found that waist-to-height ratio is at least equal to and sometimes better than BMI at predicting cardiovascular risk factors, and it outperforms BMI specifically in people with diabetes.¹⁹ Its practical appeal is obvious: all you need is a tape measure, and the 0.5 threshold applies across ages and sexes, unlike BMI’s population-specific cutoffs.²⁰

A more sophisticated option is A Body Shape Index, or ABSI, which uses waist circumference adjusted for both BMI and height. ABSI is designed to capture abdominal fat independently of overall size. In a large European cohort study, it offered a useful advantage: because ABSI is not strongly correlated with BMI, it could identify high-risk subgroups within every BMI category, including people in the “normal” weight range who carried disproportionate abdominal fat. The highest quartile of ABSI was consistently associated with higher mortality across every BMI band.²¹ Tools like these are not meant to replace BMI entirely but to layer on the information it misses.

Shifting Clinical Attitudes

The medical establishment has begun to formally acknowledge what researchers have been saying for decades. In June 2023, the American Medical Association adopted a policy stating that BMI is an imperfect measure that loses predictive value across different age, sex, and racial groups, and should not be used as the sole criterion for clinical decisions or insurance policies.²² Eligibility for bariatric surgery or newer weight-management medications, the AMA said, should involve more comprehensive assessments. The statement was notable less for its content, which researchers already agreed with, than for the signal it sent to the broader medical community.

Weight stigma in clinical settings adds a layer of harm that goes beyond misclassification. Research has found that women with higher body weight are more likely to switch doctors frequently and have less communication with their providers about sexual and reproductive health, patterns consistent with the avoidance behaviors that weight stigma produces.²³ When a blunt tool like BMI is treated as a definitive diagnosis rather than one data point among many, it can contribute to a culture in which patients feel reduced to a number.

Genetics and the Heritability of BMI

BMI is partly heritable. Twin and family studies have long shown that genetics plays a substantial role in determining body weight, and genome-wide association studies have now identified over 400 genetic loci linked to BMI variation.²⁴ None of these individual variants has a large effect on its own; the genetic contribution to BMI is the product of many small nudges spread across the genome, interacting with environment, diet, and activity. This is worth understanding because it means that two people living nearly identical lifestyles can end up with meaningfully different BMIs, and the difference is not necessarily a reflection of willpower or discipline.

Global Trends in Population BMI

Average BMI has been climbing worldwide for decades. Between 1980 and 2008, mean BMI rose by about 0.4 points per decade in men and 0.5 points per decade in women globally, though the speed of change varied dramatically by country. The fastest rises, exceeding two BMI points per decade, occurred in parts of Oceania.²⁵ The trend extends to children. A pooled analysis covering 1975 to 2016 found that age-standardized mean BMI in children and adolescents increased by roughly 0.3 to 0.4 points per decade globally, reaching a mean of about 18.5 to 18.6 by 2016.²⁶ These shifts are driving much of the global conversation about metabolic disease, health-care costs, and how societies organize food systems and urban environments. Whether BMI is the best yardstick for tracking those shifts is a separate question, but for now, it is the one most of the world’s data is built around.