BMI, honestly: what it measures, what it doesn't

BMI is the most cited and least defended health metric in modern medicine. Here's where it came from, why it's still used despite its known failures, and what to look at instead when BMI doesn't apply to you.

Where it came from (and what its inventor said about using it on individuals)

Adolphe Quetelet, a Belgian astronomer and statistician, devised "the Quetelet index" in 1832. He was studying population averages and looking for a simple measure relating weight to height. He explicitly warned against using it to assess individuals — it was a population-statistics tool.

It became "Body Mass Index" in 1972 when American physiologist Ancel Keys promoted it as a cheap, fast proxy for body fatness, useful for epidemiological studies. The WHO adopted the cutoffs (under 18.5 = underweight, 18.5–24.9 = normal, 25–29.9 = overweight, 30+ = obese) in 1995, and they've barely moved since.

The formula

Imperial:

The 703 is a unit-conversion fudge factor; the underlying math is the same.

Note the squared height. There's no principled physical reason for it — heights cubed (volume) or to the 2.5 power are both better mass-shape proxies for adults. Squared is what Quetelet picked because the data he was working with fit it slightly better than other exponents.

What BMI is actually good for

BMI's defenders are right about two things:

1. It's cheap and reproducible. Two numbers (height, weight), no equipment beyond a scale and tape. For population studies tracking trends over decades, that's invaluable. Insurance actuarial tables, epidemiology, public-health surveillance — all real uses.
2. At population scale, it correlates with bad outcomes. Across millions of adults, BMI > 30 tracks higher rates of type-2 diabetes, cardiovascular disease, and all-cause mortality. The correlation isn't an artifact.

For an individual visiting a doctor, BMI is a starting screening question — a heuristic the doctor uses to decide whether to ask follow-up questions. It's not a diagnosis.

Where BMI famously fails

Athletes. A 6-foot-tall NFL linebacker at 240 lb has a BMI of 32.6 — "obese" by every cutoff. He's also a professional athlete with single-digit body fat. The formula doesn't distinguish muscle from fat. Same problem for bodybuilders, rowers, anyone doing serious resistance training.

Ethnicity. The WHO cutoffs were derived primarily from European populations. South Asian populations have higher body-fat percentages at the same BMI — a "normal" BMI 23 person of Indian descent may carry the visceral fat of a "overweight" BMI 27 person of European descent. The WHO now recommends lower BMI thresholds (23, 27.5) for Asian populations, though uptake is uneven.

Children and adolescents. BMI in growing bodies is meaningless without age and sex adjustments. Pediatricians use BMI percentile charts (by age, by sex) instead of the raw number. The same BMI 20 is normal for a 16-year-old boy and severely overweight for a 10-year-old.

The elderly. Old people lose muscle mass faster than fat mass (sarcopenia). An older adult's "normal" BMI may mask undernutrition. The cutoffs may need to shift upward by 2–3 points for healthy 70+ year olds.

Pregnancy. Obvious. Don't use BMI on pregnant people.

Better metrics, in roughly this order

• Waist-to-height ratio. Waist measurement / height. Should be < 0.5. Simple, two numbers, much better at identifying visceral fat. Doesn't penalize muscle.
• Waist circumference. Above 40" for men or 35" for women correlates with metabolic syndrome risk independent of BMI.
• Body fat percentage. Direct measurement via DXA scan (gold standard), bioelectrical impedance (cheap, less accurate), or skin-fold calipers (operator-dependent). Our body-fat calculator uses the US Navy circumference method — surprisingly accurate for casual use.
• Lab markers. If you've got access to bloodwork: fasting glucose, HbA1c, lipid panel, hsCRP. These tell you metabolic health directly, no estimation required.

The honest summary

BMI is a fine screening number for "should I ask follow-up questions?" It's a terrible label to internalize. If you're an athlete, the formula doesn't apply. If your ethnicity, age, or pregnancy status puts you outside the populations the cutoffs were derived from, the formula doesn't apply. If you're "normal" but eating a diet that produces metabolic dysfunction, the formula is missing the signal.

Use it as one input among several — not as a verdict.