You may have seen the study – recently reported on in the Washington Post (they get interested in population/cohort health studies a couple of times a year) – that says that a meat-rich, higher-protein diet in middle age is associated with greater cancer risk.
If you already believe that eating more meat is bad for you, then a study like this is a reassuring reaffirmation to adopt a certain notion of a better diet – more whole-grains, I imagine, with fruit, vegetables, and maybe some soy protein (or at the worst, lean meat).
Sure, there are dissenting voices. A professor at NYU essentially says that this latest study has weak and contradictory findings, and ought be discarded.
I know my biases. I eat a lot of meat, and think the the Inuit have an attractive diet. I look at recent meta-analyses on saturated fat consumption and feel vindicated, and I see no paradox in the French Paradox. I’ve got my beliefs. I think they are well founded.
This shapes how I see people who say eating meat is bad. I see it as an attack on something I hold dear – a personal conviction, like a religious belief. But – for me – that conviction feels stronger than a religious belief because my belief is based on “science”, or at least the little slivers of science I choose to pay attention to. With them (vegans, PCRM, the Nurses Health Study), I see meat-animus, confirmation bias, and motivated reasoning.
And I see those things in the study we are discussing here.
As a motivated reader, how long does it take to find something that permits me to discount this study? A few minutes.
We notice that the initial run of this data didn’t produce the result the author’s expected. If you just start by asking your self “why such a contorted headline?” you don’t have to wait long for an answer – in the initial data run, high and moderate protein consumption was not associated with all-cause mortality, or heart disease, or cancer.
So the authors mined the data for the result they wanted.
And even here they faced a terrible problem with their data – by isolating the age sub-group (50-65) that did show an asssociation between increased cancer mortality with increased protein consumption, they were stuck with the left-overs (ages 66+) in a data-set that, remember, showed no overall association between protein intake and mortality. And that set up the following problem for the overall hypothesis – low protein diets look like a death sentence for those over age 66.
That’s a crazy result, like finding strong evidence in your data that cigarette smoking from age 50-65 is really bad for you, but after age 66 cigarette smoking is really good for you. It is the kind of thing that might make you wonder about your data, or your hypothesis, or what it would take to make you consider the possibility that a belief you hold dear might actually be wrong.