Like most people, I was mildly disturbed by the headlines last month proclaiming the results of a study that claimed that people with overweight friends tended to become overweight over time. Like a lemming, I mentally thought over the size of my friends, and then shook my head. I know my weight gain is based in a bad reaction to a medication. It has nothing to do with how much my friends weigh.
So I was pleased to run across an article helping to put the results in context. I'm not going to say "debunk," although that's close. An excerpt:
I wish I had been clever enough to notice the observational setup of the "experiment" when reading the original write-up. After all, I spent that whole semester in Research Methods in graduate school.
On what firm foundation of scientific evidence are they basing this almost comical, yet somehow still unnerving claim? In fact, it is based on an observational study of some 2300 people from the "offspring cohort" (siblings) of the original Framingham Study. The researchers looked at the body mass indices of individuals that these cohort offspring indicated on their original administrative tracking sheets for the Framingham study as people they would recommend to contact (otherwise known as friends and families) to help facilitate follow up of the study. They then compared BMI changes measured in these individuals during numerous three year periods between 1971 and 2003 with BMI changes of the people in the cohort. What were the startling findings? An individual's chances of becoming obese were 57% greater if they had a friend who became obese over a certain period of time. And what was the authors' conclusion?
"It's not that obese or non-obese people simply find other similar people to hang out with, rather, there is a direct causal relationship".
How does one conclude a direct causal relationship from an observational study? Bald men are more likely than men with a full head of hair to have a heart attack. Can we conclude from this that they should buy a toupee or begin using Rogaine lotion to lower their risk? And what about that little nasty episode with hormone replacement therapy just a few years ago? Remember when women were told that since there was much more heart disease observed after menopause the reason must be the loss of hormones - and that therefore hormone replacement therapy must be the answer? Oops, wrong again. The bottom line is that what you get from observational studies like this one are hypotheses which then must be validated by research that actually implements experimental interventions to prove or disprove them.