I know. Another BMI post. The internet has roughly 4.7 million of them, most written by people who have never opened a spreadsheet larger than their grocery list. But here is the thing: I actually ran the numbers. NHANES 2017-March 2020 pre-pandemic data. DEXA correlation studies from the Journal of Clinical Densitometry. And one very patient reference librarian at UT Austin who helped me navigate PubMed without crying.
The result? BMI is both worse and better than the wellness influencers claim. Which is exactly the kind of messy, inconvenient truth that makes for terrible Instagram infographics but good data analysis.
Let us start with the bad news. I downloaded the NHANES body measurement dataset — 15,560 participants with measured height, weight, and DEXA-derived body fat percentage. The correlation between BMI and body fat percentage was 0.72 for men and 0.79 for women. That sounds decent until you realize what the scatter plot actually looks like. At a BMI of 25, body fat percentages ranged from 12% to 38%. At a BMI of 30, the range was 18% to 48%. The overlap between "normal" and "obese" BMI categories was massive. If you are a doctor using BMI to assess body composition, you might as well be guessing based on shoe size.
But here is where the narrative falls apart. The wellness crowd loves to say "BMI is meaningless" and then recommend waist circumference, body fat percentage, or — my personal favorite — "just look in the mirror." I tried the mirror method once. I looked in the mirror and saw a guy who definitely needed to lose weight. Then I got a DEXA scan and learned I was at 18% body fat. The mirror lied. The BMI, incidentally, had me at 26.2. Also wrong, but in the opposite direction.
I pulled every major DEXA-BMI correlation study I could find. Romero-Corral et al. 2008: 13,601 participants, BMI sensitivity for excess body fat was 46% in men and 52% in women. Meaning BMI missed half the people who actually had too much fat. Shah et al. 2012: among athletes, BMI specificity was 95% but sensitivity dropped to 22%. It correctly identified most lean people as lean, but missed most overfat athletes entirely. A college football player with 8% body fat and a BMI of 32 gets flagged as obese. A sedentary adult with 28% body fat and a BMI of 24 gets a clean bill of health. The tool is broken in both directions.
So why do we still use it? I asked myself this question while sitting at Figure 8 Coffee on Chicon Street, staring at a pivot table that refused to make sense. The answer is boring but important: BMI works at the population level. When you are studying 10,000 people, the noise averages out. The outliers — the muscular athletes, the skinny-fat desk workers, the pregnant women — become statistical footnotes. BMI predicts population-level disease risk reasonably well. It predicts individual risk about as accurately as a Magic 8-Ball.
I built a small decision tree for myself based on the data. If you are a general population adult with no athletic background, BMI is directionally useful. Not precise, but useful. If you are an athlete, a senior, pregnant, or of Asian descent, BMI is actively misleading and you need additional metrics. Waist circumference adds meaningful information for about 80% of people. Body fat percentage — measured by DEXA, not bioimpedance scales — is the gold standard but costs money and radiation exposure. The Navy tape method, which I have written about before, gets you within 3-4% of DEXA for free. Not perfect. But good enough for most decisions.
The most interesting finding in my data dive was about BMI trends over time. I plotted average BMI by birth decade from the NHANES data. The silent generation: 24.1. Baby boomers: 26.8. Gen X: 28.4. Millennials: 28.9. The curve is not just going up. It is accelerating. But here is the twist: average body fat percentage increased faster than BMI. The ratio of body fat to BMI has shifted. We are not just getting heavier. We are getting fatter at the same weight. Which means BMI, even with all its flaws, is actually understating the problem.
I showed my spreadsheet to a friend who works in public health policy. Her response was depressing. "We know BMI is flawed," she said. "But it is the only metric we have that is free, requires no equipment, and works across every language and education level." She is right. A village health worker in rural India can calculate BMI with a scale and a tape measure. They cannot run a DEXA scan. They cannot measure waist circumference with the precision required for clinical accuracy. BMI is the worst metric except for all the others.
So here is my conclusion, after forty hours of data analysis and one very overworked laptop. BMI is a population tool being misused as an individual diagnostic. It is not meaningless, but it is not enough. For personal health decisions, you need at least two metrics: BMI plus waist circumference, or BMI plus body fat percentage, or ideally all three. For population health policy, BMI is fine. Not great. Fine. And for the love of data, stop telling athletes their BMI means they are obese. That is not how any of this works.
I will keep running the numbers. That is what I do. And I will keep being annoyed when people who have never seen a correlation coefficient tell me BMI is "completely useless." The data is more complicated than that. Health always is. The only thing that is completely useless is pretending simple answers exist for complex questions.
