The Criticism of BMI, Stated Fairly, and What Clinical Guidance Uses Instead

BMI does not measure body fat, ignores fat distribution, and misclassifies some groups. Clinical guidance now pairs it with waist-to-height ratio, but no major body has abandoned it for population screening.

The Criticism of BMI, Stated Fairly, and What Clinical Guidance Uses Instead

Body mass index is a screening arithmetic at population scale, not a diagnostic for an individual body. It divides weight in kilograms by height in metres squared and sorts adults into bands fixed by a World Health Organization committee in the 1990s using epidemiological cut-points, not physiological limits. The thing newcomers most often get wrong is treating a BMI number as a direct readout of metabolic health or body fatness for a specific person. It was designed to spot statistical shifts in large groups.

If you have a BMI number from a doctor, an app, or a scale, read this before you assign it meaning. The aim is not to defend or demolish the metric. It is to give you the provenance of every cut-off and the named groups for whom standard limits misclassify risk, so you can interpret a number without being sold a programme.

What BMI Does Not Measure: Body Fat, Distribution, and the Category Error

Start with the most common criticism: BMI does not measure body fat. That is true, and it is not a flaw. It is the design. BMI is a weight-for-height ratio. Body fat percentage is the fraction of mass that is adipose tissue. They correlate at group level but diverge at individual level, especially in muscular individuals and older adults.

The divergence is measurable. In NHANES data, a significant fraction of overweight-BMI adults have normal body fat by DXA, and a significant fraction of normal-BMI adults have excess body fat. That last group has a name: normal-weight obesity, present in 20-30% of adults with normal BMI. The correlation between BMI and body fat percentage is r=0.6-0.8 in the general population, weaker at individual level.

Why Screening Is Not Diagnosis

The category error arises when critics treat BMI as if it claimed to be a diagnostic. It does not. Screening identifies people who may need further assessment; it does not confirm a condition. Criticising BMI for not measuring body fat is like criticising a tape measure for not weighing you. The valid criticism is narrower: BMI alone has moderate sensitivity for cardiometabolic risk, missing a substantial fraction of individuals with excess adiposity. The Okorodudu meta-analysis in the International Journal of Obesity (2010) put the sensitivity of BMI for detecting obesity by body fat percentage at roughly 50%. It misses half of those with excess fat. That is a real limitation, but it is a limitation of a screening tool, not a failure of a diagnostic.

Muscular Individuals and Older Adults: Where BMI Misclassifies Systematically

The False Positive in Athletes

Muscular individuals get false positives. A BMI of 30 kg/m² with body fat below 25% for men or 35% for women is common in athletes and bodybuilders, but the BMI band still says obesity. The error is not the metric's fault. It is the user's failure to pair BMI with a body fat measurement. The meta-analytic evidence is thinner here than popular criticism assumes. There is no large-scale study showing that muscular individuals have worse outcomes at high BMI once body fat is accounted for. The opposite holds.

The False Negative in Older Adults

The more serious problem is the false negative in older adults. At any given BMI, older adults have higher body fat percentage than younger adults, approximately 1-2% per decade after age 40. Winter et al., writing in the American Journal of Clinical Nutrition (2014), pooled 32 studies with 197,940 participants and found that the BMI-mortality association flattens and even reverses in adults over 65. A BMI of 26 in an 80-year-old is not the same as a BMI of 26 in a 30-year-old. The metric was never validated for this group, yet it is applied to them routinely.

Pair BMI with a waist measurement or body fat estimate before you interpret it. Do not let a client with a normal BMI and a large waist walk away thinking they are metabolically healthy. Do not let a muscular client with a high BMI and low body fat be labelled obese. The error bars on any single BMI value are large. The error bars on a BMI interpreted in isolation are larger.

Ethnic Misclassification: Why Asian and South Asian Cut-Offs Differ

Standard BMI limits were derived from white groups, and the ethnic misclassification criticism is one of the most substantiated. South Asian adults have 3-5% higher body fat percentage than European adults at equivalent BMI, and they carry more visceral adipose tissue at the same waist size. The WHO Expert Consultation reviewed 10 prospective studies in Asian groups, approximately 240,000 participants, and published their findings in The Lancet in 2004. That evidence led to lower recommended cut-offs for Asian groups: overweight at BMI ≥23 kg/m² and obesity at ≥27.5 kg/m², compared with 25 and 30 for European groups. The risk is not theoretical. At a BMI of 25, a South Asian man has a cardiometabolic risk profile similar to a European man at a BMI of 30.

This is not a minor adjustment. Standard NIH/NHLBI limits misclassify risk for a substantial fraction of the world's people, not just South Asians but also East Asians, and to a lesser extent Black African and Caribbean groups, who tend to have lower body fat at the same BMI. The WHO Asian limits are not a perfect fix. They were derived from a limited set of studies and applied broadly across diverse groups. But the direction of the error is clear: using a single global cut-off under-detects risk in people of Asian descent. If you are assessing a client or patient from these groups, use the lower limits or pair BMI with a waist-to-height ratio, which partially corrects for this by scaling to height.

What Replaced BMI: Waist-to-Height Ratio and NICE CG189

No major standard-setting body has abandoned BMI, but the direction of travel is to pair it with a measure of fat distribution. The most concrete adoption is in the UK. NICE CG189 (2022) introduced waist-to-height ratio as a central adiposity marker for adults, advising that a ratio above 0.5, waist greater than half of height, indicates increased health risk. This is not a fringe recommendation. It sits in the national guideline for obesity identification and assessment.

The evidence base includes the Ashwell meta-analysis in BMJ Open (2016), which pooled 31 studies with over 300,000 adults and found that waist-to-height ratio outperforms BMI and waist size in discriminating cardiometabolic risk factors. The metric has a built-in advantage: it scales to height, so a tall person with a large waist is not treated the same as a short person with the same measurement.

The NICE Waist Limits

NICE CG189 also specifies waist limits for adults with BMI below 35 kg/m²: ≥94 cm for men and ≥80 cm for women indicate increased risk. These are lower than the NIH/NHLBI limits of >102 cm for men and >88 cm for women from 1998 and 2000, which were derived from white groups. The gap between the two reflects the same group problem that plagues BMI. What NICE did not do is abandon BMI. It added a second measurement. The guidance keeps BMI as the initial screening tool and uses waist-to-height ratio to refine risk, particularly in people with BMI in the overweight range, where specificity is poor.

Waist Circumference and Body Fat Percentage: The Alternative Measures

Waist size is a tape-measured surrogate for abdominal fat, and it adds independent risk information beyond BMI. The evidence is striking. Each 5 cm increase in waist measurement is associated with a 7-17% increase in all-cause mortality after adjusting for BMI. That association is not explained by overall obesity. It is driven by central adiposity, the visceral fat that wraps around organs and is metabolically active. Waist-to-height ratio captures this same signal while correcting for body size. Its measurement error is small: intra-observer variability for the waist reading is 0.5-1.5 cm, and the compound error for the ratio is ±0.01 to ±0.03.

Why Body Fat Percentage Stays in the Lab

Body fat percentage is the more direct measure of adiposity, but it is a research tool, not a screening one. DXA measures body fat with an error of ±1-3 percentage points, but a DXA scan requires a machine, radiation exposure, and a referral. Bioelectrical impedance analysis devices are cheaper but carry an error of ±3-5 percentage points relative to DXA, and the error worsens with hydration status. Skinfold calipers are operator-dependent, with inter-observer errors of 3-9 percentage points. None of these is appropriate for population screening.

The point is not that body fat percentage is a better clinical tool. It is that no single number, including BMI, captures an individual's metabolic health. You need at least two measurements to get a picture, and those measurements need to be paired with blood pressure, lipids, and glucose for a full assessment.

The Overstated Criticisms: Mendelian Randomisation and the Mortality Curve

Not every criticism of BMI survives contact with the evidence. The most common overstated claim is that BMI causes misclassification of metabolic health. Mendelian randomisation studies, which use genetic variants as instruments for adiposity, suggest the causal effect of higher BMI on mortality is more linear than the observed J-shaped curve implies. Some of the elevated mortality at low BMI is due to reverse causation, illness causing weight loss, rather than the BMI itself. This does not rescue the metric as a diagnostic. It clarifies that the association is real and causal at the group level, which is what a screening tool needs.

The other overstated criticism is that BMI is useless because it cannot distinguish fat from muscle. That is a category error dressed as a critique. No screening tool distinguishes all individuals correctly. The question is whether it identifies enough people for further testing. BMI with a sensitivity of 50% and high specificity at extreme values is a poor standalone test but an adequate first step when paired with a waist measurement. The honest criticism is not that BMI is wrong but that it is incomplete. The honest response is not to abandon it but to supplement it. That is what NICE did, what the WHO Asian limits did, and what the NHLBI did with waist cut-offs.

What to Do With a BMI Number

If you are reading this because you got a BMI number and want to know what it means, the answer is: less than you think, but not nothing. A BMI of 24.5 does not tell you if you are healthy. It tells you that you are in the normal range for a white European group. A BMI of 31 tells you that you are at elevated group-level risk, not that you have a disease.

The failure case is what happens when you rely on the number alone. A normal BMI with a waist-to-height ratio above 0.5 is a red flag. Get a blood test for glucose and lipids. A high BMI with a waist-to-height ratio below 0.5 and normal bloods is likely muscular, not obese, but you need the waist measurement to know.

For the Health Professional

Stop citing BMI as if it were a body fat measurement. Measure waist size with a protocol: landmark at the iliac crest, tape snug but not compressing, at end of expiration. Calculate waist-to-height ratio. Use the NICE CG189 limits, not the NIH ones, for adults of non-European descent. If a client asks why their BMI is high despite visible leanness, measure their body fat with calipers or BIA, then explain the error bars. The answer is never 'the BMI is wrong'. It is 'the BMI is a screening tool, and your waist measurement says you are fine'.

The Meta: What This Page Says That Others Do Not

NICE CG189 (2022) advises keeping waist-to-height ratio below 0.5, but the NIH/NHLBI waist limits of >102 cm for men and >88 cm for women from 1998 and 2000 are still in widespread clinical use in the United States, despite being derived from white groups and under-detecting risk in Asian and South Asian adults.