Over at Pure Pedantry, Jake has a nice post about a study showing that the ever-popular Body Mass Index measure is not a good predictor of the risk of heart disease. He’s got a lot of details about the study, including this graph of risk vs. BMI:
Now, here’s the thing. This is the second study I recall hearing about that has a similar result– there was a flurry of articles a while back about a large study (or maybe one of those meta-studies) showing that people who were slightly overweight according to BMI had lower mortality than those of “normal” weight. And now, this study shows that people who are overweight according to BMI have a slightly lower risk of heart disease.
My question is, who’s the idiot who decided what “normal” is? (Mroe after the cut…)
I mean, as a physicist, I would like to think that you would set the “normal” level by basically making the sort of graph shown above, only without the descriptive labels, and then choose the point of lowest risk as your “normal” range, and set “overweight” and “underweight” relative to that.
Ideally, you would repeat this for several different risk factors– heart disease, diabetes, whatever else you associate with being overweight– making graphs and choosing the local minimum for each as your “normal” point. I mean, that would seem like a sensible definition of “normal” weight, in the sense that this is being used, namely the ideal that healthy people are supposed to be striving for. Ideally, you’d want everybody to try to reach the point of minimum health risk.
And yet, it doesn’t seem like that’s what was done. Or if it was done, it wasn’t done terribly well, as there are at least two measures by which the range currently being called “normal” is a higher risk than the range currently being called “overweight.”
So how did they pick the current value? It’s not some graven-in-stone number that’s been handed down through the generations, I know that, as nobody talked about “BMI” back when I was in high school, and I remember at least two rounds of news stories over the last fifteen or twenty years talking about the government lowering the values considered “normal.” But if they didn’t take the sensible empirical appoach (they can’t have, otherwise the recent studies wouldn’t be publishable), what did they do?
What is the population in the graph? Is it all people or people with existing CAD? <clicky clicky>. From the summary originally linked:
Thus, the graph reflects mortality in persons with a “history of percutaneous coronary intervention, coronary artery bypass graft, or myocardial infarction,” not in the general population. That said, BMI is still a very crude metric.
My guess on the “underweight/normal/overweight” decision is that it was arrived at in the same way that “heat index” was — a bunch of assumed averages. I believe that there are a significant number of perfectly healthy people who fall outside the normal BMI range.
Dang! Take a look at those error bars! Looks like there’s a lot of variance in the health of “severely obese” people.
Craig: Thus, the graph reflects mortality in persons with a “history of percutaneous coronary intervention, coronary artery bypass graft, or myocardial infarction,” not in the general population.
Sure, but if obesity is a major risk factor, you’d still expect the “normal” people to fare better than the “overweight” people. It might reduce the magnitude of the effect, but the correlation should still be there. The “overweight” group is pretty clearly doing better, though.
My guess on the “underweight/normal/overweight” decision is that it was arrived at in the same way that “heat index” was — a bunch of assumed averages.
The fact that the category breaks just happen to occur at nice round numbers– 20, 25, 30– tends to suggest that there’s a lot of fudging going on here. I wouldn’t expect real biological processes to give you such convenient divisions.
Dave: Dang! Take a look at those error bars! Looks like there’s a lot of variance in the health of “severely obese” people.
Some of that’s probably statistical– there are probably fewer people in that group than the others.
I bet they did talk about BMI, or something like it, when you were in high school. When I was in high school, they did have this thing they called the “fatometer”, or skin fold tester, that they subjected us to in PE. It did something similar, told you how fat you were, and then how much you should worry about how fat you were.
The cynical part of me believes that what is defined as “normal” is defined by those commercial interests who need more and more people to suffer from anorexia and similar diseases, so that they can stay in business selling all kinds of quackish weight loss remedies.
-Rob
The study as I read it concludes that a BMI in the “overweight” to “obese” range could be mildly beneficial for those with CAD. I would like to see a similar study (and google would probably be sufficient for this, but I’m feeling lazy) on the risk of developing CAD in those without it over the same BMI range (and junk the labels, use the numbers.)
I’m sure you’re right on that. I remember seeing a cartoon with Kate Moss as the person in charge of [some agency which defines the government standard for weight,] with the caption of “Let them eat celery.”
If they used the term “normal” in the study, it is incorrect unless normal BMI is, indeed, the most frequently-occurring condition. Surely the more appropriate term would be something like “the recommended BMI” (based on whoever does the recommending). If “recommended” is used, then surely (I hope) there is a reason based on some kind of morbidity statistics. If that be true, then it’s hard to argue with the recommended BMI. If the recommended BMI is not based on any kind of health data, then it must be based on some reasoning other than medical.
I dislike BMI. In college, I was an athelete and in very good shape. My BMI classified me as overweight. Now, I’m not in such great shape, my fat-to-muscle ratio has increased, but my BMI hasn’t really changed.
That said, this one study is hardly a good reason to say that “normal” (or “recommended”) was put in the wrong place. A study released today measured mortality and, if anything, suggests that just a point or two into the overweight range produces a real rise in risk.
http://www.npr.org/templates/story/story.php?storyId=5691917
That one study doesn’t prove anything either (I’ve heard what sound like reasonable criticims of it), but maybe whoever decided on “normal” wasn’t an idiot after all.
Rob: I bet they did talk about BMI, or something like it, when you were in high school. When I was in high school, they did have this thing they called the “fatometer”, or skin fold tester, that they subjected us to in PE. It did something similar, told you how fat you were, and then how much you should worry about how fat you were.
Yeah, they did that. And that measurement, I’m more inclined to trust, as it actually measures something real about a specific individual– body fat percentage, I believe.
I don’t recall hearing about “Body Mass Index” as a simple function of height and weight until I was in grad school.
Mark: If they used the term “normal” in the study, it is incorrect unless normal BMI is, indeed, the most frequently-occurring condition. Surely the more appropriate term would be something like “the recommended BMI” (based on whoever does the recommending).
Well, the graph above uses “Normal Weight” for one of the category labels, so…
Colst: That said, this one study is hardly a good reason to say that “normal” (or “recommended”) was put in the wrong place.
It’s not just this one study– there was another one a year or so ago that found something similar, and also the stories mentioned here quite a while ago about how the BMI classifies most NBA players as obese. But you’re right, it’s a complicated issue.
A study released today measured mortality and, if anything, suggests that just a point or two into the overweight range produces a real rise in risk.
That one study doesn’t prove anything either (I’ve heard what sound like reasonable criticims of it), but maybe whoever decided on “normal” wasn’t an idiot after all.
The news story makes it sound like that study relies on self-reported data mailed in by AARP members, with no empirical confirmation of their numbers. Those are methodological holes you could drive a truck trhough. A really fat truck, full of yummy, greasy burgers.
You’re probably right, though, that “idiot” is a little harsh. Then again, “idiot” is somewhat more likely to get people to read the rest of the post, and we’re traffic whores, here at scienceblogs.com…
I do believe the BMI in the US changed in 1998. It added 30 million Americans to the obese range. Alicia Mundy’s book “Dispensing the Truth” suggests the diet industry helped push such changes to increase demand for their products. Her book focuses on Fen-Phen. It doesn’t seem to be the most scientific way of figuring whether someone in “overweight” or not. Some people do have broader shoulders than others, not to mention head size, etc. The BMI assumes avg. proportions. Also where you carry your fat is important it appears with regards to health as well. Plus who is to say that all human beings have the same “optimal” fat percentage. At the very least, the BMI should include more dimensions such as skeletal frame size. Of course then it wouldn’t be as easy to look at and useable as a PR tool for public health.
BMI exists almost solely because it’s easy to calculate, and is somewhat accurate for most people. Bodyfat percentage is likely to be much more closely related to health issues, but requires the skinfold measurements described in earlier comments. Those take time and cost money and require gadgets. On the other hand, everyone knows how tall they are and how much they weigh, and most people can multiply and divide.
I’ve seen other graphs — and I’m sorry I can’t give you a citation or even remember exactly what risk we’re talking about (possibly mortality), but I could track it down if you really want to know — that show WILDLY different risk-vs-BMI profiles for smokers and nonsmokers. Which makes sense, really. Nonsmokers have to be much heavier than smokers before they reach the same level of risk.
You seem to be only a couple of years older than me, and I definitely remember BMI from high school. But I’ve often wondered how they came up with “normal”, myself. On growth charts (currently under review, IIRC) they do have percentiles, so perhaps “normal” just means “the middle of the distribution of people we measured at the time we measured them”. Or something.
Given that BMI scales with height squared IIRC, it’s pretty much useless in reality. Colst’s anecdote is a perfect example.
(No, I’m not gonna spend the time to hunt down links for that.)
The BMI ratings for normal were a translation of height/weight actuarial charts that were developed by the life insurance industry decades ago. If one really wants to make any kind of bet on the overall outcome, one has to argue against the people who have the most at stake in the outcome. It is true that there have been changes – the original tables had separate categories and risk levels for light, medium, and heavy frame sizes. This opens up a question about the accurate measurement of frame type, so the three tables were simplified into one. The same thing has to be said about the break-points at 20, 25, 30 etc. – simplification.
One of the factors that have allowed the deepest point of the U-curve to shift rightward has probably been the treatment of hypertension in those in the overweight and moderately obese categories. One of the factors that almost certainly has effected the “underweight” category is cachexia from cancer [or more recently from AIDS] changing blood chemistry to increase the risk of some types of cardiovascular disease.
Most clinicians are now realizing that it is the distribution of fat that is the real issue, with central abdominal/intraperitoneal fat being dangerous — the most easily recognizable sign of metabolic syndrome [hypertension, hyperglycemia, hyperinsulinism, hyperlipidemia, hyper-LDL, hypo-NDL]. Being in the obese category and having metabolic syndrome has been definitively shown to increase risk. Subcutaneous or peripheral fat does not seem to be nearly as risky. Skinfold thickness measures sub-q fat at defined sites and assumes a certain amount of intraperitoneal fat in the regression equations used to estimate fat composition [the equations were derived from cadaver studies].
I have tended to assume that ‘normal’ got defined by the insurance industry as per their actuarial data. And, that the BMI, as we know it, was derived from there. Putting “BMI waist hip ratio” into a Google search window turned up a fair amount of conflicting (conventional wisdom vs scientific) reports without much consistency on position as reported by source. That said, there was an interesting bit at: http://www.consumeraffairs.com/news04/2005/obesity_waist.html which cited the Lancet. “… The authors state that compared with BMI, waist-to-hip ratio is three times stronger than BMI in predicting the risk of a heart attack. Larger waist size (which reflects the amount of abdominal fat) was harmful, whereas larger hip size (which may indicate the amount of lower body muscle) was protective….” There does seem to be a tentative position that adipose tissue in the abdominal region is not identical to adipose tissue deposited elsewhere. And, particularly in women, adipose tissue in the abdominal area is closely related to the incidence reporductive cancers. The suspicion being that there is a concentration of hormones in that adipose tissue that is inconsistent with health. I would argue that, short of skin calipers in the hands of someone professionally trained, most measures of ‘weight,’ as it relates to health, are crude at best. And, at the risk of inviting the scorn of those more adept at measurment in the physical sciences, I’ll posit that it is hard to come to good conclusions when you have poor measures. [Okay; I’m braced for it.]
Thanks to natual cynic who said it faster and better than I did.
Dave: Dang! Take a look at those error bars! Looks like there’s a lot of variance in the health of “severely obese” people.
Chad: Some of that’s probably statistical– there are probably fewer people in that group than the others.
I suspect that there’s multiple populations aggregated into that group. that the distribution of outcomes is actually multimodal. Sometimes obesity is a symptom of some underlying condition. Such people would have the worst modal outcome.
That would suggest, by the way, that weight, on its own, simply isn’t an adequate diagnostic tool. Which is what the study said.
Mary Kaye: Skin calipers are not a very effective way to estimate risk as it relates to body fat. It just estimates body fat to within about 2.5% of the true value in the hands of the best practitioners. And, since it works on a simple linear regression equation it does not give a clue about the intraperitoneal fat. To truly measure i.p. fat you have to have a CAT scan or some other device that can distinguish fat between the abdominal organs and scan the liver. The fat area can give a good value. Studies have been done that show vasly different amounts of i.p. fat in people with nearly identical total fat. As a cheap, dirty and effective proxy measurement a waist circumference or a waist/hip ratio gives a pretty good indication of dangerous abdominal fat. Circumferences is conjunction with blood lipids, glucose and pressure give an easy indication of cardiovascular risk.
(Dup post from somewhere)
Ni Hao! Kannichi Wa!
BMI/central obesity is only one of a number of whole organism parameters of metabolic syndrome (syndrome X) whose etiology is chronic positive imbalance of calories in versus calories burned.
Whether the imbalance of where the excess calories come from–fat, carbohydrate or protein–make a difference is a second generation question.
BMI, although only one parameter, receives most attention because it is visual (like skin color, ugly/pretty, etc.) rather than unseen parameters as chronic high blood pressure, high LDL/HDL, high cholesterol, glucose intolerance, insulin resistance (type II diabetes) and fatty liver diseases.
Each of these are correlative observations for which it is unknown whether they are independent, upstream or downstream of the others in terms of cause and effect.
I wonder if the dietary habits of these patients, e.g. calories in versus calories out, source of calories, and the parameters of metabolic syndrome were measured in this study.
Western and developing societies are faced with a major dilemma of imbalance of calories in v. calories burned, the genetic drive for food and as much of it as possible (not unlike the reproductive drive), addiction of taste (not unlike the pleasure associated with reproductive events), and increasing cost-effectiveness in marketing to satisfy those drives (not unlike things like the diverse cosmetic, pornography, Hollywood-based industry, etc.) which only in the case of food drive, the tip of the iceberg was approached in Morgan Spurlock’s Super-Size Me.
Ironically, obesity is an example of a selected trait in evolution for survival that has become unnecessary or a detriment in developed societies (perhaps similar to the drive to reproduce).
Metabolic Syndrome may be an adaptive response intrinsic to current stage of societal evolution in eating and exercise habits. The degree of reversibility versus irreversible imprinting is a key unknown as well as its inheritability.
MOTYR
I don’t remember hearing about the BMI until perhaps ten years ago, but I do remember something called a “ponderosity index” at least as far back as 1989 or 1990. It was a similar metric but used inches and pounds rather than SI units.
As far as comparing outcomes in patients with known CAD vs. those in healthy people, higher BMI values are clearly associated with greater CAD risk factors, and high BMI may be an independent risk factor in itself (see this, this and this. Like the study under discussion, most of the studies I dug up examined risk and outcomes in people already diagnosed with CAD.
Chad:
“It’s not just this one study — there was another one a year or so ago that found something similar…”
You’re most likely thinking of the study done by Katherine Flegal’s team at the CDC and published in JAMA last year. This was the equivalent of an internal review designed to pin down the number of deaths per year attributable to obesity (a figure which will always be nothing more than a decent guess at best). After the findings were reported, docs at the Harvard School of Public Health took Flegal to task for poor methodology and sketchy assumptions.