(or: Why Drawing Conclusions From a Limited Data Set is Bad.)
The news is out, folks. According to several click-baity articles sweeping across the intertubes, 72% of Americans are overweight or obese. Seventy-two percent! THAT’S LIKE, ALMOST ALL OF US. It’s crazy that any country, especially one as large and wealthy as the US could be so woefully, piteously, ludicrously fat. Our fast-food-living, sedentary lifestyle-having, beer-swilling nation is the poster child for dietary excess. It’s just embarrassing. THINK OF THE CHILDREN!
The numbers come directly from the CDC, too. This is not a scam perpetrated by some right or left-wing think-tank or diet industry juggernaut. Our own CDC posted the tables and the news is pretty damn grim. There is no debate to be had here. America is out of shape.
Pump the Brakes
But before you make that New Year’s Resolution you have no intention of keeping, let us folks here at Bullshido clarify those numbers and graphs for you just a skosh. To be honest, my skeptic senses are tingling just a teeny tiny bit here. It’s that damn 72% number and something I heard once that goes like this: “Extraordinary claims require extraordinary proof.”
Something smells wrong. When we look at the data, we find that the CDC classifies anyone with a Body Mass Index (BMI) greater than 25 as “overweight.” Find yourself over 30 and you are “obese.” This has the stink of objective numerical classification to it, and normally mathematical tools that defy qualitative adjustment make me all warm in my engineer-parts. So let’s talk about BMI. How solid is this arbiter of health and fitness?
It’s a formula that looks like this:
Let’s look at the math, first. BMI is your mass in kilograms divided by the square of your height in meters. This is not a complex algorithm. It is one data point divided by another. This bizarre lack of mathematical sophistication in so widely used a metric was a deliberate choice; and the reason for this is both simple and justifiable.
The BMI was developed somewhere between 1830 and 1850 by Adolphe Quetelet, a Belgian astronomer, mathematician, statistician, and sociologist. You will note that his qualifications do not include “medical doctor,” “physician,” “dietitian,” or anything that might have a tangible correlation to the science of human metabolism and its relationship to obesity. His primary goal was to categorize the relationship between height and weight across multiple countries and populations. This had nothing to do with obesity or health, but rather it was his sincere (and possibly racist) attempt to classify human beings by their physical traits. He and several other statisticians were trying to create a referenceable codex of physical characteristics by race and region for the scientific study of anthropomorphic differences. He was a statistician doing statistics. Cool. Statistics is scientific stuff.
The main problem he faced was the huge variation in the data. To reduce the noise, Adolphe condensed the unruly mess of height and weight data for an entire population to a simple ratio. He incorporated a square function to allow the ratio to account for volumetric changes, but in the end this calculation results in a measure of thickness rather than density or obesity. Squaring any dimensional quantity calculates the area in units of a square. It’s why we call it squaring. Adolphe squaring the height of a subject effectively converts us all to two-dimensional quadrilateral polygons of uniform height and width. This was a quick and simple way of reducing the entirety of human physical uniqueness to the average of the average average. Why? Why would so blunt an instrument be employed? Because Adolphe was only looking to sort people by geography and ethnicity and did not care about variations in physiology that went beyond average height as a function of average weight.
For his purposes, using this basic ratio squashes a vast quantity of data into a very simple number. This is a common statistical tactic useful for making millions of data points readable, at the cost of drawing meaningful conclusions. The most critical failure of this metric as a public health tool is that there was literally no consideration made for fitness, fatness, health, or any other metric beyond the ratio between a person’s height and weight. The health, shape, and actual body composition of the populace was irrelevant to his goal and remains entirely unrelated to the purpose of the metric.
The first modern person to apply the BMI to public health was controversial health researcher Ancel Keys. His failures as a public health professional are an article unto themselves, but his “Seven Countries” study on fat consumption has been categorized by many as bad science perpetrated for personal gain. He is single-handedly responsible for fifty years of damaging public health policy linked to corporate intervention. One of his lesser-known crimes against health is his endorsement of the BMI as a health index, which he leaned on heavily in 1972 to bolster his claims.
It worked. By 1985, the BMI was entrenched in public health policy. Part of the reason for this lay in the sheer convenient simplicity of the BMI as a metric. It compressed giant populations into a neat scale with some basic zones. Doctors could use this to parse their patients into risk categories quickly and easily. To that effect, the BMI does have actual merit. There is some correlation between higher BMI and certain health conditions across populations. It is not a particularly reliable correlation, but it does exist.
Where the BMI has very real utility is at your annual physical. When you go to your annual physical, your doctor can look at your BMI and use the number there to direct the examination. If your BMI is over 35, the examining physician has a good idea of what to check first to ensure your health needs are being met. The doctor can take a good look at you and compare you to this handy reference I just made.
If the patient looks like the guy on the left, the discussion should focus on ensuring that the patient is training properly and not mainlining HGH through the eyeballs. If the patient resembles the guy on the right, then a discussion of healthier habits is probably warranted.
The doctor sees hundreds of patients a year. A quick reference point for starting a conversation about health and weight is handy and useful if you have a lot of people to examine. If that was how the BMI got used all the time, it would be a perfectly good heuristic. It’s a fast way to determine if a patient’s diet and exercise habits need a closer look. What it doesn’t do is tell you if they are in shape, healthy, or any other descriptor pertaining to medical recommendations.
Where the Standards Fail:
Take a look at this chart.
According to the BMI standards of the CDC, a male human who is 5’9” tall should weigh between 125 and 160 lbs. Here is a picture of a male human who is 5’9” tall and 155 lbs.
What do we notice about this person? Maybe I’m being unfair. Let me find another. Here’s one:
Okay. That’s MMA champion Conor McGregor. Sensing a pattern. To be perfectly clear, the CDC considers both of these people to be perched on the very precipice of “overweight.” Process that. Let it marinate. Now let me stop posting fitness gods and see what a truly “Overweight” person (BMI > 25) looks like.
This is the type of person the CDC calls “overweight.” This is not an uber-athlete or weightlifter, either. This is a regular dude who posted a “before” picture on the internet to motivate himself to get in better shape.
Hey ladies, think it might be better for you? Has society not body-shamed you enough yet? Try this on for size. Both the following are pictures of “overweight” women. If that makes you want to scream into your chocolate double-whip venti macchiato, then go ahead. We get it.
The taller you get, the worse it gets, because our boy Adolphe did not have an excess of folks over six feet tall to deal with in 1840. Mathematically, because the BMI forces all heights into a square, it scales extremely poorly with height. The taller you are, the larger your square is, but the ‘standards’ zones do not not square, and the weight component scales linearly. What does all that mean? It means that the taller you are, the lighter you need to be per unit of height to stay “not overweight.” So a man who is 6’2” has to weigh under 195 to escape the “overweight” label. Picture a 6’2” guy weighing less than 195. Sure, they exist. But are they any healthier than the two porkers pictured below?
One of these is legendary actor and wrestler Dwayne Johnson (6’5”). The other is me (5’10”). Can you tell which is which? Either way, according to the CDC, with BMIs between 32.5-35, We are both way,way, fat, bro. “Obese,” even.
Now, this is where “72% of the US is overweight” claim starts to make sense. It is actually pretty challenging to get into the ideal weight range on the BMI chart. If you have ever spent even cursory time in the presence of a barbell, you ain’t gonna make it. Why? Because BMI does not account for density and the standards are strict in order to drive policy. Reality is a distant third in this calculus.
Once again we find ourselves banging up against the limitations of so basic a metric. We know that an entire spectrum of body types exist across each individual BMI number. The BMI sacrifices all that information at the altar of simplicity. Adolphe did not apply “standards” to the original version because there were none and he did not care. But Ancel Keys and the CDC needed standards for their charts. So instead of parsing the groups by nation or ethnicity, they drew some lines on the graph and decided who was fat and who was not.
How did these standards get decided? Well, you’d have to talk to Ancel Keys and the CDC about that, because Adolphe did not give one whit about standards when he built the BMI. The CDC has acknowledged the limitations of the BMI in several publications, and defends the use by pointing out that as a metric for evaluating a population, it’s no worse than any other.
This is, of course, untrue. Several easy body composition tests are readily available across the world, and almost all provide better data than the BMI. While each has limitations, every single one of them produces a better analysis of body composition than the BMI. None of them, however, produce a single convenient number that can be indexed across a large population with quite so little effort. In defense of the CDC, BMI is so limited in its utility that they admit they really are just doing the best they can with it. In fact, some research has been done that shows that BMI and total adiposity are only loosely related.
Comparing body mass index (BMI) and percent body fat (“%BF”) in a 1994 study of 8,550 men. See the upper left and lower right quadrants for the limitations of BMI in assessing body fat. If there was a strong correlation between BMI and body fat, then the blue dots would align along the black line without wandering too far away from it. Instead, we have a big ol’ Rorschach that is slightly oblong. This indicates a correlation, but the messier the blob, the weaker it gets. This is a pretty messy blob. For 2,105 out of the 8,550 participants (24.6%), BMI not only failed to provide good information, it actually provided BAD information.
So I’m Not Fat, Then?
“Fat” is a subjective term. Statistically speaking, you are more likely to be “out of shape,” or “unfit.” Both of which are things the BMI cannot differentiate from “in shape” or “fit.” So before you sigh and wipe your forehead in relief, consider all the ways you CAN effectively evaluate your health and try some of those on for size. The only way to determine your health and fitness state in real time and in the real world is to go to a doctor. Get some bloodwork done, check your body composition with equipment and standards designed for that purpose. Now you have relevant data to work with. If your BMI is high, and you don’t score so well on any number of available body composition metrics, then you have some work to do. What will really cheese you off is if your BMI is low and you STILL score bad on body composition analysis. Guess what? You’re still out of shape. You’re just light and out of shape at the same time. There is no prize for that.
So What About the ClickBait?
As far as the “72% overweight” statistic is concerned, you need to look at that with a very squinted eye. By itself the statement is meaningless. Both the metric and the standards applied with it are so deeply flawed as to be useless… by themselves. What if we applied the BMI the way Adolphe intended it to be applied? What is the AVERAGE BMI of the american populace versus others?
What might we find, then?
We would find that the US, with a mean BMI of 28.5 (rank 20/190), is at best slightly overweight by the CDC’s standard. A standard that feels rather stringent when examined closely. This is how the BMI was meant to be employed: as a static, dimensionless reference point for comparing populations. In this case, it lets us know that we are a touch heavier per unit of height than many other countries. Like… that’s it. That’s all it means. Literally no other assertions can be made from the data without applying conjecture with unreliable correlations.
I’ll be blunt, here. We are at a point where BMI has rapidly diminishing returns as a metric. It had purpose and utility at one point, no question. Before computers and the internet, brevity and clarity were extremely important. Sifting through endless lines of data and trying to parse meaning from stacks of papers was a task fraught with peril. It made a lot of sense to condense data into indexable metrics. We don’t have that excuse anymore. Nevertheless, BMI is still used today to influence public policy, as well as affecting healthcare and insurance rates.
If all it did was create clickbait then it would not be so bad. But the next time you try to buy life insurance and they ask you for your height and weight to set your premium, think about how it might be harming you. When the next round of soda taxes or food regulations percolate through the various legislatures, check and see if they used BMI standards to make the decisions. Public policy and health care costs are a big deal. The metrics used to drive either should be as robust as the potential consequences.