Monday 23 March 2015

Nina Teicholz's Big Fat Surprise

Nina Teicholz's book The Big Fat Surprise received favourable reviews from the Economist and from former BMJ editor Richard Smith so I decided to read it. I wish I hadn't bothered. Teicholz's thesis in a nutshell is that we have been lied to for years about saturated fat, leading Americans to adopt a low-fat, high-carb diet that has made them obese and diabetic and probably given them cancer. She concludes that we should go back to eating lots of red meat and dairy products like people did in the good old days.

I became suspicious of this book almost immediately when the author nonchalantly dismisses America's increasingly sedentary lifestyle as a factor in the rise of obesity between 1970 and the present day, saying:

These eight words in a parenthetical aside is the only reference to physical activity in the book. It is unreferenced and untrue.

Suspecting that Teicholz might not be fully on top of her brief, I searched out a critique online and found a forensic fisking by Seth Yoder at The Science of Nutrition who makes a compelling case for viewing Tiecholz as a hopelessly biased, cherry-picking plagiarist. More of that in a moment, but first let's return to the basic premise that Americans used to eat lots of fat and now they don't.

Teicholz repeatedly claims that "Since the 1970s, we have successfully ...  reduced the amount of fat we eat from 43 percent to 33 percent of calories or less." Lord knows where she gets the 43 per cent figure from*, but she compounds the error by claiming that this shows that Americans have reduced their fat consumption by 25 per cent. Neither claim is true.

According to the Centers for Disease Control, in the USA between 1970 and 1994, average calorie intake rose in the USA for both men and women (in contrast to the UK) as a result of an increase in carbohydrate intake. As a result, fat as a percentage of total calories fell, from 36.9% to 32.8% for men and from 36.1% to 32.8% for women, but fat consumption fell little, if at all. Indeed, the CDC clearly states: "The decrease in the percentage of kcals from fat during 1971--1991 is attributed to an increase in total kcals consumed; absolute fat intake in grams increased."

As for saturated fat, the CDC notes that between 1970 and 2000, "the percentage of kcals from saturated fat decreased from 13.5% to 10.9% for men and from 13.0% to 11.0% for women." By my calculations, this means that the number of calories men consumed from saturated fat fell from 331 to 285 and the number consumed by women rose from 200 to 206. Hardly a dramatic change but, again, Teicholz refers only to the percentages. She does not mention that the decline for men was tiny, nor that there was no decline at all for women. Nor, indeed, does she mention that the percentage of saturated fat in the American diet is still higher than "less than 10%" recommended in the official recommendations. 

In other words, and contrary to Teicholz's endless assertions, Americans have not "dutifully" followed government guidelines, they do not have a low-fat diet and they certainly do not have a "near-vegetarian diet".

The "near-vegetarian diet" claim, which is made more than a dozen times in The Big Fat Surprise, is so patently ludicrous that one wonders why her editor didn't pull her up on it. Here's a chart of the world's biggest meat-eaters. Looking from the top down, it won't take you long to find the USA...

 On page 116, Teicholz shows US meat consumption since 1909...

So a "near-vegetarian diet" means eating more meat than nearly any other country and eating more meat than Americans have eaten since records began? Teicholz defends this bizarre claim by saying "about half is poultry" (it's actually more like a third if this graph is correct**) and then berates the US Department of Agriculture for stating, perfectly accurately, that meat consumption is at a "record high". She claims that this is "misleading because they lump together red meat and chicken into one category" (p. 116). Yeah, they do: the category of 'meat'. You know why? Because chicken is meat. But even if you think that chicken is a vegetable, it is still clear that Americans are eating more red meat—which Teicholz claims is "virtually banned" in the USA! (p. 5)—than they did for most of the twentieth century. (Her claim that Native Americans ate "a diet of predominantly meat, mainly from buffalo" is also very dubious.)

Since the premise is untrue, the conclusion she draws from it—that Americans suffer from obesity and diabetes because they've been living off celery and mung beans since the 1970s—must also be untrue. But between the premise and the conclusion we have the, er, meat of the argument which revolves around the evidence for the belief that saturated fat causes heart disease. This is fertile ground for a popular science book, which is why several popular science books have already been written about it, notably Gary Taubes' Good Calories, Bad Calories which has given Teicholz a great deal of inspiration, to say the least.

Saturated fat is no longer seen as the singular dietary villain that it once was. It seems clear that it raises levels of 'bad cholesterol' which, in turn, increases the risk of heart disease, but the risk may not be as great as was previously believed. The most recent Cochrane Review on the subject concluded:

The findings are suggestive of a small but potentially important reduction in cardiovascular risk on modification of dietary fat, but not reduction of total fat, in longer trials. Lifestyle advice to all those at risk of cardiovascular disease and to lower risk population groups, should continue to include permanent reduction of dietary saturated fat and partial replacement by unsaturates. The ideal type of unsaturated fat is unclear.

This is not enough for Teicholz, who wants the reader to believe that saturated fat is not a risk factor for anything and should instead be viewed as a disease prophylactic. To switch from one extreme to the other she has to take some astonishing liberties with the evidence. I'm not sufficiently interested in the topic to check Teicholz's references—no casual reader should have to—and so I would have been deceived time and again had it not been for the fact-checking of the aforementioned Seth Yoder. I recommend you read his two blog posts, even if you are not interested in reading Teicholz's book. It delivers a heavy blow to Teicholz's credibility and her lame response to him suggests that she knows she hasn't a leg to stand on. (She promised a point-by-point rebuttal by mid-March but that has yet to materialise.)

The Big Fat Surprise: A critical review part one

The Big Fat Surprise: A critical review part two

Seth has identified many examples of Teicholz borrowing from other people, but especially from Gary Taubes. She not only uses many of the same sources as Taubes (which is often fair enough), but she tends to take the exact same quotes and makes the exact same mistakes as Taubes does in a way that suggests she hasn't even read some of the original sources. For example...
BFS, page 112:
[W]hen Senator McGovern announced his Senate committee’s report, called Dietary Goals, at a press conference in 1977, he expressed a gloomy outlook about where the American diet was heading. “Our diets have changed radically within the past fifty years,” he explained, “with great and often harmful effects on our health.”

The problem here is that Teicholz cites the source of this quote as “Select Committee on Nutrition and Human Needs of the United States Senate, Dietary Goals for the United States (Washington, DC: US Government Printing Office, 1977); 1.” However, this quote does not appear on page 1. It appears on page XIII. Normally I would chalk this up to a simple citation error. The reason I mention it is because Taubes uses the same exact quote on page 10 of GCBC, and also mistakenly cites the source of the quote as being on page 1. I would argue (as I have done previously many times) that this is good evidence that Teicholz is simply lifting sentences from others and simply citing what they cite – likely without ever even seeing the source material.

The accusation of plagiarism does not reflect well on Teicholz, but they do not destroy her argument. However, Seth also gives numerous examples of highly selective quotation. For example...

After discussing the Ornish diet for a bit, Teicholz mentions a paper on page 145 that reviews the evidence for (very) low-fat diets:
Tufts University nutrition professor Alice Lichtenstein and a colleague reviewed the very low-fat diet for the AHA […] Lichtenstein concluded that very low-fat diets “are not beneficial and may be harmful.”
Teicholz both takes the quote out of context and mangles it somewhat. Here’s the actual quote (emphasis mine):
At this time, no health benefits and possible harmful effects can be predicted from adherence to very low fat diets in certain subgroups.
The “in certain subgroups” part is vital to the accuracy of the statement. You can’t just cut it out. And Dr. Lichtenstein’s conclusion is quite a bit more nuanced than Teicholz would have you believe. The paper is actually a very objective look at low-fat diets. It acknowledges that more research needs to be done on these diets in order for a definitive recommendation. Moreover, as alluded to above, it states that until more evidence comes in young children, the elderly, pregnant women, and those with eating disorders should probably avoid the diet. However, it also acknowledges that a low fat diet can be beneficial and there is evidence for that. Here’s an actual quote from the conclusion:
There is overwhelming evidence that reductions in saturated fat, dietary cholesterol, and weight offer the most effective dietary strategies for reducing total cholesterol, LDL-C levels, and cardiovascular risk.

Teicholz also makes statements that are not only unsupported by her own references, but which are often the polar opposite of what her sources say. For example...

 Page 317:
[I]n more than a few major studies, LDL-cholesterol levels were found to be completely uncorrelated with whether people had heart attacks or not.
Let’s take a look at these “major studies” she cites, shall we?

The first is by de Lorgeril et al. I won’t go into detail, but those in the intervention group had fewer heart attacks and also had lower LDL. From the text: “[T]he trend with time was a decrease in total and low density lipoprotein (LDL) cholesterol […]” Although it was not statistically significant, so we’ll give this one to Teicholz.

The second is not a study, but a short commentary by Despres. It argues that we should not focus exclusively on LDL, which is not the same as saying LDL is not correlated with anything.

The third is a statin trial that showed reducing LDL cholesterol also reduced coronary events. In other words, the opposite of Teicholz’s claim. Some choice quotes from the paper:
This trial provides evidence that the use of intensive atorvastatin therapy to reduce LDL cholesterol levels below 100 mg per deciliter is associated with substantial clinical benefit in patients with stable CHD.

Our findings indicate that the quantitative relationship between reduced LDL cholesterol levels and reduced CHD risk demonstrated in prior secondary-prevention trials of statins holds true even at very low levels of LDL cholesterol.
The fourth is a meta-analysis on statins and all-cause mortality. Basically irrelevant because it’s only slightly related to the claim of no relationship between LDL and heart attacks.

The fifth is by Castelli et al and is again pretty much the opposite of what Teicholz said. Want some more choice quotes?
There is a very regular increase of CHD prevalence rates with increasing LDL cholesterol level at each level of HDL cholesterol.

The inverse relationship between HDL cholesterol and CHD, when taken over the three levels of LDL cholesterol, is significant (P < 0.001) by a method of Mantel, as are the positive trends of CHD prevalence on LDL cholesterol level.

Cross-classification of triglyceride with LDL cholesterol level (fig. 3) leads to the conclusion that either lipid has a statistically significant association with CHD prevalence […]

In general, then, when contingency tables are constructed for the three lipids considered two at a time, HDL and LDL cholesterol emerge as consistently significant factors in CHD prevalence […]
Does Teicholz even read the studies she cites?

Moreover, the book is marred by a clear bias of interpretation which is obvious even to the casual reader. She is highly critical of observational epidemiology and one can hardly blame her. Nutritional epidemiology is so full of data-dredging, misreporting, researcher bias and vested interests that it is tempting to dismiss the whole field. That is pretty much what Tiecholz does most of the time. If she says correlation cannot prove causation once she must say it a hundred times.

But here's the thing. She only says it when the studies don't support her argument. She frequently cites observational studies—and even ecological studies—when they suit her and she does so without mentioning that they are observational studies and without the many caveats she gives when the evidence doesn't go her way. Every conceivable limitation is mentioned when she doesn't like a study's result, even if the limitation (or "flaw" as she would call it) is unlikely to have distorted the results. If you want to claim that saturated fat is bad for the ticker she'll accept nothing except randomised controlled trials (which she claims do not exist, despite the Cochrane Review mentioned above being based on nothing else). On the other hand, any old cross-sectional study will do when she is trying to prove that eating nothing but blubber is good for you.

There are other blatant double standards in this book. When a food company funds a study into trans fats, for example, she draws attention to it twice one paragraph (p. 238), but when the Atkins Foundation funds research into the Atkins diet it is a footnote (p. 310). When the IARC finds that emissions from frying oils probably cause cancer, she says that they "determined that emissions from frying oils at the temperature typically used in restaurants are 'probably' carcinogenic to humans" (p. 278), but when the World Cancer Research Fund finds that fruit and vegetables probably protect against various cancers, she says that they "found that 'in no case' was the evidence for the consumption of fruits and vegetables in the prevention of cancer 'judged to be convincing.'" (p. 143) (Both IARC and WCRF have a grading system, with 'probable' being just below 'convincing' or 'known'. In other words, the degree of certainty was the same in both instances.)

Reading this book wasn't a complete waste of time. The chapters of the Mediterranean Diet and trans fats are quite informative, although you sometimes need to read between the lines. There is a good story tucked away in The Big Fat Surprise but this is not the book to tell it. The human interest comes from the people who made saturated fat (and later trans fats) the nutritional bogeymen. In Teicholz's rendering, it is a story of "personal ambition and money" (p. 332) with the outsized egos of a handful of publicity-hungry scientists silencing opposing voices for decades. There may be a grain of truth in this, but her caricature of Ancel Keys is a transparent hatchet job and, having seen the way she mangles quotes from her written sources, I'm not sure I can even trust her with the quotes she got from her interviewees. This is my fundamental problem with the book. After seeing how she misrepresents the science, I can't trust her on anything else.

My reading of the evidence, which I don't think is terribly controversial, is this. Heavy consumption of saturated fat is a risk factor for heart disease and heavy consumption of red meat (and/or low consumption of fruit and veg) is a probable risk factor for some cancers, but neither risk is big enough to stop me eating what I like. Obesity has risen in the USA as a result of rising calorie consumption and declining physical activity and this, in turn, has caused a rise in diabetes. Trans fats are probably not as bad for you as the Center for Science in the Public Interest would have you believe, and were only banned because they were in the wrong place at the wrong time. The Mediterranean Diet was and is an poorly defined gimmick. Nutritional epidemiology is a bit of joke, filled with false positives and groupthink, but there is more to nutritional science than epidemiology.

If you want to write a book about self-censorship and junk science in the nutrition research industry, be my guest. If you want to write a book saying that saturated fat is not worth worrying about unless you are in a high-risk group, go ahead. But what you can't do is use partial and one-sided information to create what is essentially a conspiracy theory and then go way beyond any reasonable interpretation of the science to make statements like this...

...cheese is probably healthier than bread. (p. 325)

Every reliable indicator of good health is worsened by a low-fat diet. (p. 330)

...saturated fats, like all fats, do not make people fat. (p. 334)

Eat butter; drink milk whole, and feed it to the whole family. Stock up on creamy cheeses, offal, and sausage, and yes, bacon. None of these foods have been demonstrated to cause obesity, diabetes, or heart disease... Sugar, white flour, and other refined carbohydrates are almost certainly the main drivers of these diseases. (p. 335)

...a snack of full-fat cheese is better than fruit. (p. 335)

These statements are, of course, unreferenced and indefensible. Does she not realise that sweeping generalisations, exaggerations and evidence-free assertions are precisely what she accuses Ancel Keys et al. of committing? Does she not see the irony of specifically a handful of foods (white flour?!) for obesity, diabetes and heart disease, based on vastly less evidence than exists for the saturated fat hypothesis?

In the final analysis, The Big Fat Surprise is a glorified diet book that tells people what they want to hear. Its dietary advice is very dubious and arguably irresponsible, but caveat emptor and all that. What is really disappointing is that the likes of the Economist have been taken in by it.

* In the 13 pages of corrections she sent her publishers for the reprinted edition, Teicholz changed the 43% figure to 40%, although this is still inconsistent with the CDC's figures.

** Weirdly, Teicholz has changed this in the reprinted edition to say "more than half is poultry" which is even wronger.

1 comment:

Christopher Snowdon said...

Chris, funny you review this book just as I found a related book discussed in the NY Post on Sunday that got my attention. You might have already heard of it and/or the author but just in case not, I'm fairly sure it's right up your alley and you might have something to say about it. The book is Body of Truth by Harriet Brown. The blurb:

"We take as a given that fat is bad and thin is good. But what if that’s not the case? In her new book, “Body of Truth,” Harriet Brown says studies have shown that being overweight does not result in a shorter life span, and that dieting — especially among children — leads to people becoming fatter and more unhealthy."