Butter Bad, Seed Oils Good, Says New Study

Today a paper which proves my contention that governments ought to get out of the grant-making business. The “study” was funded by NIH, and conducted by “top” scientists at Harvard. Its conclusion, that replacing butter with seed oils causes you to live longer, is absurd. But I’ll have to take time to demonstrate that to you. And others, also working on the government dime, will have to deal with it, too, spending valuable resources sifting through it. It would have been better had this study never been funded.

Government grants cost more than the give.

The peer-reviewed NIH-grant paper is “Butter and Plant-Based Oils Intake and Mortality” by Zhang and a host of others in JAMA Internal Medicine.

Now I have already warned you about the sins and excesses of epidemiology (here, here, and here, in that order). Everything that I showed you that can go wrong in epidemiology has gone wrong (mostly) in this paper. Take time to review those lectures. Anybody can read them. Don’t look at another epidemiology paper until you do.

Here’s today’s “discovery”:

Findings  In this cohort study of 221,054 adults from 3 large cohorts, higher butter intake was associated with increased total and cancer mortality, while higher intake of plant-based oils was associated with lower total, cancer, and cardiovascular disease mortality.

Meaning  Substituting butter with plant-based oils, particularly olive, soybean, and canola oils, may confer substantial benefits for preventing premature deaths.

To “discover” this they scoured through questionnaires, on which, among other things, people were asked twice yearly to remember how much butter or oil they ate and in what forms. How much did you eat last November? Can you remember? Be specific. Be exact—and by exact I mean to the nearest fraction of a teaspoon. We’re doing science here.

Worse, the questionnaires when asked twice of the same people showed that people didn’t answer the same; roughly half the time they answered differently. No surprise, given the fallibility of (eating) memory. But it means a lot of error was ignored in our authors’ model, as we’ll see.

Worse again, this (my emphasis):

“The intake of plant-based oils (corn, safflower, soybean, canola, and olive) was ESTIMATED based on the reported oil brand and type of fat used for various cooking methods, including frying, sautéing, baking, and salad dressing, and all of the food composition data for calculating oil intakes were updated EVERY 4 YEARS.“

Worse again again, the uncertainty in these estimations, like in the food recall in the questionnaires, was ignored. (In the lectures, I call the claim that “X causes Y” but where X wasn’t measured, and a guessed substituted, and the whole “proved” with wee p-values, the epidemiologist fallacy.)

Worse again again again, and maybe worst, those with “with a history of CVD, diabetes, or cancer” at the start of the study were excluded from the data! Could any of those excluded people have got sick, especially diabetes, from porking out on seed oils? The end of the study did include those who died of these diseases, and therefore must have developed any of these during the study. Could a tiny pat of butter, as we’ll see, have caused cancer and killed that quickly?

They also “stopped updating dietary data after diagnoses of diabetes, stroke, or cancer”. To avoid, they say, “reverse causation”, because some might have changed diets after diagnosis. What about the idea of simply tracking who changed and who didn’t? Rejected.

They divided data into four eating-amount groups because, hey, why not. In this case (according to the Supplemental data) groups were separated by about 4 measly grams of butter a day. That’s a bit more than half a teaspoon of butter. The groups were: about zero grams to those who ate the most at 13.1 grams of butter a day. That’s two and three quarters teaspoons. ESTIMATED, you recall.

There is no way to know who had how much butter or oil to this degree of accuracy, at all, not from those questionable questionnaires, and not from the model based on estimating butter use.

Finally came the unnecessarily complex models with wee P-values. Those following the Class know that any use of P-values to make any decision or judgement about a conclusion, such as bit o’ butter can give you cancer, is a formal fallacy. What’s curious is that every scientist is taught this. But none ever remember it. Not when it is their wee P-value, which they use to “prove” cause.

Or “association”.

The authors concluded “Replacing 10 [gram per day] of total butter intake by an equivalent amount of total plant-based oil was associated with an estimated 17% reduction in total mortality”. And really just cancer, because they “found” no difference in heart disease deaths.

Recall 10 grams of butter a day is about 2 teaspoons. There was no way they could know this to that accuracy. And that reduction is of an already very low estimated chance of crapping out “instantaneously”, because that “instantaneous probability of death at any moment” is what the models were.

Is butter causing cancer? Are seed oils? You cannot tell, not from this paper. They missed the opportunity, as many of these papers do, of displaying the data graphically, so that all could see how much variability there was. Simple plots, maybe by age and sex, of the distribution of butter and oil ate from those who lived and died. Always be suspicious of papers, especially epidemiology, that don’t show you the data.

Subscribe or donate to support this site and its wholly independent host using credit card click here. Or use the paid subscription at Substack. Cash App: \$WilliamMBriggs. For Zelle, use my email: matt@wmbriggs.com, and please include yours so I know who to thank. BUY ME A COFFEE.