BMI Calculator (Standard and Asian-Indian Cut-offs)
BMI Calculator: the Figure, the Published Category It Rounds Into, and the Asian-Indian Cut-offs Shown Beside the Standard Ones
Body mass index from height and weight, in metric or imperial, printed to one decimal place and categorised on that printed figure rather than on the raw float — so the number you can see is the number that chose the row. The standard boundaries are taken from the CDC, a US Government work. The Asian-Indian cut-offs of 23 and 25 from Indian national guidance are shown alongside them and never swapped in silently, because for most readers of this site they are the more relevant pair and because moving somebody’s category without telling them is not an improvement. What BMI cannot see is on the page, not in a footnote: it is a ratio of weight to height squared, it was devised as a population statistic, and it knows nothing about muscle, bone, where fat sits, age or ancestry.
These are measurements of body size. They are not a judgement about you and they are not a diagnosis. They cannot see muscle, bone, where fat sits, or how healthy you are — two people with the same figure can be in very different health. Treat anything here as one rough signal among many, and speak to a doctor or dietitian before making a substantial change to how you eat.
A weight-to-height ratio, and the published band the printed figure falls in
170 cm, 70 kg, metric
The ratio, the exact form it is computed in, and the figure the band test is applied to
- W, H
- weight in kilograms and height in metres; Hcm is the same height in centimetres. Imperial input is converted first at exactly 2.54 cm per inch and 0.45359237 kg per pound
- × 10000 / Hcm²
- algebraically identical to W / (Hcm/100)² and numerically better. Dividing a height in centimetres by 100 produces a number that is usually not representable in binary, and squaring it compounds the error; keeping the height in centimetres means that for any height entered to the nearest millimetre the squaring is exact. Swept over 1,542,901 pairs on a 1 mm × 100 g grid, the metre form places six of them in the wrong band and this form places none
- round(10 × BMI) / 10
- the figure the page prints, and the figure the band test is applied to, so what the reader sees is what chose the band. This is the fix carried forward from the pregnancy weight-gain page in the previous batch. It costs 0.05 off each boundary and it buys the guarantee that a printed 25.0 is never classified as below 25
- 18.5 / 25 / 30
- the standard adult boundaries as the CDC prints them, with obesity subdivided at 35 and 40. A US Government work, which is why they are taken from there
- 23 / 25
- the Asian-Indian cut-offs from Indian national guidance (Misra and colleagues, 2009, revised 2025): overweight from 23.0, obesity from 25.0. Shown beside the standard pair throughout and never substituted for it. A third pair for the same population circulates widely and comes from a WHO expert consultation; WHO publications carry a NonCommercial licence term this site cannot satisfy, so that pair is named but not printed anywhere on this page
- ∂BMI/∂H
- −2·BMI/H. The index falls twice as fast with height as it rises with weight, proportionally, which is why a mis-remembered height moves the answer more than a mis-remembered weight
- 1.3 · W / H^2.5
- Trefethen’s alternative, calibrated so that it equals BMI at H = 1.69 m, since 1.69^0.5 = 1.3. Printed here not as a better index but as evidence that the exponent is a modelling choice rather than a fact
Worked example
170 cm, 70 kg, metric
The ratio, in the form the page actually uses. 70 × 10000 / 170² = 700000 / 28900 = 24.221453… kg/m², printed as 24.2. Done the other way, 70 / (1.70)², a double-precision machine gives 24.221453287197235 as well — here the two forms agree. They do not always: at 160 cm and 64 kg the centimetre form gives exactly 25 and the metre form gives 24.999999999999996, and a strict test against 25 then puts the same body in two different bands depending on which algebra the calculator was written in.
The band. 24.2 is at least 18.5 and less than 25, so the figure falls in the second band and the name the CDC prints for it is "Healthy Weight". Note what the page has and has not said. It has said which published band a number falls in. It has not said that this body is healthy, that 24.2 is better than 26.4, or that anything should change. The band is 6.5 kg/m² wide, which at this height is about 19 kg of weight, and health varies enormously inside it.
How close that is to being a different answer. The nearest boundary is 25, which is 0.80 kg/m² away. The derivative with respect to height is 2 × 24.2 / 170 = 0.285 kg/m² per centimetre, so a height mis-measured by 2.8 cm — which is roughly the difference between a height measured against a wall and one recalled from memory — moves this reader across the boundary on its own. Per kilogram the figure moves by 10000 / 28900 = 0.346, so 2.3 kg of weight does the same thing, and about a kilogram of that is the ordinary noise on a single weighing.
The same body on the Asian-Indian cut-offs. 24.2 is above 23.0 and below 25.0. On the standard boundaries it is in the band the CDC calls "Healthy Weight"; on Indian national guidance it is above the overweight threshold of 23 and below the obesity threshold of 25. Both are printed and neither replaces the other. For a reader at 170 cm the two thresholds are 66.5 kg and 72.3 kg: 5.8 kg of weight separates one country's guidance from another's on the same body. This site's readership is largely Indian, which is why that gap is on the page rather than in a footnote.
The exponent, which is a choice. On Trefethen's 2.5 exponent the same body reads 24.15, almost the same — because 170 cm is within a centimetre of the 1.69 m at which the two indices are calibrated to agree. Move the height and they separate: at 1.50 m and 70 kg the classic index is 31.1 and the 2.5 exponent gives 33.0, and at 1.90 m the classic index is 19.4 and the 2.5 exponent gives 18.3. The ponderal index, weight over height cubed, gives 14.25 kg/m³ here. Three indices, three scales, one body.
What this page refuses to print. It does not print a weight range for your height, it does not print the weight that would put you at a BMI of 25 or 23, and it does not print how much to lose or gain. Those are the outputs almost every other BMI calculator leads with and they are the ones with the worst failure mode: a number that looks like an instruction, produced by a page that knows two facts about the person reading it. What belongs there instead is a conversation with a doctor or a dietitian who can see the rest.
What the figure cannot see, in one sentence each. Muscle and fat weigh differently and this ratio cannot tell them apart; bone and frame vary and this ratio cannot see them; two people with the same figure can carry their fat in completely different places, and where it sits matters more for cardiometabolic risk than how much there is, which is the reason the waist pages in this set exist; and the relationship between the ratio and body fat differs systematically by ancestry, age and sex, which is the whole reason there is more than one set of cut-offs to print.
The standard adult bands as the CDC publishes them, and the boundary this page actually applies
| The name CDC prints | As CDC prints the range | One-decimal figures that fall here | Boundary this page operates on |
|---|---|---|---|
| Underweight | Less than 18.5 | 10.0 to 18.4 | below 18.45 |
| Healthy Weight | 18.5 to less than 25 | 18.5 to 24.9 | 18.45 to below 24.95 |
| Overweight | 25 to less than 30 | 25.0 to 29.9 | 24.95 to below 29.95 |
| Obesity, Class 1 | 30 to less than 35 | 30.0 to 34.9 | 29.95 to below 34.95 |
| Obesity, Class 2 | 35 to less than 40 | 35.0 to 39.9 | 34.95 to below 39.95 |
| Obesity, Class 3 (Severe Obesity) | 40 or greater | 40.0 and above | 39.95 and above |
The same figure read on the standard boundaries and on Indian national guidance, with the weight each threshold corresponds to at 170 cm
| BMI (1 dp) | Standard band (CDC) | Indian national guidance | Do they agree? | Weight at 170 cm |
|---|---|---|---|---|
| 17.0 | Underweight | Underweight | yes | 49.1 kg |
| 20.0 | Healthy Weight | Normal | yes | 57.8 kg |
| 22.9 | Healthy Weight | Normal | yes | 66.2 kg |
| 23.0 | Healthy Weight | Overweight | NO | 66.5 kg |
| 24.9 | Healthy Weight | Overweight | NO | 71.9 kg |
| 25.0 | Overweight | Obesity | direction yes, word no | 72.3 kg |
| 27.0 | Overweight | Obesity | direction yes, word no | 78.0 kg |
| 30.0 | Obesity, Class 1 | Obesity | yes | 86.7 kg |
| 35.0 | Obesity, Class 2 | Obesity | yes | 101.2 kg |
The floating-point defect at the band boundaries, measured rather than asserted
| Test | Grid | Pairs misplaced | Which ones |
|---|---|---|---|
| W / (Hcm/100)², compared raw against the boundary | 1 mm × 100 g, 140.0–210.0 cm, 30.0–250.0 kg (1,542,901 pairs) | 6 | 155 cm/96.1 kg at 40; 160 cm/64 kg at 25; 160 cm/76.8 kg at 30; 160 cm/89.6 kg at 35; 160 cm/102.4 kg and 155 cm/96.1 kg at 40 |
| W × 10000 / Hcm², compared raw against the boundary | the same grid | 0 | none |
| Either form, compared after rounding to one decimal place | the same grid | 0 | none, and the two forms also agree on the band for every pair |
| Either form, printed figure versus categorising figure | the same grid | 0 | the printed one-decimal figure always equals the figure the band test used |
| Pairs whose band moves because the test is on the rounded figure | the same grid, at the 25 boundary | about 1,100 | every pair whose raw ratio is from 24.95 to just below 25.00. This is the intended effect of the fix, not a defect |
What the ratio cannot see, with the published figures rather than the usual hand-waving
| The thing BMI cannot see | What the evidence says | Source |
|---|---|---|
| Fat as against everything else | Against body fat measured directly, a BMI of 30 or more identified 36% of men and 49% of women who had excess body fat by the criterion used; specificity was 95% and 99%. In other words the cut-off misses more than half the people it is taken to describe, and it misses them in the direction of false reassurance | Romero-Corral and colleagues, Int J Obes 2008;32:959–66, 13,601 adults aged 20–79.9 |
| Where the fat sits | In a 27,000-participant case-control study across 52 countries, the odds of a first myocardial infarction rose across every quintile of waist-to-hip ratio, reaching 2.52 (2.31–2.74) for the top quintile against the bottom, while BMI’s own odds ratio fell to 0.98 (0.88–1.09) once waist-to-hip ratio was adjusted for | Yusuf and colleagues, Lancet 2005;366:1640–9 |
| Ancestry | The BMI values in different ethnic groups that correspond to a BMI of 30 in white men and women range from 21.5 to 26 — a spread of four and a half units in where the same level of risk sits | Ashwell and Gibson, BMC Med 2014;12:207 (CC BY 4.0) |
| What it was built for | Quetelet produced weight over height squared in the 1830s while studying the statistical distribution of human measurements in populations, not to assess an individual. Keys and colleagues named it the body mass index in 1972 after testing candidate indices for the one that correlated best with weight and least with height in population samples | Quetelet 1835; Keys and colleagues, J Chronic Dis 1972;25:329–43; Wells, Int J Epidemiol 2014;43:672–4 |
| The exponent | Nothing in anatomy fixes the exponent at 2. On 1.3 × W / H^2.5 the same 70 kg body reads 33.0 at 1.50 m and 18.3 at 1.90 m, where BMI reads 31.1 and 19.4 | Trefethen, University of Oxford, 2013 (New BMI) |
A population statistic from the 1830s, what it is for, what it cannot do, and why this page prints two sets of cut-offs instead of choosing one
Body mass index was not invented to describe a person. Adolphe Quetelet was a Belgian astronomer and statistician who spent the 1830s doing something nobody had done systematically before: measuring the distributions of human characteristics across populations and asking whether they followed regular laws. In the course of that work he noticed that weight does not scale with the cube of height, as a simple geometric argument about similar solids would predict, and that dividing weight by height squared produced a quantity that was roughly stable across adults of different heights. That is the whole provenance of the ratio. It was a tool for comparing groups, published in a book about the statistics of human populations, and it carried no cut-offs, no categories and no clinical claim. The name came much later: in 1972 Ancel Keys and colleagues tested the candidate height-adjustments against each other and chose weight over height squared because it correlated best with weight and least with height in the samples they had, and it is their paper that calls it the body mass index. Even there the argument was explicitly about which index best served population comparisons.
So the honest framing of this page is that it reports a number, says which published band the number falls in, and then spends most of its space on the things the number cannot see. That is not modesty for its own sake. It is because the gap between what BMI measures and what readers want it to measure is large and has been quantified. In 13,601 American adults with body fat measured directly, a BMI of 30 or more picked out 36% of the men and 49% of the women who had excess body fat by the study’s criterion. Specificity was excellent, which is to say that almost nobody the cut-off flagged was flagged wrongly; sensitivity was close to a coin toss, which is to say that the cut-off’s most common error is telling people there is nothing to see. A calculator that prints a band and stops has not told that reader anything they can use.
The ratio cannot distinguish muscle from fat, and that is not a quibble about bodybuilders. Weight is weight. A person who has gained five kilograms of muscle and a person who has gained five kilograms of fat move the same distance up the same scale, and at the population level the second case is so much more common that the index works anyway. At the level of the individual in front of the screen it does not. The same arithmetic also cannot see bone density or frame width, cannot see where fat is distributed, and cannot see that the relationship between the ratio and the amount of fat shifts with age as muscle is lost and with sex throughout life. Of all of these the distribution of fat is the one with the strongest evidence behind it: in a case-control study of 27,000 people across 52 countries, the risk of a first heart attack rose steadily across quintiles of waist-to-hip ratio and BMI’s own association disappeared entirely once waist-to-hip ratio was accounted for. That result is why this set has waist pages at all, and why the waist-to-height page is worth reading next: it asks a related question with better evidence behind it, and it needs one extra measurement.
Now the two sets of cut-offs, which for this site’s readers is the most consequential thing on the page. The standard boundaries — 18.5, 25 and 30, with obesity subdivided at 35 and 40 — are printed by the CDC, which is a US Government work, and that is where this page takes them from rather than from a WHO classification table. Indian national guidance sets different thresholds: a consensus statement led by Anoop Misra and published in 2009 put overweight at a BMI of 23.0 to 24.9 and obesity at 25.0 and above for Asian Indians, on the evidence that body fat, insulin resistance and the cardiometabolic consequences of adiposity appear at a lower BMI in this population. A 2025 revision by largely the same group carried those thresholds forward and added a two-stage definition in which a raised BMI on its own is distinguished from a raised BMI accompanied by measurable effects on function or an accompanying condition. There is a third pair in circulation for the same population, which comes from a WHO expert consultation. WHO publications are licensed CC BY-NC-SA 3.0 IGO and the NonCommercial term applies to a site carrying advertising, so this page neither uses that pair nor prints it — if you have met an Asian BMI cut-off above 25 somewhere else, that is where it is from, and it is not Indian national guidance.
This page prints both sets and switches neither, and the reason is worth stating because the alternative looks helpful. A calculator that detects an Indian reader and quietly applies 23 and 25 has changed the label on that reader’s body without telling them, and has also made their figure incomparable with every other calculator, every research paper and every clinical note they will ever encounter. A calculator that applies only the standard boundaries tells a very large number of South Asian readers that there is nothing to look at when the national guidance for their own population says otherwise. Printing both, with the distance to each threshold in the rows above, is the only version of this that leaves the reader better informed rather than differently misled. Between 23.0 and 24.9 the two positions give opposite answers, and at 170 cm that is a band 5.4 kg wide; from 25.0 upward they agree on the direction and differ only in which word they attach. Which set applies to a particular person is a question for a clinician who knows their family history, their blood pressure, their lipids and their glucose, none of which this page has.
Two numbers this page deliberately does not print. The first is a weight range for your height. It is the output almost every BMI calculator leads with, it is trivial to compute from the boundaries, and it is the single most harmful thing such a page can produce: a specific number of kilograms, generated from two measurements by a web page that knows nothing else, which arrives looking exactly like an instruction. The second is the number of kilograms between the reader and the next band. Both are targets, and a target produced by an instrument this blunt is not information. What replaces them above is the sensitivity arithmetic: how much the figure moves per centimetre and per unit of weight, and how large a measurement error would be enough on its own to change the band. That tells you how much to trust the figure, which is the question a reader is actually in a position to act on.
The arithmetic, briefly, because two details in it are usually wrong elsewhere. The first is the algebra. Writing the ratio as weight divided by the square of a height in metres means dividing a centimetre measurement by 100 and squaring the result, and neither step is exact in binary floating point; the classic consequence is that 64 kg at 160 cm, which is exactly 25, computes as 24.999999999999996 and a strict test against 25 puts it in the band below. Computing it instead as weight times ten thousand divided by the square of the height in centimetres is algebraically the same thing and is exact for any height entered to the nearest millimetre. Over 1,542,901 height-and-weight pairs on a one-millimetre by one-hundred-gram grid, the first form puts six in the wrong band and the second puts none. The second detail is the rounding. This page applies the band test to the one-decimal figure it prints, which moves each effective boundary down by 0.05 and reclassifies roughly 1,100 pairs on that grid at the 25 boundary. That cost is paid deliberately in exchange for a guarantee: no reader is ever shown a figure of 25.0 and told they are in the band for figures below 25.
For children and teenagers this page is the wrong instrument entirely, and not by a little. A growing child’s normal BMI changes every year, so there is no fixed cut-off to compare against; the reference is a percentile on an age-and-sex growth chart, and applying an adult band to a twelve-year-old produces a label with no meaning. A BMI-for-age percentile page is being built for this set and is deliberately not linked here yet rather than linked somewhere that does not exist. In pregnancy the index is used differently again: what matters is the pre-pregnancy BMI, which selects a published range of total weight gain, and that is a separate page with its own arithmetic. For drug dosing, the weights derived from height are different quantities with different purposes and are not interchangeable with BMI or with each other: ideal body weight was published by Devine in 1974 to size gentamicin doses and is not a statement about what anyone ought to weigh, and lean body weight responds to actual weight and is used as a size descriptor for some drugs.
If you want to go further than this page can. Two of the three obvious next measurements are here: waist-to-height ratio, which needs one tape measurement and has better evidence behind it for cardiometabolic risk than BMI does, and waist-to-hip ratio, which is the measure the heart-attack study above used and which this site presents without a threshold table for reasons that page explains at length. Body fat percentage is the third, and the honest warning there is that the field methods for estimating it carry errors of several percentage points, which is larger than most of the differences anybody cares about. Reading all four and finding that they disagree is a more accurate picture of your own uncertainty than any one of them on its own.
Frequently asked questions
Which cut-offs should I use, the standard ones or 23 and 25?
This page will not choose for you, and that is a position rather than an evasion. The standard boundaries are what the CDC publishes, what essentially all of the research literature is indexed on, and what any other calculator or clinical note you encounter will have used. The 23 and 25 thresholds are Indian national guidance, from a consensus statement led by Misra and colleagues in 2009 and carried forward in a 2025 revision, and they rest on evidence that body fat and cardiometabolic risk arrive at a lower BMI in South Asian populations. Both are on the page with the distance to each threshold, because the one thing a calculator should not do is move somebody’s category without telling them. Which applies to you in any clinically meaningful sense depends on things this page cannot see — family history, blood pressure, lipids, glucose, waist circumference — and is a conversation with a doctor. If you simply need a figure that is comparable with everybody else’s, use the standard bands and read the 23 line as additional information.
Why does this page not tell me what I should weigh?
Because it does not know, and the number it could print would not carry that caveat. A weight range derived from the BMI boundaries is arithmetic on two measurements; it has no information about muscle, frame, age, ancestry, medical history or anything else, and it arrives in the shape of an instruction. For readers who are anxious about their bodies — and a BMI page has a great many of them — a specific kilogram figure produced by a machine is the most actively harmful thing on a page like this. The sensitivity rows above are what replaces it: they tell you how much the figure moves per centimetre of height and per unit of weight, which is a statement about how much to trust the number rather than a target to chase. A weight to aim at, if one is appropriate at all, comes from a clinician or a dietitian with the rest of the picture.
Why is a BMI of exactly 25.0 counted as overweight and not as healthy weight?
Because that is how the published bands are defined: the CDC prints the second band as 18.5 to less than 25 and the third as 25 to less than 30, so 25.0 belongs to the third. The interesting part is making a computer agree. The raw ratio is a binary floating-point number, and a value that is exactly 25 in ordinary arithmetic frequently is not exactly 25 in double precision — 64 kg at 160 cm is the standard example, computing as 24.999999999999996 if the height is first converted to metres. A strict test against 25 then hands that reader the wrong row. This page defends against it twice: it computes the ratio in centimetres, where the squaring is exact for any height given to the nearest millimetre, and it applies the band test to the one-decimal figure it prints. The side effect is that the operating boundary is 24.95 rather than 25.00, which the page states wherever it matters.
Is BMI useless, then?
No, and the page is careful about this. At the level of populations BMI tracks risk well enough to be worth collecting: it is cheap, it needs no equipment beyond a scale and a tape, it is reproducible between observers, and decades of epidemiology are indexed on it, which means it is the only body-size measure for which the long-run outcome data exists at all. The problem is the inference from that to the individual. The index has excellent specificity and poor sensitivity for excess body fat, so its characteristic failure is false reassurance rather than false alarm, and the thing it is blindest to — where the fat sits — turns out to be the thing with the strongest link to cardiometabolic risk. Treat the figure as one cheap signal with a known blind spot, and take the waist measurement that covers the blind spot.
What is the 703 formula, and why is it not used here?
It is the imperial shortcut: BMI = 703 × pounds / inches². The constant 703 approximates 10000 × 0.45359237 / 2.54², which is 703.0696, so the shortcut is the same conversion with the factor rounded to three digits. The error it introduces is about one part in ten thousand, which moves a figure of 25 by roughly 0.0025 — invisible at one decimal place and incapable of changing a band. This page converts with the exact legal factors anyway, because there is no reason not to, and because a reader comparing two calculators is better served by knowing exactly what each one did.
Why do you print a “new BMI” with a 2.5 exponent and a ponderal index?
To make a point that is otherwise hard to believe: the exponent is a choice. Quetelet noticed that weight does not scale with height cubed; Keys and colleagues settled on height squared in 1972 because it behaved best statistically in their samples. Nick Trefethen of Oxford proposed 1.3 × weight / height^2.5 in 2013, calibrated to agree with the classic index at 1.69 m, on the argument that 2 under-corrects for height. At 70 kg the two indices give 31.1 and 33.0 at 1.50 m, and 19.4 and 18.3 at 1.90 m. Neither is the true one. The practical consequence is that short people and tall people are not being graded on the same instrument, and which exponent you prefer determines who the index flatters — which is worth knowing before reading very much into a figure 0.8 from a boundary.
Can I use this for a child?
No. A child’s BMI is compared against an age-and-sex percentile reference rather than a fixed cut-off, because the normal value changes with every year of growth: the same figure that is unremarkable at sixteen is not at six. Applying an adult band to a child produces a label with no meaning behind it. A BMI-for-age percentile page is being built for this set separately; until it exists this page deliberately does not link anywhere for it rather than send readers to a page that is not there. If you need a paediatric figure now, the child’s own clinician has the growth chart.
The number changed since last time and I have not changed. Why?
Almost always the height. BMI goes as the square of height, so the derivative is twice as large proportionally for height as for weight, and heights vary with the time of day, with posture, with whether shoes were on, with who was holding the measure and with whether the figure was measured at all rather than remembered. At 170 cm, one centimetre is worth 0.28 kg/m², so three centimetres of difference between a measured and a remembered height is worth about 0.85 — enough on its own to cross a band boundary from most starting points. Weight contributes its own kilogram or so of noise from clothing, food, drink, bladder and bowel. The rows above print both sensitivities for the figures entered, which is the quickest way to see whether a change is a change in a body or a change in a measurement.
Does this page carry a BMI-based body fat estimate?
Deliberately not. Equations that predict body fat percentage from BMI, age and sex exist and are widely implemented, and they inherit every weakness of BMI plus a regression error of their own, which makes the output look like a measurement of fat when it is a rearrangement of a weight-to-height ratio. If you want a body fat estimate, the circumference and skinfold methods on the body fat page are at least measuring something different from height and weight, and that page prints the published errors — which are several percentage points — beside every figure.
Related calculators
References
- Centers for Disease Control and Prevention. Adult BMI Categories. cdc.gov/bmi/adult-calculator/bmi-categories.html. Primary source for the standard boundaries implemented here, taken from a US Government work rather than from a WHO classification table: Underweight below 18.5; Healthy Weight 18.5 to less than 25; Overweight 25 to less than 30; Obesity 30 or greater, subdivided as Class 1 (30 to less than 35), Class 2 (35 to less than 40) and Class 3 or Severe Obesity (40 or greater). The page also carries the CDC’s own statement that BMI is one potential health indicator and should be considered alongside other factors, which is the position this page takes throughout.
- Quetelet LAJ. Sur l’homme et le développement de ses facultés, ou Essai de physique sociale. Paris: Bachelier, 1835 (English translation, A Treatise on Man and the Development of his Faculties, Edinburgh, 1842). The origin of weight divided by height squared, produced in the course of a statistical study of the distribution of human measurements across populations. Long out of copyright. Cited here for provenance: the ratio was a population statistic with no cut-offs, no categories and no individual diagnostic claim attached to it.
- Keys A, Fidanza F, Karvonen MJ, Kimura N, Taylor HL. Indices of relative weight and obesity. J Chronic Dis 1972;25(6–7):329–43. The paper that named the body mass index. Keys and colleagues compared candidate height-adjustments on two criteria — correlation with weight maximised, correlation with height minimised, and correlation with measured adiposity — and selected weight over height squared. Cited for the name, the selection criteria and the fact that the argument was about population indices.
- Wells JCK. Commentary: The paradox of body mass index in obesity assessment: not a good index of adiposity, but not a bad index of cardio-metabolic risk. Int J Epidemiol 2014;43(3):672–4. doi:10.1093/ije/dyu060. Cited for the historical reconstruction of Quetelet’s reasoning (including his observation that weight does not scale as the cube of height) and for the framing this page adopts: that the index is poor at the thing it is assumed to measure and better than it deserves to be at the thing it is used to predict.
- Misra A, Chowbey P, Makkar BM and colleagues. Consensus statement for diagnosis of obesity, abdominal obesity and the metabolic syndrome for Asian Indians and recommendations for physical activity, medical and surgical management. J Assoc Physicians India 2009;57:163–70. The source of the Asian-Indian cut-offs used on this page: overweight at a BMI of 23.0 to 24.9 and obesity at 25.0 and above, with abdominal obesity at a waist circumference of 90 cm or more in men and 80 cm or more in women. Cited as two figures with their provenance, not reproduced as a criteria table.
- Misra A, Vikram NK, Ghosh A, Ranjan P, Gulati S and the India Obesity Commission. Revised definition of obesity in Asian Indians living in India. Diabetes Metab Syndr 2025;19:102989. doi:10.1016/j.dsx.2024.102989. The 2025 revision, which carries the 2009 BMI and waist thresholds forward (overweight 23 to 24.9, obesity 25 or more; waist 90 cm in men and 80 cm in women) and adds a two-stage definition separating raised adiposity on its own from raised adiposity with measurable effects on organ or physical function. Also the source for the statement on this site’s waist-to-height page that Indian national guidance now prefers waist-to-height ratio with a cut-off above 0.5 to waist-to-hip ratio. Cited for its figures and the existence and shape of its recommendation; its finer BMI grading and its recommendation text are not reproduced.
- PROVENANCE AND LICENSING NOTE ON THE CUT-OFFS, recorded deliberately. Three pairs of Asian-specific BMI thresholds are in circulation and they are routinely confused. One pair, whose upper figure sits above 25, is from a WHO expert consultation (2004); WHO publications are licensed CC BY-NC-SA 3.0 IGO, the NonCommercial term applies to this site because it carries advertising, and that pair is therefore neither used nor printed anywhere on this page, on the same reasoning the waist-to-hip page in this category applies to WHO’s waist-to-hip cut-offs. Naming the figure while disclaiming it was considered and rejected as inconsistent with that page. 23 and 25 are Indian national guidance from the Misra consensus statements above and are what this page prints. 18.5, 25 and 30 are taken from the CDC, a US Government work in the public domain. Reaching WHO material through a third party that has republished it would be the same breach; copyright does not launder, and no such route was used.
- Romero-Corral A, Somers VK, Sierra-Johnson J and colleagues. Accuracy of body mass index in diagnosing obesity in the adult general population. Int J Obes 2008;32(6):959–66. 13,601 adults aged 20 to 79.9 with body fat measured directly. A BMI of 30 or more had a sensitivity of 36% (95% CI 35–37) in men and 49% (48–50) in women for excess body fat, with specificity 95% and 99%. The body-fat criterion the authors used was more than 25% in men and more than 35% in women; the provenance of that criterion is a WHO figure, so this page cites the study’s result and does not adopt its criterion, which is also why no body fat threshold appears anywhere on this page.
- Yusuf S, Hawken S, Ôunpuu S and colleagues. Obesity and the risk of myocardial infarction in 27,000 participants from 52 countries: a case-control study. Lancet 2005;366(9497):1640–9. doi:10.1016/S0140-6736(05)67663-5. Cited for the comparison between BMI and fat distribution: the odds ratio for a first myocardial infarction in the top quintile of waist-to-hip ratio against the bottom was 2.52 (2.31–2.74), while BMI’s own odds ratio of 1.44 (1.32–1.57) fell to 0.98 (0.88–1.09) after adjustment, and the population attributable risk for the top two quintiles was 24.3% for waist-to-hip ratio against 7.7% for BMI.
- Ashwell M, Gibson S. A proposal for a primary screening tool: ‘Keep your waist circumference to less than half your height’. BMC Med 2014;12:207. doi:10.1186/s12916-014-0207-1. Open Access under CC BY 4.0, which permits reuse on a commercial site with attribution. Cited here for one figure: the BMI values in different ethnic groups equivalent to a BMI of 30 in white men and women range from 21.5 to 26.
- Trefethen LN. New BMI (New Body Mass Index). University of Oxford, 2013, people.maths.ox.ac.uk/trefethen/bmi.html, following a letter in The Economist, 5 January 2013. Source of the alternative index printed in the rows above, 1.3 × weight(kg) / height(m)^2.5, with the constant 1.3 chosen as the square root of 1.69 so that the new index equals the classic one at a height of 1.69 m. Printed on this page as evidence that the exponent is a modelling choice, not as a recommended replacement. An arithmetic rearrangement of two measurements is nobody’s intellectual property.
- Floating-point behaviour at the band boundaries: established for this page by sweeping 1,542,901 height-and-weight pairs on a one-millimetre by one-hundred-gram grid (140.0 to 210.0 cm, 30.0 to 250.0 kg) in the live engine rather than reasoning from the specification. Computing the ratio as weight / (height in cm / 100)² and comparing the raw result against the boundary misplaces six pairs; computing it as weight × 10000 / (height in cm)² misplaces none; and categorising on the one-decimal figure misplaces none in either form. Roughly 1,100 pairs at the 25 boundary have their band changed by categorising on the rounded figure, which is the intended effect of that choice rather than a defect. The figure of 94 disagreeing pairs given in this batch’s shared brief could not be reproduced.
CalcEngines health calculators are for education and for checking arithmetic that has already been decided elsewhere. They are not medical advice, they do not decide what to give, and they do not replace the judgement of a doctor, nurse, midwife or dietitian who knows the person in front of them. Every figure depends on the values you enter and on the assumptions stated on the page — check it against the prescription, the product label and your local policy before acting on it.