Understanding Body Mass Index: Screening, Measurement, and What BMI Means

BMI divides weight in kilograms by height in metres squared, then sorts large groups into bands fixed by a World Health Organization expert committee in 1995. That arithmetic was not designed to diagnose an individual body. It was designed to spot statistical shifts across groups. Treat a BMI number as a direct readout of metabolic health or body fatness, and you will misread both the number and the person who carries it.

Every threshold named here carries the body and year that published it. The frame throughout: BMI screens groups, not individuals.

• Formula: Weight in kilograms divided by height in metres squared (kg/m²). Imperial: 703 × weight (lb) / height (in)².
• WHO Adult Classification: Underweight <18.5; normal weight 18.5-24.9; pre-obesity 25.0-29.9; obesity class I 30.0-34.9; class II 35.0-39.9; class III ≥40.0. Source: WHO Technical Report Series 854 (1995), reaffirmed 2000.
• WHO Asian BMI Trigger Points (2004): Increased risk at 23.0-27.4; high risk at ≥27.5. Published in The Lancet, 2004, after review showing elevated type 2 diabetes and cardiovascular risk at lower BMI in Asian groups.
• CDC Child Growth Chart Percentiles: Underweight <5th; healthy weight 5th, <85th; overweight 85th, <95th; obesity ≥95th; severe obesity ≥120% of 95th or BMI ≥35. Source: CDC 2000 Growth Charts, revised 2022 for severe obesity.
• NICE Waist-to-Height Threshold (CG189, 2022): Keep waist-to-height ratio below 0.5 for adults with BMI <35 kg/m², as a supplementary measure to BMI.
• NIH/NHLBI Waist Circumference Thresholds (1998): Increased risk: >102 cm men, >88 cm women. Derived from prospective cohort studies in US adults.
• IDF Waist Circumference (Metabolic Syndrome, 2006): Europid men ≥94 cm, women ≥80 cm; South Asian, Chinese, Japanese men ≥90 cm, women ≥80 cm; Sub-Saharan African, Eastern Mediterranean, Middle East men ≥94 cm, women ≥90 cm.
• Self-Reported vs. Measured BMI Error: Self-report underestimates BMI by 0.5-1.5 kg/m² on average (source: Connor Gorber et al., Obesity Reviews, 2007). Weight underreported 0.5-2.0 kg, height overreported 0.5-2.0 cm; error larger in women and higher BMI groups.

The BMI section is organised by the question you bring. If you just got a BMI number from a doctor, an app, or a scale and need to know what it means and what it does not mean, go to the guide that interprets a single result. It walks you through the WHO adult BMI classification.

If you have a child's centile from a school screening or a paediatrician, the paediatric BMI guide explains how growth charts work, why a centile is a position on a reference distribution and not a diagnosis, and what the CDC 2000 growth charts versus WHO Child Growth Standards actually measure.

If you are a fitness or health professional who needs to explain thresholds without overclaiming, the guide that names every issuing body and year for cut-offs by population is the one for you. If you have seen BMI criticised and want to know which criticisms are supported by meta-analyses and which are category errors, the evidence guide covers the J-shaped mortality curve, the obesity paradox, and Mendelian randomisation studies. If you are a writer or editor needing the provenance of every threshold, the authority guide lists which body published which cut-off in which year and on what evidence.

Who Should Use This Section

The person who wants to understand measurement and screening without being prescribed a diet or a workout. The parent who needs to interpret a paediatric centile without panic. The professional who needs verifiable thresholds. The data-literate reader who wants the real gaps in the research. This section is not for someone looking for a calorie target, a weight-loss drug decision, or a personal health assessment. Those require a registered dietitian, a prescribing clinician, or a medical consultation. One anthropometric number cannot answer whether you are healthy, and this section explains why.

The BMI formula is mathematically fixed, but the inputs are not. A measured BMI requires standing height without shoes on a stadiometer (error ±0.1-0.5 cm with proper technique) and weight in light indoor clothing on a calibrated scale (error ±0.1-0.5 kg). The WHO STEPS protocol (2008) specifies these conditions. A self-reported BMI, which is what most apps and scales that ask you to type in your height use, systematically underestimates the real number by 0.5-1.5 kg/m² on average. Weight is underreported. Height is overreported. The gap is larger in women and in people with higher actual BMI.

Do not treat self-reported height and weight as equivalent to measured values. For children under two years, recumbent length on a length board and weight on a calibrated infant scale replace standing height, per WHO Child Growth Standards (2006).

Measurement Error Sources You Cannot Ignore

Waist circumference, when measured by a trained observer using the WHO STEPS protocol (midpoint between lower costal margin and iliac crest), has an inter-rater error of 1-3 cm. Self-measured waist circumference, without landmark training, produces errors commonly exceeding 3-5 cm.

Bioelectrical impedance analysis (BIA) body fat percentage estimates carry individual errors of ±3-5 percentage points compared to dual-energy X-ray absorptiometry (DXA), and the error varies with hydration status, recent food intake, and device algorithm. DXA itself has an error of ±1-3% against the 4-compartment model, making it the practical reference method in research, not a routine clinical screening tool. Skinfold measurement with callipers by an untrained operator has an error of ±5-10%, while a trained operator achieves ±3-5%. Every BIA reading taken under uncontrolled conditions is noise that looks like signal. Trust only measured, protocol-adherent numbers when you need to interpret a screening result.

Tool	What It Measures	Population Screening Strength	Individual Misclassification Risk
BMI (kg/m²)	Weight-for-height ratio	High specificity at extreme values; good for tracking shifts in large groups	Moderate sensitivity; misclassifies muscular individuals and older adults with sarcopenic obesity
Body fat percentage (%)	Adipose tissue fraction of total mass	Better correlates with metabolic risk than BMI alone	No universal healthy-range standard across methods or groups; method-dependent thresholds
Waist circumference (cm)	Abdominal girth, surrogate for central adiposity	Adds information beyond BMI; risk thresholds by sex and ethnicity	Protocol-dependent (landmark, tension, clothing); self-measured values unreliable
Waist-to-height ratio (dimensionless)	Central adiposity scaled to stature	Single cut-off (0.5) removes need for sex-specific thresholds; NICE CG189 recommends it	A population heuristic, not a personalised risk score; not validated for all ethnicities at the same cut-off
Visceral fat area (cm², CT/MRI)	Fat inside the abdominal cavity around organs	Strong independent association with metabolic disease	No universal clinical cut-off; thresholds differ by sex and ethnicity; not a routine screening tool

The correlation between BMI and DXA-measured body fat percentage is r=0.6-0.8 in adults (Flegal et al., 2009, meta-analysis of 32 studies, n=44,000+). That is a moderate to strong association at the group level. At the individual level, it leaves room for substantial misclassification.

In NHANES data, a significant fraction of adults with a BMI in the overweight range have normal body fat by DXA, and a significant fraction of adults with a BMI in the normal range have excess body fat, especially at older ages and in certain ethnic groups. The distinction is between being overweight by BMI and being overfat by direct measurement. A person can be both, neither, or one without the other. BMI cannot distinguish fat mass from lean mass; that is not a failure of the tool, it is a description of what the arithmetic does. Muscular individuals can have a BMI ≥30 with a body fat percentage below 25% (men) or 35% (women). Older adults can have normal BMI while carrying excess body fat masked by sarcopenic obesity, a condition that affects an estimated 4-10% of adults over 65, depending on the definition used.

BMI Thresholds By Population: The WHO Asian Cut-Offs

In 2004, the WHO Expert Consultation published lower trigger points for public health action in Asian groups: increased risk at BMI 23.0-27.4, high risk at ≥27.5. The evidence base was a review of cross-sectional and cohort studies showing that type 2 diabetes and cardiovascular disease risk elevates at lower BMI than in European and North American groups. This is not a minor adjustment. A person with a BMI of 26 would be classified as pre-obese by the global WHO cut-off (25-29.9) but as high-risk by the 2004 Asian trigger point. Countries including Singapore, India, and Japan have adopted their own national guidelines that lower the thresholds further. Japan uses a BMI of ≥25 to define obesity, per the Japan Society for the Study of Obesity. The principle is settled even if the exact cut-offs are periodically reviewed. BMI thresholds are not universal; they are specific to the group on which the evidence was collected.

If you need a single additional measurement that adds information beyond BMI, the NICE guideline CG189 (2022) recommends waist-to-height ratio. The threshold is 0.5: keep your waist circumference below half your height. This removes the need for sex-specific cut-offs and adjusts for frame size. The evidence base is a meta-analysis of 31 studies (Ashwell et al., 2012, n=300,000+) showing that waist-to-height ratio predicts type 2 diabetes, hypertension, and dyslipidemia with an AUC of 0.70-0.75, outperforming BMI alone. It is a population heuristic, not a personalised risk score, but it is the most practical single add-on for a screening tool that does not require imaging equipment.

The NIH/NHLBI waist circumference thresholds (1998: >102 cm men, >88 cm women) serve a similar function but were derived from US white groups and do not transfer to other ethnic groups without adjustment. The International Diabetes Federation (2006) uses ethnicity-specific waist circumference thresholds for metabolic syndrome: ≥94 cm men, ≥80 cm women for Europids; ≥90 cm men, ≥80 cm women for South Asian, Chinese, and Japanese groups. The World Health Organization has not issued a single global waist circumference standard. If you measure waist circumference, you must also state whose threshold you are using and the measurement protocol followed. Without that context, the number is meaningless.

What To Do Next

If you got a BMI number and need to interpret it, go to the guide that explains that single result. If you are measuring a child, go to the paediatric BMI guide that walks through growth charts and centiles. If you are comparing methods, go to the measurement guide that covers DXA, BIA, skinfolds, and waist circumference protocols. If you want the evidence behind a threshold, go to the authority guide that names the issuing body, year, and evidence base for every cut-off.

BMI was designed to spot statistical shifts in large groups. Its strength is in epidemiology, not in the clinic. The WHO adult BMI cut-offs were set by an expert committee reviewing observational studies linking BMI to morbidity and mortality, primarily in European and North American groups, in 1995. Those cut-offs were reaffirmed in 2000. They have not been revised since, despite accumulating evidence of ethnic differences, age-related body composition changes, and the systematic bias of self-reported data.

The principle that BMI is a population screening tool, not an individual diagnostic, is embedded in WHO, CDC, and NICE guidance. That principle is also the reason you cannot use a BMI number to decide whether a specific person is healthy. If you want a personal health assessment, you need a medical consultation that includes body composition measurement, metabolic markers, and clinical history. No anthropometric number alone answers that question.