Last updated on April 12th, 2017
Life is full of risks — some real and some statistical.
Most of the risks claimed by observational studies fall into the statistical category. By “statistical,” I mean “imaginary.”
For instance, a study published yesterday in JAMA Internal Medicine (online) entitled “Vegetarian Dietary Patterns and the Risk of Colorectal Cancers” suggests that eating a vegetarian diet will reduce a person’s risk of getting colon cancer by over 20% . Or to put it the other way around, regularly eating red meat will increase a person’s odds of getting colon cancer by that alarming percentage. (View the abstract.)
Actually, to be fair to the spirit of the study, I should change the word “will” in the above sentence to “may.” But I’m not going to. I’m allergic to weasel words, and besides, the strong implication of the study and the various news accounts of it is that eating red meat will give you colorectal cancer.
They all stop short of saying you deserve it.
(Let me say that I take the threat of colon cancer seriously — even personally. Early in my adult life –long before I started eating my low-carb, high-fat diet — I discovered I had a tendency to produce colon polyps, which could become cancerous. Thankfully, none ever have, but I go in for a colonoscopy every few years. I urge everyone to consult with their medical professionals about colon cancer screening.)
The study followed a group of 77,659 Seventh-Day Adventists for a little over seven years. A bunch of these folks were either vegetarians or semi-vegetarians, while another bunch ate meat on a regular basis. Now, the news accounts and the study’s own abstract will tell you that after 7.3 years, the sample group as a whole had suffered 490 cases of colorectal cancer. Those easily accessible accounts will also tell you that the vegetarians had a 22% lower risk of developing such cancers relative to the non-vegetarians.
However, for some reason, those easily accessible accounts omit the raw numbers of cancers in each group. To get those numbers, you need to have access to the full study article, which will cost you money if you don’t have the right credentials.
Luckily, I do have the credentials.
Even when you read the study, it takes a little digging to find the raw numbers. I found them in Table 2. They are presented below, along with my calculation of the colorectal cancer rate for each group:
40,367 vegetarians — 252 cancer cases (0.624%)
37,292 non-vegetarians — 238 cancer cases (0.638%)
That’s right. The absolute difference in cancer rates is 0.014% over seven years.
To put it another way, in a group of 1,000 vegetarians, we might expect 6.24 people to get colorectal cancer in a period of seven plus years.
In a group of 1,000 non-vegetarians, we might expect 6.38 people to get colorectal cancer. That’s an additional one-seventh of a person getting cancer out of 1,000 people over a period of 7.3 years. Maybe. I mean, all of this is hypothetical. That’s the best you can get from an observational study like this — hypotheses to be tested in actual scientific experiments.
Yes, the difference is bigger than zero. Is it statistically significant? I have no idea, and, as a practical matter, I’m not sure I care much. It’s a difference. I’ll let you decide what it means. Personally, it’s nothing that would cause me to give up bacon or beef-steak.
My question is, why aren’t these absolute risk numbers in the abstract? Why aren’t they in the media accounts?
Why is the difference in risk stated as 22% and not 0.014%?
I have to say, it seems dishonest to me. (Yes, “seems” is a weasel word, and I’m a bit itchy already.)
However, I have to give the study’s chief author credit. He is quoted as saying, “Our vegetarians not only ate less meat than the non-vegetarians, but also less sweets, snack foods, refined grains and caloric beverages” (CBS News).
That is a frank, honest admission that the diets of the two groups differed in a number of ways, not just in the matter of eating red meat. The problem is usually referred to as one of “confounding variables.”
Why does the study focus only on the supposed red meat association? You have to wonder how many other conclusions a reasonable person could generate about diets and cancer from the same set of data.
Wait — did I say “conclusions”?
I meant hypotheses.
Also see: “Baked, broiled, or deep-fried: how do you like your variables?”
You must be logged in to post a comment.