Why BMI Lies, and Lies Differently Depending on Your Ancestry

Two scientists once put themselves through a DXA body scanner to settle a point. Chittaranjan Yajnik, an Indian diabetologist, and John Yudkin, a British one, had near-identical body mass indexes. On paper they were the same size of person. The scans said otherwise: Yajnik carried more than double the body fat of his European colleague at the same BMI. They wrote it up as a 2004 Lancet letter that became a small classic, mostly because the picture is so hard to argue with. Same number, wildly different body.

That's the whole problem with BMI in one image. It's a ratio of weight to height squared, invented in the 1830s and validated mostly on European bodies, and somewhere along the way it got promoted from a cheap population statistic to a personal verdict on your health. It was never built for that job. And the ways it fails aren't random. They track, quite specifically, with ancestry.

What BMI Actually Measures, and What It Doesn't

BMI knows two things about you: how much you weigh and how tall you are. That's it. It can't tell muscle from fat, and it has no idea where the fat sits. A lean, heavily built rugby player and a soft, sedentary office worker can post the identical BMI while living in completely different bodies. This is old news for athletes. What's less discussed is that the same blind spot opens up along population lines.

The clearest evidence comes from a 1998 meta-analysis by Deurenberg and colleagues, which pooled body-fat measurements across American Blacks, Chinese, Ethiopians, Indonesians, Polynesians, Thais, and Caucasians. For the same amount of body fat, at the same age and sex, BMI landed lower in nearly every non-Caucasian group. The offsets weren't trivial: about 1.9 BMI units lower in Chinese subjects, 2.9 in Thais, 3.2 in Indonesians, and 4.6 in Ethiopians, all relative to Caucasians. Run that logic backward and it means these groups hit a given fat level well before BMI flags them as overweight.

A later 2002 review by the same group sharpened the Asian picture. At any given BMI, the Asian populations they studied carried roughly 3 to 5 percentage points more body fat than Caucasians. Put differently, for the same body-fat percentage, their BMI ran about 3 to 4 units lower. A BMI of 23 in one body can mean what a BMI of 27 means in another.

Why? Not race in any essentialist sense. The differences trace to body build and development: limb-to-trunk proportions, how much muscle a population tends to carry, and early-life nutrition that shapes where fat gets stored decades later. These are averages across groups with enormous individual overlap, not fixed properties stamped on anyone at birth. Plenty of individuals sit nowhere near their population's average. None of this makes the number destiny. A single universal cutoff just quietly assumes everyone converts weight into fat the same way, and they don't.

The Asian Indian Case, and Why Institutions Moved

South Asians are the sharpest example, which is why Yajnik used himself as the demonstration. The pattern has a name in the literature, sometimes the "thin-fat" phenotype: comparatively normal weight, but with more total body fat and, more importantly, more of it packed into the abdomen and around the organs. A 2009 consensus statement led by Misra describes exactly this, excess body fat with increased abdominal and intra-abdominal adiposity showing up at BMI levels that look reassuring by Western standards. Visceral fat is the metabolically nasty kind, the sort that drives insulin resistance and cardiovascular risk, and BMI is completely blind to it.

This isn't a fringe observation anymore. In 2004 the WHO expert consultation, published in the Lancet, looked at exactly this evidence and found that substantial risk of type 2 diabetes and cardiovascular disease was appearing in Asian populations below the standard overweight threshold of 25. The observed-risk cutoff ranged from about 22 to 25 across different Asian groups. Rather than throw out the international 25/30 system, the WHO recommended adding lower public-health action points, at BMI 23.0 and 27.5, so that screening in these populations catches risk the standard bands miss. It's a rare and honest institutional admission that a number treated as universal isn't.

The same theme runs through MacroMentor's piece on ethnicity and BMR: population averages are a useful nudge, but the real driver is body composition, and once you know that directly, the ancestry proxy has less work to do.

The Better Tool Is a Tape Measure

So if BMI is a shaky personal verdict, what's the upgrade? Reach for something that actually cares where your weight sits. The strongest candidate is waist-to-height ratio, and the evidence for it is genuinely good. A 2012 systematic review by Ashwell and colleagues, pooling more than 300,000 adults across several ethnic groups, found waist-to-height ratio improved the discrimination of cardiometabolic risk by about 4 to 5 percent over BMI, and beat plain waist circumference too, across diabetes, hypertension, and cardiovascular disease in both sexes.

The rule of thumb is almost insultingly simple: keep your waist under half your height. One boundary, roughly 0.5, that works across populations without a lookup table, precisely because it measures the thing BMI can't. MacroMentor leans on this exact ratio for the same reason, and its waist-to-height piece digs into where the 0.5 line comes from.

None of this means BMI is useless. As a cheap way to screen a whole population, it's fine. As a statement about you, it's a flag, not a diagnosis, and one that reads differently depending on whose body it's measuring. Pair it with a tape measure and you get a far more honest picture. Run your own numbers, waist included, over at /calculator.