7 The Journalist Gary Taubes 7: Anomaly Hunter 1
Anomaly hunting, as we learn here from Steven Novella, lies at the heart of pseudoscience. Anomaly hunting is simply the practice of seeking and collecting apparent inconsistencies between our observations of the world and our scientific understanding of the world. It's a good thing to draw attention to the odd and unexpected in the pursuit of knowledge, but that’s not how pseudoscientists use anomalies. They use them to sow doubt and confusion. They use them to try to discredit legitimate research. They use them to fabricate false claims in support of their preconceived notions.
Good scientists are drawn to the quirky and unexpected in their efforts to better understand our world. The natural world is sufficiently complex that we should expect to find apparent anomalies in it, but Dr. Novella points out that the only true anomaly is one that cannot be explained even after serious scrutiny.
Gary Taubes is an anomaly hunter of the pseudoscientific variety. In his book, Good Calories, Bad Calories, he presents the reader with a litany of presumed exceptions to the diet-heart idea, exceptions that he says should have undermined diet-heart long ago. Taubes never applied serious scrutiny to his anomalies, however. He is not interested in providing himself or his readers with a deeper understanding of the science of diet-heart. Instead, he uses his fake anomalies to induce the intellectually passive reader to adopt his irrational cynicism toward mainstream nutrition science.
Taubes pretentiously quotes Francis Bacon as he lectures us on the perils of our common tendency to only seek out and retain the evidence that supports our opinions while shutting out the evidence that challenges our opinions. It is clear he believes that Ancel Keys and the other pioneers of diet-heart succumbed to their inclinations toward lazy, self-serving thinking. The hard-nosed journalist Gary Taubes is doing the serious critical thinking they didn’t have the integrity or intelligence to do themselves. Ironically, in his attempt to expose them, Taubes has provided us with an astonishingly mindless demonstration of unfettered motivated reasoning. Thankfully, he has concisely and clearly demonstrated his knack for self-deception in print for us all to see.
Make up your own mind as we proceed as to whether Taubes is in any position to accuse others of being prone to confirmation bias.
Before we move through the many examples of misleading anomalies Taubes has utilized, I think it is useful to recall the techniques used by the cigarette companies during their campaign to confuse the public about the harm caused by their products. They, too, used minor inconsistencies to create an illusion of scientific doubt. A very helpful viewer of my videos named Richard Arppe recommended that I read Merchants of Doubt by Oreskes and Conway to engage that history. The deceptive tactics of that industry have resonance today. One question the defenders of smoking would have asked was, “Why do cancer rates vary greatly between cities even when smoking rates are similar?” Does the logic of that question seem familiar, perhaps recalling the distractions provided by the Masai or the so-called “French Paradox”? The answer to this particularly disingenuous question is that like heart disease, cancer is affected by other factors beside smoking, and these factors differ between cities. I’ll discuss this more in my video called The Cause of Heart Disease. But think about the purpose of that question. Its purpose was to cause the lay person to doubt the public health professionals who cautioned against cigarette smoking. Should such a question cause you to doubt the harm of cigarette smoking even if it is based on a real fact? Now that you know that smoking behavior did not correlate perfectly with cancer rates in some places, do you conclude therefore that smoking is safe? I doubt it. I think you already understand that life is a little more complicated than that.
Taubes avails himself of population-level anomaly hunting of that nature in this paragraph. Do you see how he lists a whole bunch of locations in one sentence? Those are his anomalies. I’m going to look at each one of those so that you can see just how much deception this man can cram into a single sentence. Pause if you like to read the next sentence starting, “Proponents of Keys’s theory…” In that passage he admits that more could be learned by comparing populations with contrasting dietary patterns as opposed to populations with similar dietary patterns. “This might be true,” he says in the last sentence. Actually, there is no "might" about it, Mr. Taubes. If everyone in a group is eating the same way and if we learn that those people are having different health outcomes, those differences would have to be explained by factors other than diet. Is that not obvious? Taubes, tries to inject doubt into this basic principle by saying, “but perhaps fat just wasn’t the relevant factor.” Why wouldn’t fat be relevant, Mr Taubes? Did you write your book about nutrition to tell us how diet isn’t a relevant factor? How are you going to support this “perhaps” of yours? I think he would have been more straightforward had he written, “but perhaps I will believe whatever I want to believe no matter the evidence.”
He had previously given us a list of anomalies that he says, “failed to reveal any correlation between fat consumed and either cholesterol levels or heart disease.” Wait, Mr Taubes, are you saying that you think there is no correlation between dietary fat and cholesterol levels?
That’s not what you said on page 20! Saturated fats tend to raise cholesterol, you said. This is an unusually accurate statement for Taubes, but it’s not quite right to say saturated fats “tend to” raise cholesterol. They just do, straight up.
And yet, this is exactly what he wants us to call into question with these anomalies. He just won’t say so directly because he knows that that would be an obvious lie. Taubes intends to insinuate into the minds of his casual readers that his few weak examples refute diet-heart. He finds it convenient to just list these and leave them with a false impression. But I am not going to let him get away with that. We’re going to look at these studies right now. As I do this, I want you to contemplate how little effort it took Taubes to write that one sentence, and how much work I had to do to fact-check him and bring the truth to you. Bear in mind I am not being paid for my efforts by anyone.
The first anomaly on his list is Framingham. He is referring here to what he had written on his previous page…
In this paragraph here. If you look at his references, he is basing this paragraph on his interview of someone named Tavia Gordon. There is no way we can independently verify what Tavia Gordon said to him. We have to take Taubes’ word for it that these are Gordon’s thoughts, which I’m not inclined to do. The statement is made here that in research conducted in Framingham, the amount or type of fat had no effect on cholesterol levels.
This is not what Taubes said, is it? He said saturated fats raise cholesterol. What is his point? That Framingham overturned the results of metabolic ward experiments somehow? Viewers of my videos will know right away that a likely explanation for this comment by Tavia Gordon…
Is the effect of declining cholesterol along with ill health in old age. Chances are those observations about fat and cholesterol were not properly age-corrected. I’ll have a video later on in this series about declining cholesterol in old age, but for now, here is an explanation for this effect directly applied to Framingham. It was observed in Framingham that falling cholesterol levels were “perhaps due to diseases predisposing to death.” One may counter that my reference doesn’t apply perfectly to what Taubes said, but bear in mind, I don’t have any reference to work with from Taubes. I only have that short quote. This is the closest match to it that I could find.
These couple paragraphs may explain whatever Tavia Gordon saw. Cholesterol increases in men until they hit their late 50s, and then it starts to decline. What is important to understand is that Framingham showed that high LDL is associated with an increased risk of heart disease adjusted for age up to 82 years old.
High cholesterol is a risk factor for heart disease in the elderly when comorbidities that make an older person frail and lower their cholesterol are factored into the picture. This paper should have settled this question back in 1997. These authors dealt with this in definitive fashion, saying that the interpretation of cholesterol scores in the elderly is complicated and requires clinical judgment, and good judgment is not something that the cholesterol deniers have in abundance.
Realize that the men in Framingham were pretty much all consuming diets high in animal food. They weren’t going to be able to learn much about what the effects of animal foods were in their participants if there wasn’t much differentiation in their diets.
But in this case we don’t need to explore too many nuances to understand it. It is clear that Taubes can’t get his story straight. Mr Taubes, do you really mean to say that Framingham proved that saturated fat doesn’t raise cholesterol? If so, I have a suggestion for you from someone you know.
TAUBES: You do not over-interpret your data. You do not say your studies say something they do not say. And do that is bad science.
Since Gary Taubes is quoting Tavia Gordon, and since I have no idea what data is being referenced by him, the best I can do to find out what Gordon really thinks is to read what he has actually written on this subject.
Before I show you that, let’s first look at Mr Gordon’s background. He studied anthropology and mathematics. He was employed in medical research as a statistician. He was not particularly well-qualified to make pronouncements on this topic. This is not a criticism, just an observation.
He tells us that being a biostatistician is a fun career because it allows someone to enter a field of study “absolutely cold,” as he says. He would join a project in a particular field with no foundational knowledge. “I didn’t know a thing about them, except what I read in the newspaper,” he recounted. Again, this is not a criticism of the man, but this is something we must consider if we are interested in his thoughts on this topic.
Given his lack of background knowledge, what has he had to say in print about diet-heart? First, you can see here that he eventually came to be persuaded that total cholesterol and LDL cholesterol were correlated to coronary heart disease. Gary Taubes didn’t tell you that. Mr. Taubes, do you agree with your reference, Mr. Gordon, that high LDL associates with heart disease?
On the left, you see here Gordon was writing about Framingham. He said he found that even in older individuals, LDL was surprisingly strongly associated with heart disease. On the right you can see these results from Framingham spelled out more clearly. LDL was a marker of risk for both older and younger individuals. Tavia Gordon has published work that supports the lipid hypothesis. Does Gary Taubes dispute Gordon’s work? If so, why does he supply a quote from him? Shouldn't they be in disagreement? Why does he leave us with a false impression of the man’s ideas?
An even better question for Mr Taubes might be this: Is there a controlled trial out there demonstrating that older people can protect themselves from heart disease by eating lots of animal fat in an effort to raise their cholesterol? I know the answer. There is no such study. If there were Taubes would tell you about it. It's one of those huge blanks in the low carber philosophy that exposes their illogic and hypocrisy. By the way, notice at the top, he mentions Eisenhower and his age and failing health as explanations for his lower cholesterol. Even Taubes doesn’t find that paradox I showed you in video 2 to be so paradoxical. Read the rest of this excerpt and you’ll see he suggests women should get a green light to eat fatty food because cholesterol had no predictive value for the women in Framingham. Ponder that idea for its underlying logic. Is Taubes once again suggesting that fatty foods raise cholesterol? Why doesn’t he take a position and stick to it?
Now that I’ve sidetracked about women and their risk factors, I’ll just point out that I have made a video covering this already. Women have about a ten-year advantage on men for heart disease risk. Their heart health is more vulnerable to the effects of diabetes, and diabetics tend to have a characteristic lipid profile. In their heart disease risk as in other aspects of life, women are hard to figure out, and Taubes is trying to take advantage of that fact.
The next anomaly on Mr Taubes’ list is Puerto Rico.
This is the study he picked to justify the Puerto Rico anomaly, and this must be the particular paragraph in it that got his attention. The authors tell us that “when an absolute measure of fat in grams is used, there is no association between fat intake and CHD incidence.” They say the observed values would suggest an opposite, inverse relationship. So it seems fat affected incident heart disease in the opposite direction predicted by the lipid hypothesis. It seems we have an anomaly on our hands.
Maybe Mr Taubes became too excited when he saw that to read just a bit further. Had he kept reading, in the next paragraph he would have learned that sugar also failed to show any association with heart disease, which contradicts the views of low carb hero John Yudkin. Didn’t Taubes notice that? Or did he not want you to learn about that? Either way, that’s some poor journalism.
The lack of any observed effect of dietary sucrose was mentioned right there in the abstract, along with another very interesting observation. Those who did suffer major adverse events from heart disease were eating fewer starches. This study suggests that starches are protective. Mr Taubes seems to have gambled that no one would even bother to look up the abstract to this one when he stuck it in his book.
So why didn’t they find an association between saturated fat and heart disease? Why didn’t they find a strong relationship between saturated fat and cholesterol? The answer is that this wasn’t a great opportunity to find those associations. The population as a whole didn’t suffer much heart disease and their diets didn’t vary that much. Therefore any effects caused by fat would have been harder to detect.
As you know, Taubes said “this might have been true” in response to the criticism that there wasn’t much dietary differentiation in some of these studies. This was clearly the case here.
An additional problem was that they based their nutrition data on imprecise methods. This was no randomized controlled trial, Mr Taubes. This was just the dietary recall method you criticize whenever it suits you...
Like here, on page 120. You say you consider dietary recall surveys to be flawed evidence. Is this study not then flawed evidence by your own reckoning? Are you trying to have it both ways here?
Also notice, their cholesterol levels weren’t all that high. This is yet another problem with this one.
But despite all that, this study still should have given Mr Taubes reason to question his dogmatic beliefs about carbs. The authors wrote, “The above results focus on the beneficial role of increased carbohydrates.” Why did he ignore this? Is he trying to take advantage of your ignorance? Only he can say.
Next on his list of anomalies is Honolulu.
Taubes uses two references for Honolulu and here is the first. This is the whole abstract. It tells us that men who suffered from coronary heart disease there had higher intakes of protein, fat, and saturated fat than those with lower intakes. Yes, this is the reference Taubes chose. I didn’t pick it. It also says that men who did not suffer from heart disease consumed more carbs, starch, and vegetable protein. If you think it’s odd that Taubes used this one in his book, the explanation is that he selectively focused on only the rest of the abstract. First, men who consumed more polyunsaturated fatty acids suffered more heart disease, which runs contrary to expectations.
Specifically, polyunsaturated fats associated with more angina and coronary insufficiency. You know I am no defender of refined oils, but notice their caveat. These are not hard endpoints so these data were soft.
Second, Taubes probably liked that the authors said that many nutrient categories, including saturated fat, were no longer found to be significantly associated with incidence after multivariate analysis. But look at my middle highlight. One of the variables factored out in their multivariate analysis was serum cholesterol! It makes sense that the effects of saturated fat would be less apparent if they chose to systematically remove the number one observable effect of saturated fat consumption, don’t you think? I have the same comment about the other dietary factors that were said here to be no longer related to incidence after adjustment for cholesterol. Carbs and vegetable protein would have associated with lower cholesterol. That’s our second problem in this study.
A third problem is explained by the authors who said that the people in this one didn’t have high levels of heart disease or cholesterol. Compared to Americans on the mainland, they were healthier and had better risk factors. In this regard, this one is similar to the [Puerto Rico] study.
Similarly, the reason associations with cholesterol were less strong among Japanese-Americans, they said, was because they generally have low cholesterol. That explanation is right there in this paper.
Here is another paper that makes this point obvious. Pause and compare the cholesterol levels of Japanese men living in Japan, Hawaii, and California years ago. Californians had mean serum cholesterol about 43 points higher than their counterparts in Japan. Men in Hawaii were somewhere in between.
No, cholesterol did not precisely parallel heart disease in this study, but there are other risk factors for heart disease. No one argues otherwise.
I included their graph so you can see how their participants compared another way.
Let’s not gloss over their unadjusted findings. This paragraph pretty much contradicts everything Mr Taubes stands for. Carbs, starch, and vegetable protein associated with less heart disease. Saturated fat and protein in general associated with more.
The authors found that serum cholesterol associated strongly with total coronary heart disease and non-fatal heart attacks. This study was not much of an anomaly for diet-heart, was it?
This is the second reference he chose for Honolulu. This one also is an odd choice. It states that carbs were negatively associated with heart disease among Japanese men in Hawaii.
Here this is stated more clearly. “Total carbohydrate intake was the only additional diet variable that was significantly related (negatively) to CHD death.” Did he not read this? Did this not penetrate his bubble?
Taubes may have chosen this one because they also said they only found a “weak relation between diet and serum cholesterol.” This statement is bounded above and below with cautions against over-interpreting this data. Is Taubes once again trying to suggest that diet doesn’t influence cholesterol, even when he doesn’t believe that himself? On the right you see that the authors do not over-interpret this finding. Taubes isn’t opposed to over-interpreting data when it’s convenient for him.
What might have been interesting would have been if something like high cholesterol were negatively correlated with heart disease, but that’s not what happened here. They just couldn’t pull a strong signal out of this data. This study didn’t really contradict anything except for Mr Taubes’ ignorant, destructive beliefs.
Taubes gave me plenty more fake anomalies to pick over. The fun continues in the next video.