Saturday, August 25, 2007

The art of lying with statistics

daphnia
Are you afraid of this monster? I would guess no, because this is just a small water flea called daphnia. But when I was a student of preclinical medicine, sitting in a biology course and looking through a microscope, I shrug back with shock when a daphnia all of a sudden jumped into my magnified field of vision. Mountain or molehill - it is all a question of perspective.

This also holds true for an article that I just have read in my newspaper: "Overweight surgery reduces mortality." The article reports a 30 percent reduction of mortality risk after bariatric surgery in Sweden.

Thirty percent, this sounds impressive if we are not aware that this figure comes out of a statistical microscope. It seems to prove that overweight is a cause of death and that the risk can be reduced by a considerable amount if we lose weight.

But let's have a look at the facts and figures. In Sweden, 2010 people with an average body mass index of 42 have been treated by gastric surgery and have lost between 14 and 25 percent of body weight during ten years after surgery. Another 2037 people with the same BMI average have been treated conventionally, and their weight remained stable in a range of plus/minus 2 percent of body weight. After ten years, 101 persons of the surgery group and 129 of the control group have died.

How to make 30 percent out of 1.3 percent

Now we begin to see how the statistical microscope works. It ignores the vast majority of persons who survived and puts the focus on the 6.3 percent mortality in the weight-stable group and 5.0 percent mortality in the weight-loss group. That is, the risk has been reduced by 1.3 percent. This is not much, you may even call it meek. But you can calculate a relative risk: Blow up the 6.3 percent to a full hundred percent, then 5 percent equals 79 percent, and you got a relative risk reduction of 21 percent. You may do even more and blow up the 5 percent to a full hundred percent, and you get even higher figures. Then you can control for sex, age and risk factors and get an adjusted hazard ratio of 71 percent, that is, a relative hazard reduction of 29 percent. Round it up and you get the 30 percent reported by the newspaper.

The misleading term "overweight"

The Swedish study has been done with extremely obese people. I am ready to accept that extremes cannot be very healthy, be it a fashion model with a BMI of 16 or a man with the stature of a sumo wrestler and BMI 50. Therefore I am not very surprised to see a small positive effect on health with a massive change of metabolism (weight loss is not necessarily the important point) in very obese persons. But this has nothing to do with "overweight" as stated in the newspaper headline.

"Overweight" is widely used for persons with a body mass index higher than 25. The story reported by the newspaper has nothing to do with this group of people. But the headline will once again make them believe that they should lose weight in order to get healthier. A view absolutely not supported by the report.

Photo credit: flickr.com/photos/idua_japan/177006163/

No comments: