The CDC is airing its dirty laundry this week, as a new report comes out claiming that last year’s CDC report on obesity is basically hogwash. In the old numbers, obesity was this bomb descending on America that was going to wipe us out. It claimed that obesity caused 400,000 deaths/year, making it the number two cause of death. Thus, obesity wipes out the equivalent of Utah Valley every year.
Not that Utah Valley is particularly fat, I’m just giving a hypothetical.
The new report, issued by another division of the CDC, uses more recent data and pays more attention to statistical problems. They find that there are about 25,000 obesity caused deaths. So it wipes out, I don’t know, most of Palo Alto each year. This is a 14-fold downward revision. And just to rub salt in the wounds, they find that being classified as underweight causes slightly more deaths than those caused by being overweight. Don’t you hate it when that happens?
The new report also finds that having a slightly higher BMI predicts better long term health (don’t miss the graphic on this one!). Of course, this optimal BMI is technically in what is now the “overweight” range, which makes one wonder what it means to be overweight. Apparently, it means you live too long… This is called “the obesity paradox” by one source. Well, I suppose the world is full of paradoxes for some people.
Next year, the other half of the CDC might come out with a new report slamming the old one, as some people are quite convinced that obesity is an epidemic. But I just wanted to take a moment to chuckle at the audacity of it all. Researching the mortality effects of obesity is incredibly difficult. The research designs, even when done as well as possible, leave a great deal to be desired because they cannot control for so many important unobserved differences. So it is, I think, incredibly important to be up front about these limitations.
Besides the unobservables, how do we think about competing risks? You see, everybody (well, almost everybody) dies. The only question is, what will get you first? Thus, suppose I define a carcinogen as something that increases your chance of dying from cancer. Well then, bypass surgery is a horrible carcinogen. By stopping you from dying of heart failure, it massively increases your risk of dying from something else. Now, this is not my field, but similar issues come up in economics, and it’s a real pain to try to deal with in a compelling way.
There are lots of people out there who have been convinced that we have scientific proof of some favored proposition, only to have their faith dashed against the rocks of new data or new methodology. We do the best we can, but the arm of flesh, statistically speaking, has low power. Not that I’m all that impressed with the non-empirical academic approaches either. We, as a people, may not be great at getting revelation, but we’re not so hot at other methods of finding truth either!