I'm really not a fan of BMI being the determining factor in obesity. At 27 BMI, I'm apparently 3 BMI away from being considered obese. As a naturally large, fairly muscular 24 year old german/irish man, I'm 6 foot 1 and weigh 205 lbs, but you can see my abs and I'm in relatively good physical condition. Is my definition of obese wrong? I always thought obese meant someone with at least a few dozen lbs of excess fat.
Invent a way of easily measuring body fat percentage, and we can do away with BMI. Until then, muscular people can simply ignore these things. BMI works well enough for the rest of the population (the vast majority of the world). If anything, normal people have the opposite problem you do. BMI tends to underestimate obesity in the general population. http://www.cnn.com/2012/04/03/health/obesity-rates-maybe-wor...
No, there aren't. BMI can be calculated by anyone in about 2 seconds without taking any measurements they don't already know. Body fat percentage can't.
And while BMI may be useless to you, it's reasonably useful for the vast majority of people in the world.
Unfortunately, they are. However, I don't think most insurance companies put you into a lower category based on BMI alone until you hit 27 or so. It's still possible to be at a healthy body fat percentage and have a BMI of 27 or higher, but obviously it's not as likely as it is at the cutoff of 25.
Overall, I think BMI is probably a pretty good overall metric, just not a good diagnosis for an individual. For instance, a physician might notice that a patient has a BMI above 25, but then see that waist to height ratio or a bf measurement looks very favorable.
From a broad numbers perspective, I think BMI is probably very useful - for instance, if you notice that a higher percentage of a population has a BMI in excess of 27 than in previous years, you can safely estimate that the population is becoming increasingly overweight. I suspect that the situation where people have high bmi's with low body fat is limited to a small number of athletes who are not typical of the population at large.
There are tons of things which can be calculated easily and simply, which are simply wrong or utterly useless. Palmistry, astrological signs, numerology, stock charting, biorhythms, dowsing, etc., etc.
Simply because a measurement is fast, simple, and gives some arbitrary numerical precision doesn't make it useful. It's also got to, you know, like, correspond to the stuff it's supposed to be measuring. Which it doesn't.
Official discussion of BMI actually admits this.
The panel did not search for evidence to prove that
measures of body composition or metabolic rate are
useful tools.
The absence of evidence is not the evidence of absence. Especially when the absence of evidence is simply a report that is not looking for what you're looking for.
Either way, BMI clearly has some use. Or do you think this chart is completely random and devoid of correlation? http://upload.wikimedia.org/wikipedia/commons/b/b7/Correlati... Do you think the ordering of countries on the BBC link is purely random and has nothing to do with how many fat people are in each country?
That chart is largely irrelevant to the discussion in several ways. Your question is specious.
It shows population correlations. In a general population, where a range of bodytypes may be assumed, there will be a correlation between BMI and body fat percentage. There's relatively little argument over that, though as I mentioned above, the actual utility of BMI for epidemiological / population studies is being question.
- The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". For which you'd want to look at the r^2, RSS (residual sum of squares), and perhaps plotted residuals to determine the appropriateness of the model, in comparison with other estimators: waist circumference, height/waist ratio (most people also have a pretty good idea of their waist size), circumference and caliper estimators, etc.
- To paraphrase Donald Rumsfeld, in epidemiological population studies, you conduct research with the data you have, not with the data you want. Height and weight are widely collected and reported, even from sources such as DMV files (though they may also be wildly inaccurate). For large-scale or national studies, this makes for large datasets. The relevance and usefulness of the data may not be as high as, say, DEXA analysis, or even waist to height ratio (http://www.webmd.com/heart-disease/news/20050606/waist-heigh...), but it's there. And for at least some purposes, useful.
Extending the use to individuals, where in most settings (home, gym, doctor's office, clinic, health fair), availability of necessary measurement tools (tape measure, calipers, even the wildly inaccurate handheld electronic bioimpedance monitors) is readily facilitated, there's really no call to rely on BMI.
- The problem with BMI is that it's essentially saying "if you fit our model of body fat relative to height and weight, then here is where you fall on our model of obesity". Again, for a general population this relationship holds reasonably well, but for an individual, where the question specifically is "what is my body fat percentage", BMI is the wrong tool for the job. It fails notably for specific populations of interest -- it will consistently underreport obesity in less-healthy populations, and overreports obesity in more-healthy populations. It can also result in nonsensical intervention recommendations: most unfit people will benefit by efforts to increase lean muscle mass, despite the fact that for those with lower BMIs this will decrease their putative health score by increasing overall body mass. I suspect this is among the reasons strength training is under-recommended for increasing fitness (most literature addresses "diet" and "exercise" generally, but makes little or no distinction between cardio and strength training, or focuses virtually exclusively on cardio).
> The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?".
As I've been repeatedly saying, the other estimators of body fat are much more difficult to calculate. Most people already know their height and weight. Most people don't know their waist circumference and have absolutely no idea how to calculate BF%. That's why sites like this use BMI, so that their visitors can gain some information rather than no information.
I'll simply note a very strong disagreement that people don't know, or can't readily obtain, their waist measurement. Particularly in a setting at which they'd be particularly interested in some measure of general fitness: doctor's office, clinic, gym, health fair, etc. Which means that your statement that other measures are harder to obtain or less accurate is simply false.
Among the problems of BMI is that it is official health policy. It's the definition of obesity according to the US government. It's the basis of insurance coverage / premium assessments and other abuses. Changing policy would mean that doctors' offices would measure your height and waist, to give a far more accurate fitness assessment. Over time, epidemiologists would have this dataset generally available.
A simple waist circumference measurement is more accurate than BMI at making individual assessments of obesity (> 40" ~ 100cm is bad).
A set of 3-4 tape measurements (height, neck, abdomen for men; height, neck, waist, hips for women) will give a reasonable estimate of body fat. Equipment cost, about $5.
http://www.linear-software.com/online.html
Your observation of the weakness of BMI is that it tends to misreport true body fat percentage both ways. I'd hardly call that an advantage of BMI.
Note that since official US government policy defines obesity in terms of BMI, technically, BMI doesn't misreport obesity. What it does is fail to correspond to true body fat percentage in any meaningful way.
How are you positing people measure their weight and height without, you know, like, a scale, and, um, a tape measure?
As many people in the fitness/bodybuilding field will tell you, eyeball estimates of bodyfat are in fact pretty bloody accurate. If you're interested in equipment costs, I could look up eyeballs on Amazon as I did for tape measures.
And again, I agree that there are more accurate measurements, but they are all significantly more complicated. This time you are suggesting training people to visually estimate body fat percentage. That could work. But it's a heck of a lot more complicated than telling someone to type their height and weight into a website.
This is true, but it's impact is overestimate. I used to hover near the 25 level for my BMI. I was muscular, so I thought it was silly.
But I've not lost 6 pounds while gaining strength. I can see my abs. Turns out I was 12 pounds over my high school weight.
I'd now need ~8 pounds of extra muscle to become "overweight". That's a lot.
I think it's hard to go over BMI based on muscle alone. You may have less of a problem is you're BMI 25 and muscular, but you still probably could stand to lose some fat.
Well, I have a scale at home that does Body Fat %, and it is reasonably accurate. (It is consistent day to day, compared to a doctor's office one its off by maybe 1%).
I believe it was $50. There are several on Amazon for that price and lower. If a regular scale is $20, and one that measures body fat is $40, then I don't think it's a big deal. I personally watch the body fat % now rather than weight.
Before complaining about "wah wah that's BMI, BMI sucks" please think twice. Everybody knows it's BMI, BMI is a rough measurement of your ratio height to weight.
"The medical establishment has acknowledged major shortcomings of BMI.[...] found that BMI-defined obesity was present in 21% of men and 31% of women. Using body fat percentages (BF%), however, BF%-defined obesity was found in 50% of men and 62% of women."
Take it as a statistical game and stop getting self conscious about it.
Note also: If you are very muscular with a crazy BMI despite no fat it's not healthy either.
No "everyone" doesn't know that BMI is discredited, the BBC should have put a huge disclaimer on it saying "these numbers are nothing to do with how healthy you are, in fact they are completely meaningless".
I like how I got downvoted. Please read everything that is on the website:
The BBC say it's an estimate, with a link on estimate leading to its definition and a disclaimer about BMI.
They never say that you are fatter or slimmer than the rest of the population but they talk about your BMI and your BMI only, just a quote:
"You have a lower BMI than 88% of males aged 15-29 in your country"
"You have a lower BMI than 54% of males aged 15-29 in the world"
I really don't see anything about weight or any scheme to make you think you're fat if you have a high BMI from the BBC like some imply in the comments. It's very funny how hyper defensive everybody gets around the BMI thing, especially the ones who don't have the "right" BMI.
Note also: If you are very muscular with a crazy BMI despite no fat it's not healthy either.
I hear this claim parroted quite a bit. I haven't seen any evidence of this in medical literature, though, except in cases of really big guys like NFL lineman. Do you have any references for the claim?
BMI is an utterly bogus means of determining individual fitness assessments, and there's some reason to believe it's not very useful for population measures of fitness.
Dude, thanks for linking to Dr. Keith Devlin's column. Probably the only thing I read religiously in the MAA.
BMI is just one of several mistakes attributed to the great mathematician Quetelet. You can credibly implicate Quetelet for a variety of modern day statistical crimes such as profiling ( attempting to figure out if a person is a criminal based on statistical analysis of race/sex/height/etc...),"normal" age of marriage, statistical rates of divorce in different countries and their implications on fidelity, and such other sociological nonsense.
Dr. Devlin labels these things "junk math, numerological nonsense". But you can take a more charitable view - given that Quetelet practically invented both statistics & sociology ( which he called "physics of society" ), he came up with a number of formulae that seemed true at that time ( mid 1800s ). On hindsight much of this stuff is just silly mathematical manipulation ( Instead of dividing weight by square of height & multiplying by the constant 703, you can square the weight & divide by the cube root of height and put in some other constant...)
If you are interested in the kind of statistics practiced in Quetelet's timeframe, here's some wacky references.
btw, much of what we do today in quant finance is suspiciously similar to what Quetelet suggests :) We build statistical "factor models" that credibly relate earnings of a company in california to the price of chicken in China :) We take a rigged number called the LIBOR & price interest rate derivatives on the spead between this rigged fixed rate & the floating rate. Then the public loses money & we scream - hey who the fuck rigged the LIBOR in the first place ?:) Talk about bootstrapping... honestly, we are no different from Dr. Quetelet. Might as well base Google's quarterly earnings on Page's BMI.
Quetelet's work was reasonably appropriate for his time. One source I recall (cannot find it now) stated that he was looking for a rapid paper-and-pencil method for estimating aggregate troop strength. As a basic, good-enough first approximation, BMI works reasonably well. The problem is that it's now been grossly misapplied.
Quants (and economists) spend, I suspect, a lot of time chasing what turn out to be transitory relationships between various measurements. The relationships may very well be statistically valid, when first measured, but either changing real-world conditions, or, more likely in the financial world, adoption of the relationships by others seeking profits, rapidly erodes their validity.
Thanks for the links, I've got some reading for later :)
An enterprise I had some association with in the past spent some years and considerable capital trying to establish a strongly predictive relationship between a physical property and a health measurement. It turns out that there was a relationship, but not one which could be reliably used to determine an accurate quantification of the health measurement. This didn't keep them from torturing the data extensively.
By moving to a city with a lot of S.E. Asian people I'm suddenly above average BMI - I need to move back to Texas.
On the other hand, all those little old Japanese women mean that the average life expectancy here is among the highest in the world - so I should live to a 100 like them
I was really interested to see this data until I saw that they use the BMI; possibly the dumbest and most loudly touted measure of health. According to these numbers I am overweight, heading towards obese, and I don't think any doctor on Earth would say I was overweight if they looked at me.
40 comments
[ 3.5 ms ] story [ 38.4 ms ] threadI weight lift... a lot. So BMI is utterly useless for me (as you said)
And while BMI may be useless to you, it's reasonably useful for the vast majority of people in the world.
Overall, I think BMI is probably a pretty good overall metric, just not a good diagnosis for an individual. For instance, a physician might notice that a patient has a BMI above 25, but then see that waist to height ratio or a bf measurement looks very favorable.
From a broad numbers perspective, I think BMI is probably very useful - for instance, if you notice that a higher percentage of a population has a BMI in excess of 27 than in previous years, you can safely estimate that the population is becoming increasingly overweight. I suspect that the situation where people have high bmi's with low body fat is limited to a small number of athletes who are not typical of the population at large.
Simply because a measurement is fast, simple, and gives some arbitrary numerical precision doesn't make it useful. It's also got to, you know, like, correspond to the stuff it's supposed to be measuring. Which it doesn't.
Official discussion of BMI actually admits this.
http://www.nhlbi.nih.gov/guidelines/obesity/ob_gdlns.htmEither way, BMI clearly has some use. Or do you think this chart is completely random and devoid of correlation? http://upload.wikimedia.org/wikipedia/commons/b/b7/Correlati... Do you think the ordering of countries on the BBC link is purely random and has nothing to do with how many fat people are in each country?
It shows population correlations. In a general population, where a range of bodytypes may be assumed, there will be a correlation between BMI and body fat percentage. There's relatively little argument over that, though as I mentioned above, the actual utility of BMI for epidemiological / population studies is being question.
- The question isn't "is there a correlation?", but "is the correlation between BMI and body fat stronger or weaker than for other estimators of body fat?". For which you'd want to look at the r^2, RSS (residual sum of squares), and perhaps plotted residuals to determine the appropriateness of the model, in comparison with other estimators: waist circumference, height/waist ratio (most people also have a pretty good idea of their waist size), circumference and caliper estimators, etc.
- To paraphrase Donald Rumsfeld, in epidemiological population studies, you conduct research with the data you have, not with the data you want. Height and weight are widely collected and reported, even from sources such as DMV files (though they may also be wildly inaccurate). For large-scale or national studies, this makes for large datasets. The relevance and usefulness of the data may not be as high as, say, DEXA analysis, or even waist to height ratio (http://www.webmd.com/heart-disease/news/20050606/waist-heigh...), but it's there. And for at least some purposes, useful.
Extending the use to individuals, where in most settings (home, gym, doctor's office, clinic, health fair), availability of necessary measurement tools (tape measure, calipers, even the wildly inaccurate handheld electronic bioimpedance monitors) is readily facilitated, there's really no call to rely on BMI.
- The problem with BMI is that it's essentially saying "if you fit our model of body fat relative to height and weight, then here is where you fall on our model of obesity". Again, for a general population this relationship holds reasonably well, but for an individual, where the question specifically is "what is my body fat percentage", BMI is the wrong tool for the job. It fails notably for specific populations of interest -- it will consistently underreport obesity in less-healthy populations, and overreports obesity in more-healthy populations. It can also result in nonsensical intervention recommendations: most unfit people will benefit by efforts to increase lean muscle mass, despite the fact that for those with lower BMIs this will decrease their putative health score by increasing overall body mass. I suspect this is among the reasons strength training is under-recommended for increasing fitness (most literature addresses "diet" and "exercise" generally, but makes little or no distinction between cardio and strength training, or focuses virtually exclusively on cardio).
As I've been repeatedly saying, the other estimators of body fat are much more difficult to calculate. Most people already know their height and weight. Most people don't know their waist circumference and have absolutely no idea how to calculate BF%. That's why sites like this use BMI, so that their visitors can gain some information rather than no information.
Among the problems of BMI is that it is official health policy. It's the definition of obesity according to the US government. It's the basis of insurance coverage / premium assessments and other abuses. Changing policy would mean that doctors' offices would measure your height and waist, to give a far more accurate fitness assessment. Over time, epidemiologists would have this dataset generally available.
Really. It's long past time to retire BMI.
Why are you so wedded to it?
A set of 3-4 tape measurements (height, neck, abdomen for men; height, neck, waist, hips for women) will give a reasonable estimate of body fat. Equipment cost, about $5. http://www.linear-software.com/online.html
Your observation of the weakness of BMI is that it tends to misreport true body fat percentage both ways. I'd hardly call that an advantage of BMI.
Note that since official US government policy defines obesity in terms of BMI, technically, BMI doesn't misreport obesity. What it does is fail to correspond to true body fat percentage in any meaningful way.
Your point is that a better measurement can be made if everyone goes out and buys some new thing and takes some new measurement.
That seems to be in agreement with my point.
> since official US government policy defines obesity in terms of BMI, technically, BMI doesn't misreport obesity
The US government defining something doesn't make it true.
>What it does is fail to correspond to true body fat percentage in any meaningful way.
It is very meaningful. It's much better than a random guess.
As many people in the fitness/bodybuilding field will tell you, eyeball estimates of bodyfat are in fact pretty bloody accurate. If you're interested in equipment costs, I could look up eyeballs on Amazon as I did for tape measures.
And again, I agree that there are more accurate measurements, but they are all significantly more complicated. This time you are suggesting training people to visually estimate body fat percentage. That could work. But it's a heck of a lot more complicated than telling someone to type their height and weight into a website.
But I've not lost 6 pounds while gaining strength. I can see my abs. Turns out I was 12 pounds over my high school weight.
I'd now need ~8 pounds of extra muscle to become "overweight". That's a lot.
I think it's hard to go over BMI based on muscle alone. You may have less of a problem is you're BMI 25 and muscular, but you still probably could stand to lose some fat.
I believe it was $50. There are several on Amazon for that price and lower. If a regular scale is $20, and one that measures body fat is $40, then I don't think it's a big deal. I personally watch the body fat % now rather than weight.
I'm a Marathon runner who is apparently fatter than 99% of the world...
This is going to tell me I'm fat.
I hovered around 25 BMI, was muscular, thought it was silly. Everyone considered me "lean". But, no abs.
Now I can see mine, am very strong, and have a 23 BMI.
It's possible BMI effectiveness changes with height though, which would make it even worse. I'm 5'6".
http://en.wikipedia.org/wiki/Body_mass_index#Limitations_and...
"The medical establishment has acknowledged major shortcomings of BMI.[...] found that BMI-defined obesity was present in 21% of men and 31% of women. Using body fat percentages (BF%), however, BF%-defined obesity was found in 50% of men and 62% of women."
Take it as a statistical game and stop getting self conscious about it.
Note also: If you are very muscular with a crazy BMI despite no fat it's not healthy either.
The BBC say it's an estimate, with a link on estimate leading to its definition and a disclaimer about BMI.
They never say that you are fatter or slimmer than the rest of the population but they talk about your BMI and your BMI only, just a quote:
"You have a lower BMI than 88% of males aged 15-29 in your country"
"You have a lower BMI than 54% of males aged 15-29 in the world"
I really don't see anything about weight or any scheme to make you think you're fat if you have a high BMI from the BBC like some imply in the comments. It's very funny how hyper defensive everybody gets around the BMI thing, especially the ones who don't have the "right" BMI.
I hear this claim parroted quite a bit. I haven't seen any evidence of this in medical literature, though, except in cases of really big guys like NFL lineman. Do you have any references for the claim?
- Top 10 Reasons Why The BMI Is Bogus : http://www.npr.org/templates/story/story.php?storyId=1062684...
- Beyond BMI, Why doctors won't stop using an outdated measure for obesity : http://www.slate.com/id/2223095/
- Do You Believe in Fairies, Unicorns, or the BMI? : http://www.maa.org/devlin/devlin_05_09.html
BMI is just one of several mistakes attributed to the great mathematician Quetelet. You can credibly implicate Quetelet for a variety of modern day statistical crimes such as profiling ( attempting to figure out if a person is a criminal based on statistical analysis of race/sex/height/etc...),"normal" age of marriage, statistical rates of divorce in different countries and their implications on fidelity, and such other sociological nonsense.
Dr. Devlin labels these things "junk math, numerological nonsense". But you can take a more charitable view - given that Quetelet practically invented both statistics & sociology ( which he called "physics of society" ), he came up with a number of formulae that seemed true at that time ( mid 1800s ). On hindsight much of this stuff is just silly mathematical manipulation ( Instead of dividing weight by square of height & multiplying by the constant 703, you can square the weight & divide by the cube root of height and put in some other constant...) If you are interested in the kind of statistics practiced in Quetelet's timeframe, here's some wacky references.
http://en.wikisource.org/wiki/Popular_Science_Monthly/Volume...
http://statistiks.net/wp-content/plugins/downloads-manager/u...
btw, much of what we do today in quant finance is suspiciously similar to what Quetelet suggests :) We build statistical "factor models" that credibly relate earnings of a company in california to the price of chicken in China :) We take a rigged number called the LIBOR & price interest rate derivatives on the spead between this rigged fixed rate & the floating rate. Then the public loses money & we scream - hey who the fuck rigged the LIBOR in the first place ?:) Talk about bootstrapping... honestly, we are no different from Dr. Quetelet. Might as well base Google's quarterly earnings on Page's BMI.
Quetelet's work was reasonably appropriate for his time. One source I recall (cannot find it now) stated that he was looking for a rapid paper-and-pencil method for estimating aggregate troop strength. As a basic, good-enough first approximation, BMI works reasonably well. The problem is that it's now been grossly misapplied.
Quants (and economists) spend, I suspect, a lot of time chasing what turn out to be transitory relationships between various measurements. The relationships may very well be statistically valid, when first measured, but either changing real-world conditions, or, more likely in the financial world, adoption of the relationships by others seeking profits, rapidly erodes their validity.
Thanks for the links, I've got some reading for later :)
An enterprise I had some association with in the past spent some years and considerable capital trying to establish a strongly predictive relationship between a physical property and a health measurement. It turns out that there was a relationship, but not one which could be reliably used to determine an accurate quantification of the health measurement. This didn't keep them from torturing the data extensively.
And Page's health is in the news today: http://www.washingtonpost.com/google-ceo-larry-pages-health-...
By moving to a city with a lot of S.E. Asian people I'm suddenly above average BMI - I need to move back to Texas.
On the other hand, all those little old Japanese women mean that the average life expectancy here is among the highest in the world - so I should live to a 100 like them
But whether you're muscular or fat, this scale is a pretty good reminder of how well nourished we are compared to everyone else.
"You're most like someone from DR Congo"
They clearly mean I'm heavier than most people, but lighter than the average BMI. Very large people skew the average upwards.
Edit: I am wrong. See the comment below. I thought average referred to arithmetic mean.
You're messing up "average" and "mean" :)