How To Trash Your Own Study Results

      43 Comments on How To Trash Your Own Study Results

Suppose you were an idiot. And suppose you were a nutrition scientist. But I repeat myself.

Sorry, just couldn’t resist borrowing from Mark Twain. Let me try again.

Suppose you’re a nutrition scientist. And suppose you conduct a study, all the while expecting the results to support a hypothesis you already believe.  But then  — @#$%!! the results undermine the hypothesis.

Oh, dear, what to do?

We saw one way to handle that sticky situation in our last episode: just don’t publish the results. Ancel Keys conducted a clinical study in which people who consumed vegetable oils instead of arterycloggingsaturatedfats!! had higher mortality rates – including higher mortality from heart disease. Keys didn’t like that result, so the data gathered dust for 40 years.  Nice move, Ancel.

Here’s another way to handle results you don’t like: explain them away. I found a couple examples of that method while looking through my database of studies recently.

In one study, researchers looked for a link between consuming dairy fat and heart disease. Here are some quotes from a Brown University press release:

Dairy products can be high in harmful saturated fat but not necessarily in risk to the heart.

Okay, let’s stop right there. Take a moment and wrap your head around that sentence. Saturated fats are harmful. Why? Because according to the Lipid Hypothesis, they cause heart disease. But they don’t necessarily pose a risk to the heart.

A newly published analysis of thousands of adults in Costa Rica found that their levels of dairy consumption had nothing to do statistically with their risk of a heart attack.

To conduct the study, [researchers] Aslibekyan and Baylin analyzed data on 3,630 middle-aged Costa Rican men and women who participated in an epidemiological study between 1994 and 2004 by co-author Hannia Campos of the Harvard School of Public Health.

Ah, so it was an observational study. Perhaps the researchers would be justified in explaining away their own results.

The researchers looked not only at the subjects’ self-reported dairy intake, but also at measurements of dairy fat biomarkers, namely 15:0 and 17:0, in their bodies.

Whoops. Not just an observational study. They actually measured biomarkers that told them how much dairy fat the participants consumed. Let’s look at the results.

What they found is that the dairy intake of people who had heart attacks was not statistically different than the intake of people who did not. After breaking people into quintiles, based on their dairy consumption amount, there was no significant linear relationship between consumption and heart risk, even among the most voracious consumers. The highest consumption quintile consumed an average of 593 grams of dairy foods a day.

Once again, stop and wrap your head around that. We’ve been told for decades to stop consuming cream and butter because the saturated fats cause heart disease. But in this study of more than 3,600 people – a study in which researchers directly measured biomarkers of dairy fat consumption – there was no relationship between consuming those arterycloggingsaturatedfats!! and heart disease.

So how do we explain this result?

Rather than suggesting that the saturated fats in dairy products are harmless, Aslibekyan and co-author Ana Baylin, an adjunct assistant professor of community health at Brown, hypothesize that other nutrients in dairy products are protective against heart disease, for all but perhaps the highest dairy consumption quintile in their study. The potentially beneficial nutrients include calcium, vitamin D, potassium, magnesium and conjugated linoleic acid (CLA).

Well, there you have it: saturated fats cause heart disease, ya see, but people who consume a lot of saturated dairy fats — even the most voracious consumers — don’t have higher rates of heart disease because … uh … because something else in the dairy products is protecting their hearts!

Awesome. Coming soon: nutrition scientists decide there must be heart-protective substances in bacon, eggs and sausage that offset the heart-killing effects of the arterycloggingsaturatedfats!!

Here’s another example of explaining away embarrassing results. In a 2010 study, researchers randomly assigned subjects to one of three groups: no change (the control group), 60 grams of whole grains for 16 weeks, or 60 grams of whole grains for 8 weeks followed by 120 grams of whole grains for 8 weeks.

These are the markers of cardiovascular health the researchers measured, according to the abstract:

BMI, percentage body fat, waist circumference; fasting plasma lipid profile, glucose and insulin; and indicators of inflammatory, coagulation, and endothelial function.

If whole grains are the wunnerful, wunnerful, health-enhancing food we’ve all been told they are, the second group should have shown improvement those health markers, and the third group should be well on its way to immortality. So let’s check the results:

Although reported WG intake was significantly increased among intervention groups, and demonstrated good participant compliance, there were no significant differences in any markers of CVD risk between groups.

No significant difference in any markers. That would mean not one.

Oh dear … and here we have our beloved USDA telling us all we need to eat more whole grains. How do we explain this result? Here’s how:

A period of 4 months may be insufficient to change the lifelong disease trajectory associated with CVD.

Riiiiiiight. Four months of consuming large servings of whole grains doesn’t affect BMI, percentage body fat, waist circumference, fasting plasma lipid profile, glucose, insulin, or indicators of inflammatory, coagulation, or endothelial function. But those whole grains may still have a positive effect on the lifelong disease trajectory associated with CVD.

Yeah, that makes perfect sense.

Well, at least these studies were published. If Ancel Keys had conducted them, the data would still be sitting in a dusty attic somewhere.


43 thoughts on “How To Trash Your Own Study Results

  1. Jim Frake

    Tom, please refrain from referring to Keys as a scientist because to qualify, use of the scientic method is required. From Wikipedia,

    “To be termed scientific, a method of inquiry is commonly based on empirical or measurable evidence subject to specific principles of reasoning.”

    I’m generally not a fan of MORE rules and regulations. However, IMO we need addition to the Penal Code for those that spend tax payer dollars to fund studies and then either obscure, manipilate or bury the results.

      1. Bob Niland

        Anyone tempted to assume that scientific papers are credible needs only follow for a while. Sometimes the data isn’t being hidden – it just never existed:

        But even when a trial was actually run, it’s very dangerous to be a black swan* in far too many labs: you could get murdered, tossed in the data dungeon, or simply painted white.

        On the matter of tax-supported research, the scandal is even wider than it appears. It appears that ⅓ of trials never publish at all, and another ⅓ take more than 2 years to do so.

        Then they often publish in pay-walled journals, which might cause the concerned citizen to ask: why do I have to pay for these results twice?

        Then they hide the raw data, so no one can perform any authentic peer review. Does anyone suppose that if Ancel were still alive, that MCE data would have been available for re-analysis?
        * Black swan: a data point that falsifies a hypothesis (under the allegory, the hypothesis that all swans are white – 1 black swan suffices for falsification).

        1. Tom Naughton Post author

          I’m a libertarian, but if researchers are taking tax dollars, I’m all in favor of requiring them to publish all results, including the raw data. They shouldn’t get to take my money for research and then decide they own the data.

          1. Thomas E.

            The US used to be really good at that. Way back 20 (or more) years ago, a friend of mine was a research assistant for GIS data processing/collection (Geographical data) in Ontario, Canada. And even in Canada, using US based data was as simple as paying for the media and the data would be sent to you. I think it was like 10 or $20 for the CD set.

            Where as in Canada, in her federal government, there were no controls on the product based on the tax payer subsidies. So what wound up happening, is every research project across the nation, would collect the data it needed, and charge a premium if anyone wanted it. Almost without exception is was cheaper to have undergrads collect new data that was good enough for the research, so you had dozens, if not hundreds of data sets that were just good enough for the purpose they were collected, and one or 2 great data sets that no one could afford to purchase.

    1. Tom Welsh

      Jim, I think we should distinguish between two common meanings of the noun “scientists”.

      1. Someone who believes whole-heartedly in the scientific method, and is dedicated to the search for objective truth for its own sake.

      2. Someone who has a PhD in a “science” subject, and intends to do well in life – have a big house, wear expensive clothes, drive a big car, travel to exotic places (preferably at government expense)… and do lots of other things at government or corporate expense.

      The former would rather poison himself than fake or hide results. The latter… well, he is really more interested in money and success, and somehow it turns out that the disinterested pursuit of truth doesn’t always maximize your revenue or your reputation. (Just look at Socrates and Jesus, for a start. Not to mention Ignaz Semmelweiss).

      1. Mike (another one)

        There is also that category of scientists who do try to publish meaningful data, but their experiments are flawed because they have to study what they study indirectly, because the funders of their research won’t fund direct study.

      2. Walter Bushell

        Well, for example, Mr. Keys was a scientist, sometimes. He did recant on dietary cholesterol, for example, but the sickness care industry ignored him. He was hoist by his own petard, as it were.

  2. Desmond

    “No one trusts a model except the man who wrote it; everyone trusts an observation, except the man who made it.” — Harlow Shapley

    This may hold true for astronomers. Unfortunately, only the second half seems to apply to nutritional science, and only if it conflicts with the model.

    1. Tom Naughton Post author

      I’m reminded of U.S. Grant’s reply to a comment about Senator Charles Sumner, who was known for his huge ego:

      “Senator Sumner doesn’t believe in the Bible.”

      “Well, I’m not surprised. He didn’t write it.”

      1. Bob Niland

        Speaking of paradox, Credit Suisse is focusing on nutrition again, with this recent hour-long video discussion with Lustig and Attia, linked from DietDoctor:

        In the introduction, CS moderator Stefano Natella makes it clear that he recognizes that scientists who use the word paradox are highly suspect, as they usually use it as an excuse to exclude irksome data.

        He didn’t put it this way, but it might be useful to think of it as:
        Paradox is often a euphemism for “black swan denier”.

        I might further add that the video stream has almost no useful visual content (the videographer rarely caught the slides they were using), so it might make good lawnmowing headphone fare. You might be tempted to take notes when Peter discusses his “8 things”, but it appears he skipped #5, so I’m going to see if I can find the complete list elsewhere.

  3. Firebird

    Replace “Nutrition” with “Global warm, er, Climate Change” and you get the same result.

    “Cold cause warmth”

    “Just because it is cool where you are.”

    “Sea levels rose 2′ in 2500 years. We’re doomed!”

  4. Justin

    Are you planning on releasing your database, or a software or book version of it, at some point? I also do that kind of stuff regularly to keep track of things for my own personal sanity, but in your case, it seems like something that could really help people, at the very least, continue to convince themselves that they’re not killing themselves, contrary to everything they hear on a regular basis.

  5. Dianne

    Tom, I wish that every high school student had to watch your “Science for Smart People” video and read “The Art of Clear Thinking” by Rudolph Flesch. It wouldn’t hurt for them to do the same as college freshmen and again a couple of years later. If enough average citizens had a basic knowledge of how to sort out fact from fiction, and how to spot twisted “reasoning,” the freakin’ liar– er, nutrition scientists — wouldn’t dare try to publish this garbage. Even politicians might have to clean up their acts.

    Nah — THAT ain’t never gonna happen.

    1. Tom Naughton Post author

      Interesting question: which people are less likely to clean up their acts … nutrition scientists or politicians>

      1. Dianne

        Actually, when I consider some of the nonsense my now-very-liberal niece got fed in college, I’m not sure the schools even want their students to be proficient in rational thinking. It all seems to be about emotions and feelings these days — which may help explain some of the mess we’re in.

        1. Tom Naughton Post author

          When colleges set aside “safe spaces” where disagreements and challenges are prohibited, they’re certainly not promoting rational thinking. That’s what happens when left-wing loons take over the faculty.

          1. Firebird

            I just got a warning ban from a sports blogging website. I had to click an agreement before I could continue to the site. They told me that I had no right to criticize a writer or an article. If I didn’t like it, they said I should ignore it and not criticize it.

            My response? Get a back bone, grow a pair and accept my right to free speech. If your writers can’t take criticism for what they write (which they have no problem dishing out) then don’t write the damned article!


            1. Tom Naughton Post author

              Well, the right to free speech and a free press includes the right to decide what content will appear on your site. But yes, they’re weenies.

      2. Mike (another one)

        To be a working nutrition scientist with funded research, you are almost necessarily a politician.

  6. Tom Welsh

    “Well, there you have it: saturated fats cause heart disease, ya see, but people who consume a lot of saturated dairy fats — even the most voracious consumers — don’t have higher rates of heart disease because … uh … because something else in the dairy products is protecting their hearts!”

    Just the nutrition science equivalent of Ptolemy’s famous epicycles. If the observations don’t exactly fit the model… just add a few more epicycles.

    On the other hand, Karl Popper would have strangled them with his bare hands.

    1. gallier2

      Not really. Epicycles were the accepted theory of celestial motion for nearly 2000 years because it fit the observation quite well*. The arterycloggingsatfat hypothesis can not even claim that.

      * it was only when observation quality increased at a suffisant level, with Tycho Brahé, that epicycles started to break down.

  7. Carlos

    “Although reported WG intake was significantly increased among intervention groups, and demonstrated good participant compliance, there were no significant differences in any markers of CVD risk between groups.”

    The obvious explanation is that WG products must contain some substance that counteracts the wunnerful-wunnerful health benefits of whole grains.

      1. Walter Bushell

        Carbohydrates?! Dr. Jason Fung[1] (the fasting guru) states that fiber is the antidote to carbs. So. But it seems to me wouldn’t it be easier to eschew the poison, so that there is no need for the antidote?

        [1] Who I greatly respect and has a great web site at

  8. erdoke

    In the WHOLEheart study they only report results vs. the Control group. Body fat went up 1.0 % for the Intefvention 1 group in 16 weeks (rather big change) and insulin increase might also be significant within the Intervention 2 group. Those holy grains…

    1. Tom Naughton Post author

      Heh-heh … yeah, researchers are big on spotting “trends” when they can’t cite significant results.

  9. Mike (another one)

    Part of the problem is that Ancel Keys believed his claims. He convinced himself that the science was right. It would be difficult to make fraud under those circumstances.

    One aspect that he didn’t believe at the end of his life was the notion that high cholesterol levels came from eating stuff with cholesterol in it, like eggs. He tried to publish research refuting this notion, but he was ignored by the same wall of nonsense that he himself helped build.

  10. Wayne Gage

    After making a hypothesis, a real scientist will search for information that does not support his hypothesis, he will search for information that destroys his hypothesis.

  11. Paul Miller

    Your brand of intelligent and entertaining mockery is so badly needed across so many fraudulent cultural edicts. I hope and believe it will open more than a few people’s eyes. Go Naughtons!

  12. Cameron

    Tom, I picked up “How Not To Be Wrong” after reading about it here. Ellenberg rips apart one study in which the authors projected that every American would be obese by 2048, but black men wouldn’t reach obesity until 2089.

    That’s the most relevant example related to this post, but the book is excellent overall.

  13. Walter Bushell

    Humans are imitation machines rather than thinking machines. Human parents expect to teach and human children expect to be taught. Chimps are better at taking short cuts, than human children. There is a famous experiment showing that. Given a candy dispensing machine the chimps went directly to get the candy, but the human children followed the pattern they were taught.

    This may explain a lot of the behaviors we deplore. The head ape said it , therefore it is true.


Leave a Reply

Your email address will not be published.