Monday, April 16, 2007

Do Newspaper Recipes Make you Fat?

My media scanners picked up this headline on Friday,

"Community Obesity Rates Linked To Calories From Newspaper Dessert Recipes"
and of course, I immediately closed the Citizen's food section.

All kidding aside, what the headline did do was spur me to read the actual article published in the Wisconsin Medical Journal entitled:
"Calories from Newspaper Dessert Recipes are Associated with Community Obesity Rates"
After reading the methodology and the paper, I can say that without a doubt this is the weakest article I have ever read that was actually published in a peer-reviewed medical journal.

Here are their methods:
"3 cities with populations of 400,000 or more were selected from 4 geographic areas within the United States....All recipes published in major newspapers for those cities in the last week of August 2000 were accessed....and the nutrient content for each recipe was calculated....data were entered into an Excel spreadsheet and analyzed with SPSS using linear regression and correlation."
So basically a bunch of large cities had their obesity rates plotted against one week's worth of local newspaper recipes' Caloric contents.

The authors concluded,
"In this study, we found that caloric intake of recipes published in major newspapers are correlated with community obesity rates."
The authors do admit that there are limitations to the study,
"Limitations of this study stem primarily from the correlational study design that may lead to the "ecological fallacy" where the observed correlation may be due to some underlying differences between the communities included in the study, such as demographics or health care"
Um, yah think?

Two things blow me away about this study. Firstly that any self-respecting scientist would publish correlational data based on a SINGLE week's worth of data boiling down an incredibly complex social and medial condition with dozens, if not hundreds of possible hypothetical contributors, to have cause from a single observational variable, and secondly that somehow this thing passed peer review.

[note to self, if I ever have a poorly designed study that I can't get published elsewhere maybe consider the Wisconsin Medical Journal]

This study reminds me of a story published a few months ago in the Economist where Toronto based scientist Peter Austin "proved" that those burn under the astrological sign of Sagittarius were 38% more likely to be admitted to a hospital with a broken arm. The difference between Austin's ridiculous conclusion and these authors' being that Dr. Austin purposely designed his study to promote a ridiculous result so as to call to question conclusions like the one above and highlight the need for proper study methodologies and statistical analyses.

Statistics can be incredibly misleading.