BMI is flawed because it reduces a complex picture of health to a single number based on height and weight, ignoring body composition, fat distribution, age, sex, and ethnicity. It was never designed to assess individual health, and a growing body of evidence shows it routinely misclassifies people in both directions: labeling fit, muscular people as overweight while giving a clean bill of health to people carrying dangerous amounts of body fat.
In 2023, the American Medical Association formally acknowledged these problems, stating that BMI “loses predictability when applied on the individual level” and should not be used as a sole measure in clinical decisions. Here’s a closer look at where BMI falls short and why it matters for you.
It Was Built for Populations, Not People
BMI traces back to 1832, when Adolphe Quetelet, a Belgian statistician and astronomer, divided weight by height squared to define the characteristics of the “normal man.” Quetelet wasn’t a physician. He was looking for statistical averages across large groups, not trying to diagnose disease in individuals. The formula was designed to describe a population-level pattern, and it was based on data from Western European men.
Nearly two centuries later, that same formula is used in doctor’s offices, insurance assessments, and military fitness screenings worldwide. The WHO’s standard obesity classes (class I starting at a BMI of 30, class III at 40 and above) were built primarily on evidence from studies of white populations. Applying a tool meant for 19th-century population statistics to a diverse, modern individual is where the trouble starts.
It Can’t Tell Muscle From Fat
BMI treats every kilogram the same, whether it’s muscle or fat. A 200-pound person who is 5’10” gets the same BMI regardless of whether that weight comes from years of strength training or years of inactivity. This creates a well-documented problem for athletes, military personnel, and anyone with above-average muscle mass.
A study of U.S. Army soldiers illustrates this clearly. Soldiers classified as “obese” by BMI did carry more body fat than their normal-weight peers, but they also carried an average of 12.1 kg (about 27 pounds) more lean muscle mass. In the overweight category, soldiers had roughly 5.5 kg more muscle than normal-weight soldiers. The additional muscle was enough to push them past BMI thresholds even when their body fat wasn’t dramatically elevated. Research on college athletes found similar problems, concluding that BMI’s standard cutoffs significantly overestimate body fat in active populations and that different classification systems are needed.
These muscular individuals are essentially treated as statistical noise in BMI’s framework. The formula simply has no mechanism to distinguish between a bodybuilder and someone who is sedentary and overfat.
Normal BMI Doesn’t Mean Normal Body Fat
The opposite problem is just as serious. A large study using body fat measurements found that 33% of men and 52% of women with a “normal” BMI (20 to 25) actually had excess body fat by clinical standards. That means roughly one in three normal-weight men and one in two normal-weight women were carrying enough fat to qualify as having obesity, yet their BMI gave no warning.
This phenomenon, sometimes called “normal weight obesity,” is particularly dangerous because these individuals are unlikely to receive screening or counseling for metabolic conditions. Their BMI looks fine on paper, so the conversation about cardiovascular risk or blood sugar often never happens.
It Ignores Where Fat Lives in Your Body
Not all body fat carries the same health risk. Visceral fat, the type that wraps around organs deep in the abdomen, drains directly into the liver through the portal circulation and is the primary driver of insulin resistance. It’s closely linked to type 2 diabetes, chronic inflammation, and higher mortality. Subcutaneous fat, the layer just beneath the skin, is far less metabolically harmful and may even improve insulin sensitivity.
BMI cannot distinguish between these two types of fat. Two people with identical BMIs could have radically different health profiles: one storing fat primarily under the skin on their hips and thighs, the other packed with visceral fat around their liver and intestines. A waist circumference measurement captures this distinction. BMI does not.
It Doesn’t Work the Same Across Ethnicities
People of different ethnic backgrounds develop metabolic disease at different BMI levels. Asian and South Asian populations, for instance, tend to accumulate more visceral fat at lower body weights. Recognizing this, both the WHO and the UK’s National Institute for Health and Care Excellence recommend lower BMI thresholds for Asian populations: 27.5 for class I obesity instead of 30, and 37.5 for class III obesity instead of 40.
The AMA’s 2023 policy specifically called out BMI’s reliance on data from non-Hispanic white populations and acknowledged historical harm from applying those standards universally. For people of Asian, Black, or Hispanic descent, the standard cutoffs can either underestimate or overestimate risk depending on the population and the specific health outcome in question.
It Gets Less Accurate as You Age
After about age 50, most people begin losing height due to narrowing of the spaces between vertebrae, compression of spinal discs, and changes in posture. Women lose an average of 5 to 7 cm (2 to 3 inches) between ages 40 and 80. Men lose 3.6 to 6 cm. Since BMI divides weight by height squared, even a modest loss in height inflates the number. By age 70, height loss alone can bump BMI up by about 1 point, enough to push someone from “normal” into “overweight” without any actual change in body composition.
The misclassification is substantial. At age 70, roughly 20% of women and 15% of men are incorrectly categorized as overweight based on their current measured height. By age 80, about 30% of those labeled obese by BMI are misclassified when their original adult height is used for the calculation. On top of this, older adults lose muscle mass with age (a process called sarcopenia), meaning they can have a “normal” BMI while carrying a higher proportion of body fat than BMI suggests.
Better Alternatives Exist
Several measurements do a better job of estimating body fat and health risk. Waist circumference is the simplest: it takes seconds, requires only a tape measure, and captures visceral fat accumulation that BMI misses entirely. Waist-to-hip ratio is another option, though a large meta-analysis of over 82,000 participants found that its advantage over BMI in predicting cardiovascular death was marginal, less than 1% improvement in discrimination.
A more promising tool is the Relative Fat Mass index, which uses only height and waist circumference. In validation studies, RFM explained 84% of the variation in body fat percentage measured by DXA scans (the gold standard for body composition). BMI, by comparison, explained just 36%. That’s a dramatic difference in accuracy from a formula that’s nearly as simple to calculate.
Body fat percentage itself, measured through methods like DXA scans or air displacement plethysmography, provides the most direct answer. These are more expensive and less accessible than stepping on a scale, but they measure what BMI only guesses at.
What BMI Still Does Well
For all its flaws, BMI remains useful as a quick screening tool across large populations. It’s free, fast, and requires no equipment beyond a scale and a ruler. At the population level, it correlates significantly with body fat. The problem isn’t that BMI exists. It’s that it’s treated as a diagnostic tool for individuals when it was only ever meant to identify broad trends.
If your BMI puts you in the overweight or obese category, it’s worth investigating further with waist circumference, blood pressure, blood sugar, and cholesterol rather than accepting the label at face value. And if your BMI is normal but you carry weight around your midsection, that number may be giving you false reassurance. The single most useful thing you can do is stop treating BMI as a verdict and start treating it as one data point among many.