BMI, explained properly

Where BMI came from

The formula is Adolphe Quetelet's, from the 1830s. He was looking for the 'average man' in Belgian census data, not diagnosing anyone. It languished as an actuarial curiosity until 1972, when physiologist Ancel Keys tested competing height-weight indices against actual body fat measurements and found Quetelet's ratio the least bad of the simple options. Keys named it 'body mass index' and, importantly, recommended it for population studies. He was explicit that it was a poor tool for individuals.

The WHO adopted BMI for international classification in the 1990s, drawing the now-familiar lines at 18.5, 25, 30, 35 and 40. Those thresholds came from mortality and disease curves in mostly European cohorts: risk rises noticeably above 25 and again above 30. They were always population landmarks, chosen partly for memorability: nothing biologically special happens at 24.9 versus 25.1.

Two things follow from that history and are worth holding onto. First, BMI was never designed to measure body fat; it was designed to remove height from a weight comparison so that populations of different stature could be compared at all. Second, the cut-offs describe where risk curves bend in the data they were drawn from, which is why other cut-offs exist for populations whose curves bend somewhere else.

The formula, in metric and in imperial

In metric the calculation is BMI = weight in kilograms ÷ (height in metres)². In imperial it is BMI = 703 × weight in pounds ÷ (height in inches)². The 703 is not a fudge factor: it is the unit conversion built in, because 39.3701 inches per metre squared, divided by 2.20462 pounds per kilogram, comes to 703.07. Rounded to 703 it shifts the result by about one part in ten thousand, which is invisible at one decimal place.

Work an example through both. Someone 178 cm tall weighing 82 kg: square the height in metres, 1.78 × 1.78 = 3.1684, then divide, 82 ÷ 3.1684 = 25.88, so BMI 25.9. The same person in imperial is 70.08 inches and 180.8 pounds: square the height, 70.08² = 4911, multiply the weight by 703, 180.8 × 703 = 127,102, then divide, 127,102 ÷ 4911 = 25.88, BMI 25.9 again. If a calculator gives you two different answers in two unit systems, it is rounding your height or weight before it squares them, not using a different formula.

One consequence of squaring height is worth knowing. Real bodies do not scale as perfect squares, so dividing by height² slightly overstates BMI in tall people and understates it in short people. The effect is small (a point or so at the extremes of adult height), but it is systematic rather than random, and it is one more reason to treat the first decimal place as noise.

What the WHO categories actually say

The full WHO classification has more rungs than the four most people know. It subdivides the underweight range into severe, moderate and mild thinness, calls 25.0–29.9 'pre-obese' rather than simply 'overweight', and splits obesity into three classes. The wording matters: 'pre-obese' was chosen to describe a statistical position on a risk curve, not a prediction about any individual.

Read the table as a set of boundaries in a population distribution, not as a set of diagnoses. Someone at 24.8 and someone at 25.2 differ by a fraction of a kilogram, and nothing in the underlying evidence distinguishes them. What the table does well is tell you roughly where you sit on a curve that most people are somewhere on.

What the number actually captures

Across a population, BMI correlates with body fat percentage at roughly r = 0.7–0.9 depending on age and sex. That is strong, which is why it works for public health surveillance. For an individual, that leftover scatter is the entire problem: two people with a BMI of 27 can carry anywhere from athletic-lean to clinically concerning body fat, because the formula sees only total mass. Muscle, bone density, fluid, fat and its location are all invisible to it.

Height itself introduces a subtler bias. Dividing by height squared slightly overstates BMI in tall people and understates it in short people, because real bodies don't scale as perfect squares. It is a small effect, but one more reason not to treat decimals as destiny.

The sharper limitation is that BMI cannot see where fat sits, and location is most of the risk. Fat stored viscerally (around the liver, pancreas and intestines) behaves metabolically very differently from the same mass stored subcutaneously on hips and thighs. Two people at BMI 29 with identical body fat percentages can carry quite different cardiometabolic risk purely on distribution, which is exactly the gap a waist measurement fills for the price of thirty seconds and a tape.

When BMI is the wrong tool

Muscular and athletic people read high while lean, covered in depth in our athletes guide. Older adults can read 'normal' while carrying little muscle and much fat (sarcopenic pattern), which the number cannot see. Many Asian populations face elevated cardiometabolic risk at BMI levels the standard table calls normal, which is why separate Asia-Pacific cut-offs exist. Pregnancy, amputation and significant oedema each break the formula's assumptions outright. And for anyone under 18, adult categories are simply invalid. Children are assessed on BMI-for-age percentile charts.

Since January 2025, mainstream clinical guidance has caught up with these limits: a Lancet Commission and its wide endorsements now require BMI plus at least one fat-distribution measure (waist circumference, waist-to-height or waist-to-hip ratio) before concluding excess adiposity. Our BMI calculator runs that combined check for you.

• Trained and muscular people: BMI counts muscle as if it were fat, and reads high.
• Older adults with low muscle mass: BMI can read normal while body fat is high.
• People of East, South and Southeast Asian ancestry: risk rises at lower BMI than the standard table assumes.
• Anyone whose weight has changed fast for reasons other than fat: illness, fluid shifts, a new diuretic.
• People at the extremes of height, where the height-squared term biases the result in a known direction.

BMI after 65: the same number, a different meaning

The BMI thresholds were drawn largely from middle-aged cohorts, and they do not transfer cleanly to later life. Winter and colleagues pooled 32 studies covering 197,940 community-dwelling adults aged 65 and over, with an average of twelve years of follow-up, and found the mortality curve sitting in a different place. Using BMI 23.0–23.9 as the reference, mortality was 12% higher at BMI 21.0–21.9 and 19% higher at 20.0–20.9. Across the range the standard table calls 'overweight', no increase in mortality risk was detected. Risk began climbing again above BMI 33.

The reason is not that fat becomes harmless with age. It is that BMI in older adults is measuring something different: below about 23 it increasingly reflects lost muscle and bone rather than lost fat, and low muscle mass predicts falls, fracture, slow recovery from illness and death. A BMI drifting down through the twenties in someone over 70 is a signal to investigate, not a success.

The practical reading is that a BMI of 26 in a 75-year-old is not the same finding as a BMI of 26 in a 35-year-old, and a BMI of 21 is more worth a conversation in the older person than in the younger one. What matters more at that age is the direction of travel, grip strength, and whether the person is losing weight without trying. Unintentional loss of more than about 5% of body weight in six months warrants medical attention regardless of where BMI sits.

Children, teenagers and BMI-for-age

Adult BMI categories are not merely imprecise for under-18s; they are invalid. A healthy child's BMI falls through early childhood, bottoms out somewhere around five or six, and then climbs steadily through puberty. A BMI of 17 is unremarkable for a seven-year-old and low for a seventeen-year-old. Applying 18.5 and 25 to a growing body produces a number with no defensible interpretation.

Children and adolescents are therefore assessed as a percentile against a reference population of the same age and sex. On the US CDC growth reference, which covers ages 2 to 20, below the 5th percentile is underweight, the 5th up to the 85th is healthy weight, the 85th up to the 95th is overweight, and the 95th percentile or above is obesity. Above the 97th percentile the percentile scale compresses badly, so the CDC now expresses severe obesity as a percentage of the 95th percentile instead: 120% marks class 2 and 140% class 3. Our child BMI calculator does that arithmetic, including the percent-of-the-95th figure.

Two cautions come with the percentile. It is a position in a reference distribution, not a diagnosis. A muscular fifteen-year-old and a sedentary one can land in the same band for entirely different reasons. And the reference matters: the CDC charts are built from US survey data collected between 1963 and 1994, while the WHO publishes its own growth reference for ages 5 to 19 drawn from a multi-country sample. The two disagree by a few percentile points, most noticeably in adolescence, so a child's band can shift simply by changing charts.

BMI in pregnancy

BMI calculated during pregnancy is meaningless. The formula assumes the mass it is weighing belongs to one body; from the second trimester onwards a large and growing share of it is foetus, placenta, amniotic fluid, expanded blood volume and breast tissue. There is no adjustment that rescues it, which is why antenatal care never asks for a current BMI.

What clinicians do use is pre-pregnancy BMI, calculated from weight before conception or at the first booking appointment, because that number sets the recommended range for total gestational weight gain. The Institute of Medicine's 2009 guidelines, still the reference used across most of the English-speaking world, tie the recommended gain for a singleton pregnancy to that starting category: the higher the pre-pregnancy BMI, the lower the recommended gain.

After delivery, BMI stays unreliable for a while: fluid shifts, blood volume returning to baseline and the physiology of lactation all move the scale independently of fat mass. Waiting six to twelve weeks before reading anything into a BMI, and comparing against the pre-pregnancy figure rather than a chart, gives a much more honest picture. Our pregnancy weight gain calculator works from the pre-pregnancy number and the IOM ranges below.

When to ignore BMI entirely

There is a difference between BMI being imprecise and BMI being invalid. In the situations below the formula's assumption (that total mass is a reasonable proxy for fat mass on a body of standard composition) is not slightly wrong but structurally broken, and the number produced should not be interpreted at all. Reaching for a different tool is not a workaround; it is the correct procedure.

In most of these cases a waist measurement, a body fat estimate or a purpose-built chart does the job instead. Where none of them applies, the honest answer is that no home calculation will tell you anything useful and the question belongs with a clinician who can measure directly.

• During pregnancy, and for the first six to twelve weeks after delivery.
• Anyone under 18: use BMI-for-age percentiles instead.
• Significant oedema, ascites or fluid retention from heart, liver or kidney disease, where kilograms are water.
• After limb amputation, or with a congenital limb difference: the height-to-mass relationship the formula assumes no longer holds.
• Skeletal dysplasia and other conditions causing disproportionate stature, where height² is not comparable with the reference population.
• Advanced muscle-wasting conditions, and during or shortly after a serious illness with rapid weight change.

How to actually use it

Treat BMI as the first sentence of a conversation, not the verdict. Compute it, then pair it with a waist measurement, thirty extra seconds that resolves most of the number's blind spots. If both are elevated, the signal is real and worth acting on. If they disagree, the waist-side measures are usually the truer guide, and a body fat estimate settles it. And regardless of the numbers, trends beat snapshots: a stable BMI of 27 with a shrinking waist is a very different story from the same 27 drifting upward.

If you want one habit out of this guide, make it this: record weight, height and waist on the same day every month, in the same conditions, and keep the list. Three readings tell you almost nothing; twelve tell you the direction, which is the only part of the number that reliably means something.

Frequently asked questions

Put it into practice

Run your own numbers through the BMI calculator, the waist-to-height calculator, the healthy weight range calculator, the child BMI calculator and the pregnancy gain calculator. Related reading: BMI for athletes, 2025 obesity definition and Asian BMI cut-offs.

Sources

1. Keys A, et al. Indices of relative weight and obesity. J Chronic Dis 1972;25:329–43. doi.org/10.1016/0021-9681(72)90027-6
2. WHO Expert Committee. Physical status: the use and interpretation of anthropometry. TRS 854, 1995. www.who.int/publications/i/item/9241208546
3. WHO Consultation on Obesity. Obesity: preventing and managing the global epidemic. WHO Technical Report Series 894, Geneva, 2000. iris.who.int/handle/10665/42330
4. Winter JE, MacInnis RJ, Wattanapenpaiboon N, Nowson CA. BMI and all-cause mortality in older adults: a meta-analysis. Am J Clin Nutr 2014;99:875–90. doi.org/10.3945/ajcn.113.068122
5. Institute of Medicine and National Research Council. Weight Gain During Pregnancy: Reexamining the Guidelines. Washington DC: National Academies Press, 2009. doi.org/10.17226/12584
6. CDC. Growth charts and extended BMI-for-age growth charts. National Center for Health Statistics. www.cdc.gov/growthcharts/cdc-growth-charts.htm
7. WHO Expert Consultation. Appropriate body-mass index for Asian populations. Lancet 2004;363:157–63. doi.org/10.1016/S0140-6736(03)15268-3
8. Rubino F, et al. Definition and diagnostic criteria of clinical obesity. Lancet Diabetes Endocrinol 2025;13:221–262. doi.org/10.1016/S2213-8587(24)00316-4