Is BMI the Best We Can Do?

Why the measure may be better than we thought

Published on December 26, 2012 by Steve Mooney

Any measure that says that Tom Cruise and Philip Seymour Hoffman share a similar body habitus has to have problems—that’s why body mass index (BMI—measured as the mass in kilograms divided by height in meteres squared) the most common measure of body habitus used by doctors and public health practitioners, has been receiving a lot of criticism lately.

So you might expect doctors to switch to newer measures that separate fat mass from lean mass or that takes body shape into account any day now. Or at least you might expect the World Health Organization to pick a better measure on which to define obesity.  Certainly, there are many others which doctors and patients could choose from.

But like the good guy in an action movie or the bad guy in a horror movie, BMI refuses to die.

Or maybe BMI isn’t so bad a measure after all? At least that’s what a recent study we published suggests. Across a sample of over 10,000 employed American adults, we found that BMI was about as good at predicting hypertension, dyslipidemia, and other components of the metabolic syndrome—a precursor to cardiovascular disease and diabetes—as any of five proposed alternative measures to which we compared it head-to-head in the same population (fat mass index, waist circumference, waist-to-height ratio, percent body fat, and fat-free mass index).

One reason why BMI seems to perform as well as other measures may be that most American adults are shaped more like Philip Seymour Hoffman than Tom Cruise. In this study, waist circumference explained more than 90 percent of the variation in BMI, suggesting that high BMI was typically due to excess fat rather than excess muscle.

Also, BMI is easy to measure consistently. Scales that differentiate fat tissue from muscle tissue typically use bio-impedance, which measures the opposition to electrical current through the body, giving an estimation of total body water, which in turn can be used to estimate fat-free mass. But as you might expect, these measurements are prone to variation on a host of extraneous factors, like how much water the subject drank recently. And comparing waists isn’t as easy as you might think—it requires finding the equivalent point on every waist.

So for all of BMI’s flaws, it might just be good enough.

But that may not be the case forever. For example, improvements in fat-measuring technology may make Percent Body Fat easier to measure precisely, and therefore a better clinical discriminator in the future. But even as other measures might improve, BMI will never be better than it is today.

Also, components of the metabolic syndrome are only a problem if they lead to premature death or disability. If BMI predicts metabolic syndrome components just fine, but fails to predict obesity-related mortality as well as another measure because, say, the cases metabolic syndrome that BMI catches aren’t as severe, then we may need to pull that fancy scale back out of the trash.

Finally, for any given individual, how consistently other measures perform is less of an issue. For example, I can pick a point to measure my waist, and then measure that same point every day, which will give me a reliable estimate of my changes over time. Similarly, if I always check my percent body fat before I eat breakfast, I can assume my hydration level will be pretty consistent. That means that for individuals, BMI may not be as good as other measures, even though it is among the entire population.

In the end, nobody is going to claim BMI is perfect, yet it remains a valid and reliable indicator of health risks and is a measure that should be taken seriously.

Edited by Abdul El-Sayed.

Steve Mooney
Steve Mooney is a doctoral student in epidemiology at Columbia University. His research interests include cancer prevention and the appropriate use of machine learning techniques in epidemiology.

You might also like


The BMI reliably and demonstrably generates both false positives (i.e. – fit folk have a higher BMI) and, much more critically, false negatives (i.e. – unfit folk who have small bones and/or little muscle but are clinically ‘overfat’ may well have a BMI within accepted ‘norms’).

Yes, the BMI correctly identifies overweight fat folk — but all you need to do that is a mirror. It’s correlative measurement masquerading as predictive science.

It is demonstrably NOT “a valid and reliable indicator of health risk”, nor should it be taken seriously.

The BMI is much worse for most people than this article implies. Today I found a free site that gives better estimates of how much I should weigh – I thought it was very smart.

Leave a Reply

Your email address will not be published. Required fields are marked *