The Incredible Shrinking Obesity/Morbidity Link

What is up with these obesity statistics?  Last March, Fox News ran this:

Group Challenges CDC's Obesity Claims

..."America is now suffering from an epidemic of obesity myths much more than an epidemic of obesity," said CCF senior analyst Dan Mindus.

Mindus authored a report that attempts to shatter "obesity myths" and takes direct aim at what Mindus says is the principal culprit for obesity hysteria in America — the Centers for Disease Control and Prevention (search).

Last year, the CDC publicized a study that said 400,000 deaths annually can be attributed to obesity

"The CDC has misled the American public, has compared obesity to the black death, has told us that obesity is going to kill more people than tobacco," Mindus said. "They're completely wrong. They know they're completely wrong and yet they're trying to sweep under the rug all the evidence to support that."

But Fox also provided a "fair and balanced" counterpoint:

The Center for Consumer Freedom also has its skeptics. The group gets most of its funding from some of the biggest names in food, including fast-food chain restaurants and food manufacturers.

One medical director who treats overweight people full-time as director of George Washington University's weight management program said quibbling about the numbers blurs the true story.

"Let's assume that it's not 65 percent, let's assume that it's 50 percent. Let's assume that it's 40 percent. It's still a lot of people," said Dr. Arthur Frank. "I think [the consumer group is] being petty and I think they're being silly."

"Silly"?  Well, who looks silly now?  This latest CDC study, as described in the Times, is a shocker:

People who are overweight but not obese have a lower risk of death than those of normal weight, federal researchers are reporting today.

The researchers - statisticians and epidemiologists from the National Cancer Institute and the Centers for Disease Control and Prevention - also found that increased risk of death from obesity was seen for the most part in the extremely obese, a group constituting only 8 percent of Americans.

...The new study comes just 13 months after different researchers from the disease control centers published a paper warning that obesity and overweight were causing an extra 400,000 deaths a year and were poised to overtake smoking as the nation's leading preventable cause of premature death.

That conclusion caused an uproar, and scientists, particularly those who examine the consequences of smoking, questioned the study's methods. In January, the agency's researchers corrected calculation errors and published a revised estimate of 365,000 deaths.

Now the new study says that obesity and extreme obesity are causing about 112,000 extra deaths but that overweight is preventing about 86,000, leaving a net toll of some 26,000 deaths in all three categories combined, compared with the 34.000 extra deaths found in those who are underweight.

Dr. Donna Stroup, director of the Coordinating Center for Health Promotion at the C.D.C., noted that the previous study had used different data and different methods of analysis.

"Counting deaths is not an exact science," Dr. Stroup said.

"Not an exact science".  That, I believe.

Now, these doctors know more about this than I, and presumably have given it more thought.  However, two points strike me:

(1)  the Body Mass Index is based purely on height and weight, with no adjustments for age, gender, or physical activity.  This leads to odd results, such as the news that First Fitness Fanatic George Bush (6 ft, 190 lbs.) is overweight, or that Alex Rodriguex is nearly obese.

(2)  Correlation is not causation: Physical fitness seems to be a much better predictor of health than a simple overweight/underweight calculation, because it is possible to be fit and fat, or sedentary and skinny.  However, "overweight" and "sedentary" tend to be correlated, which may be confounding the stats.

Dr. Williamson, one of the authors of this study, would also like to thank you for not smoking.

Comments

[Aric]

I've always thought that the fundamental problem with BMI is that it's based on the *square* of your height -- human beings are three-dimensional creatures, so if you scale a linear dimension, the mass should increase or decrease with the *cube* of the scaling factor.

Has anyone ever looked into a BMI based on height^3 instead? Would that give more reasonable results?

[Cecil Turner]

That's funny, I always assumed they used a cubic scale, but a quick check of their calculator confirms it's square. Which sure doesn't seem to make much sense. I have a hard time believing they're really ignorant of the square-cube law. Is there some esoteric body mass function I'm unaware of?

[ryan]

I haven't seen the CDC's numbers or equations, but it seems to me that it'd be easy enough to solve the correlation problem between weight and sedentary lifestyle. Simply include a variable for "sedentary" in your regression and you'll have the effect of being overweight independant of being sedentary. (I have no idea if the CDC accounted for this, and since their numbers were so badly skewed before we certainly can't trust them without verifying.)

[Cousin Dave]

I can well believe that being slightly "overweight" according to the government charts is actually healthier for you, because it appears to me that the proscribed weight targets are actually underweight for most normal people of a given height and build. Weight targets have been revised downward in the last several years or so, creating (intentionally?) the perception of a hugely growing problem where, most likely, no significant problem exists.

[ryan]

Cousin Dave,

Regarding the perception of a hugely growing problem, the logic of bell curves suggests that any increase at all leads to apparent exponential growths in obesity rates. That is, if the weight of the average person increases even 1 pound every decade, a graph of obesity rates will show an apparently exploding obesity rates, leading to the conclusion that some massive change must be occurring.

[Paul Zrimsek]

I know of some 8,000-200,000 Iraqis who might agree with the bit about counting deaths not being an exact science.

[dsquared]

The real trouble here (and this is not a criticism of the study; just an observation that computing actuarial tables is *difficult*!) is that the death rates in this dataset are being driven by people who are old. People who are old and overweight are not likely to have been overweight their whole life. People who are, today, young and overweight, are likely to be overweight for their whole life. So the study leaves us slightly better informed, but perhaps not as much as we'd like, about what we might be doing to ourselves by being so fat. But as I say, this is not a criticism of the study - the issue is intrinsic to the problem of trying to make guesses about mortality rates from causes which have their effect a long time in the future. I suspect that people on this thread are correct to say that what this study shows is that BMI as calculated here, isn't really a good instrument for "overweightness".

btw, Ryan, in principle you could add a variable for "sedentary". However, there is no such variable in the database, so you would be looking at a wait of several years before you could do this even if you added it to the next major survey. And I'm also coming up short when I try to think of some objective criterion which might measure the extent to which someone was sedentary or not.

[Victor]

What is up with these obesity statistics?

I suspect that we have been and are being fundamentally misled. Simple enough. This is not unique to BMI.

The general approach is methodologically unsound. Just because we can calculate BMI doesn't mean that it is useful. And just because some people on the extremes ARE too obese and it will kill them doesn't mean that otherwise healthy people who resemble them are in danger. And just because some people do NOT resemble them (i.e., low obesity) doesn't mean that THEY are NOT in danger. Just because we can categorize doesn't mean that the categorizations are useful.

So, the number of deaths attributable to obesity is off by a factor of 14. No duh.

[kim]

Go Feast, Young Man.
====================

[lastchaos money]

I do not like to go shopping, because it always spends a lot of money, but I never hesitate to buy lastchaos money.

[mabinogi gold]

When you have mabinogi gold, you can get more!