A commenter today made me remember yet another reason that observational studies such as this one are never to be used to determine causality that I didn’t mention in the original post. It’s important enough that I want to post it here so that those who don’t read the comments and/or my responses to them will read it as well. Here is my response to the comment:
Another big problem with these studies in which researchers arbitrarily place certain foods in certain categories then use these categories to determine which is best is that there is room for a ton of hanky panky. For example, in the colon cancer study fruit juice ended up in the prudent dietary pattern category. Why? Because the researchers volitionally put it there. Many studies have shown fruit juice to be little better than sugary sodas in terms of what they do to one’s metabolism, so fruit juice could just as easily have been placed in the Western dietary pattern category. Ditto for a number of the other foods categorized. What this means is that the researchers could have (and I’m not making this accusation; I’m simply pointing out the possibility) fiddled with the data to get it to come out how they wanted it to come out. Let’s say that fruit juice was first in the Western dietary pattern, and when it was there the overall outcome changed. There was no difference in rates of death between the two categories. So one simply moves fruit juice to the other category and runs the program again, and Viola! there are more deaths in the Western pattern. With all the variables from which to chose, it’s easy to manipulate these kinds of studies to get the results one wants.