Good Calories, Bad Calories: A Critical Review; Chapter 2 – The Inadequacy of Lesser Evidence

Introduction

This is something of an ongoing review, chapter by chapter, of Gary Taubes’s extraordinarily dense book Good Calories, Bad Calories, which I usually shorten to GCBC. You might even consider this more of a fact-checking than a review, but whatever. I’m not going to get into a semantic argument. I wrote my first review of this book back in 2012, but after writing it I felt very unsatisfied. GCBC is such a dense book filled with so many unsubstantiated claims that I felt the book demanded a more thorough review. Other bloggers, like James Krieger at Weightology, seem to feel the same way and have tried to provide such a review only to eventually give up once they realize the gravity of the task. I may also give up at some point. I actually have given up a number of times only to feel compelled to hit at least one more chapter.

If you would like to read other parts of this ongoing review go to the table of contents on my Book Reviews page. FYI: All page numbers in this review refer to the hardback version of the book.

Not the Introduction

Page 24:

The press also played a critical role in shaping the evolution of the dietary-fat controversy by consistently siding with proponents of those who saw dietary fat as an unneccessary evil. These were the researchers who were offering specific, positive advice for the health-conscious reader—eat less fat, live longer. The more zealously stated, the better the copy. All the skeptics could say was that more research was necessary, which wasn’t particularly quotable. A positive feedback loop was created. The press’s favoring of articles that implied Keys’s hypothesis was right helped convince the public; their belief in turn would be used to argue that the time had come to advise cholesterol-lowering diets for everyone, thus further reinforcing the belief that this advice must be scientifically defensible.

There is no proof for this entire paragraph. None. This is just hyper-biased conjecture.

* * *

On page 24 Taubes misquotes an American Heart Association official

In 1964, when the study still hadn’t taken place, a director of the AHA described its purpose as the equivalent of merely “dotting the final i” on the confirmation of Keys’s hypothesis.

The Wall Street Journal article, from which this is taken, quotes the AHA director as saying something similar, but a different statement with a different meaning.¹ There is also no mention of Keys. From the article:

A director of the heart group explained yesterday the association decided to make the recommendation because heart disease is becoming such “a pressing public health problem.” Efforts to stem heart disease “just can’t be left until the last i is dotted and the last t is crossed,” he explained.

* * *

On page 25:

In 1957, Keys insisted that “each new research adds detail, reduces areas of uncertainty, and, so far, provides further reason to believe” his hypothesis. This is known technically as selection bias or confirmation bias; it would be applied often in the dietary-fat controversy. The fact, for instance, that Japanese men who lived in Japan had low blood-cholesterol levels and low levels of heart disease was taken as a confirmation of Keys’s hypothesis, as was the fact that Japanese men in California had higher cholesterol levels and higher rates of heart disease. That Japanese men in California who had very low cholesterol levels still had more heart disease than their counterparts living in Japan with similarly low cholesterol was considered largely irrelevant.

Damn this Keys fellow is so insistent! Does he have to insist literally everything? All kidding aside, the Japanese men in California study he cites was published in 1975. Yet Keys was apparently ignoring that evidence in 1957. Was Keys a person from the future who traveled back in time to the 50s to ignore research that had yet to be published?

Going back to the study that Taubes cites here you will find that it is an observational, cross-sectional study. (Keep that in mind because peppered throughout the book and in a recent blog post he essentially claims that all observational studies are meaningless.) Not only that but the researchers also measured other risk factors beside cholesterol, namely hypertension. Turns out that Japanese Californians also had more hypertension than their Japanese brethren, and hypertension is another major risk factor for coronary heart disease.

I can hear people screaming right now “SEE? This only proves the point that heart disease is caused by other things!” 1) No one is claiming that cholesterol is the sole cause of heart disease. Many things are considered risk factors: smoking, hypertension, cholesterol, diabetes, obesity, etc. Of course that doesn’t prove that cholesterol is not a risk factor. 2) This study is a cross-sectional study so it can’t measure causes, only associations. Taubes beats this drum later.

* * *

Pg 25-26:

The Masai nomads of Kenya in 1962 had blood-cholesterol levels among the lowest ever measured, despite living exclusively on milk, blood, and occasionally meat from the cattle they herded. Their high-cholesterol diets supplied nearly three thousand calories a day of mostly saturated fat. George Mann, an early director of the Framingham Heart Study, examined the Masai and concluded that these observations refuted Keys’s hypothesis. […] To explain away Mann’s research on the Masai, Keys then evoked more recent research suggesting that the Masai, living in nomadic isolation for thousands of years, must have somehow evolved a unique “feedback mechanism to suppress endogenous cholesterol synthesis.” This mechanism, Keys suggested, would bestow immunity on the Masai to the cholesterol-raising effects of fat.

HA HA!! Keys is such an idiot! Does he actually think that humans are capable of developing traits over time that help them adapt and survive in their environment?? A unique “feedback mechanism to suppress endogenous cholesterol synthesis. Is he serious? That’s almost as absurd as some Europeans developing a gene to digest lactose! That’s about as ridiculous as saying that some people have a gene that produces normal alcohol dehydrogenase, while other people of mainly Asian descent have a variant of the gene that codes for an enzyme that metabolizes alcohol differently. Could you imagine? Keys, don’t you understand that you sound like a jackass when saying these things? That’s almost as foolish as saying one could inherit genes that produce more melanin in the skin to protect from the sun’s UV radiation. How preposterous would that be! Oh… wait…

All sarcasm aside, the Masai do seem to be a unique bunch. Evidently much of their diet is raw, unrefrigerated milk, some sort of tea/stew made from the bark of a local tree, and some blood for good measure. I read the study Taubes cited and it’s a strange one. Take note that it is also an observational, cross-sectional study.² First of all, Mann claims he had trouble recording the diet:

The accurate measurement of dietary intake of these people proved extraordinarily difficult. We were able to make only limited measurements. This difficulty is because of the erratic intake of food, there being no fixed meal patterns in the families, because there are no uniform units of measurement or utensils and because of the disruption of usual behavior in the presence of an observer.

It seems he mainly relies on urine samples to estimate the protein intake of the Masai. Moreover, atherosclerosis was estimated by looking at ECG graphs. That’s odd, right? How accurate is that? In any case, Mann estimated very little atherosclerosis in the Masai. Cholesterol levels were made using the usual method, and they were indeed low.

So the Masai had low cholesterol levels and low atherosclerosis. Makes sense. The only thing that doesn’t really fit is thinking about a diet ostensibly high in saturated fat and cholesterol, but one that doesn’t lead to high serum cholesterol. As Mann put it, there can be one of two explanations: “Either such a diet does not in fact contribute to those consequences or the Masai have some other protecting mechanism which allows them to eat these foods with immunity.”

As it turns out the Masai do have such a protective mechanism as was demonstrated by a controlled feeding study published in the New England Journal of Medicine.^3,4 I’ll quote from the abstract:

The Masai of East Africa exhibit some unique biologic characteristics. Despite their customary diet comprised of 66 per cent calories as fat, they have persistent low serum cholesterol and beta-lipoprotein levels. Post-mortem examinations provided direct proof of a paucity of atherosclerosis. Metabolic studies revealed that the Masai absorbed large amounts of dietary cholesterol, but also possessed a highly efficient negative feedback control of endogenous cholesterol biosynthesis to compensate for the influx of dietary cholesterol. Two unusual serum-protein patterns were observed: the presence of a double alpha₂ band; and a high level of serum IgA that is apparent at an early age (four years). The high ratios of phospholipid to cholesterol in their gallbladder bile explain the extreme rarity of cholesterol gallstones. All these characteristics may reflect a long-term biologic adaptation of the tribe.

Hm. I guess both Darwin and Keys were right. Natural selection exists after all. However, Mann published another study some years after his ’64 study on the Masai. This one actually examined the hearts and aortae of recently deceased Masai (average age: 38). This newer study now claimed that the Masai actually had “extensive atherosclerosis” but for some reason the Masai’s blood vessels enlarge as they age to compensate for the blockage.⁵ Now I don’t know what to think about these people, except that they are indeed unique.

* * *

On page 27 Taubes discusses some early data from the Framingham Heart Study, noting that according to some results high serum cholesterol levels did indeed confer about a 5X greater risk of coronary heart disease.⁶ “But,” he says “there were caveats.”

As the men aged, those who succumbed to heart disease were ever more likely to have low cholesterol (as had Eisenhower) rather than high cholesterol. The cholesterol/heart-disease association was tenuous for women under fifty, and nonexistent for women older. Cholesterol has “no predictive value,” the Framingham investigators noted in 1971. This means women over fifty would have no reason to avoid fatty foods, because lowering their cholesterol by doing so would not lower their risk of heart disease. None of this was deemed relevant to the question of whether Keys’s hypothesis was true.

Talk about selection bias! This has to be one of the most selective readings of that study. The actual text says that serum cholesterol was highly predictive of heart disease.⁷ Lemme give you a few more quotes from that study that Taubes missed:

An increased risk proportional to antecedent serum cholesterol was found whether or not it was associated with elevated S_f20-400 prebeta lipoprotein.
When adjustment was made for the concomitant prebeta lipoprotein concentration and other factors related both to coronary heart disease risk and to blood lipids, a residual gradient of coronary heart disease risk proportional to the serum cholesterol was still evident.
Risk of coronary heart disease in men can be estimated using any of the lipids evaluated; however, none proved more useful than an accurate total serum cholesterol.
Cholesterol, as indicated by the size of the coefficient in Table 2, carries most of the weight as a contributor to coronary heart disease in men, whether manifested as angina or some more serious form of the disease.
Risk of each particular clinical manifestation of coronary heart disease (including angina, myocardial infarction, and sudden death) proved proportional to the antecedent serum cholesterol level in men of all ages studied.
In men and younger women the risk simply rose in proportion to the antecedent serum cholesterol concentration from the lowest to highest values recorded in this population sample (Figure 3). There was nothing to suggest that some particular level was “critical.” Hence, it does not appear logical to examine the relation of cholesterol to risk of coronary heart disease in terms of “hypercholesterolemia” but rather in terms of the actual concentration of cholesterol in the plasma.
Examination of the risk of developing coronary heart disease according to the actual serum cholesterol concentration of each subject grouped into quartiles showed an increase in risk proportional to the antecedent cholesterol concentration, not only in the general population but in persons free of factors believed associated both with hypercholesterolemia and with coronary heart disease
Even after excluding persons with hypertension, diabetes. ECG abnormalities, and the cigarette habit, a distinct gradient of risk proportional to the cholesterol concentration can be demonstrated. This tends to brand the lipid, not associated variables, as the culprit. As shown in an earlier analysis, a net effect of cholesterol is clearly apparent in men of all ages
According to this analysis the dominant effect was assigned to serum cholesterol, with an insignificant contribution of the other lipids It is apparent that in men cholesterol accounts for more of the total distance between those who developed and those who remained free of the disease than does S_f20-400 prebeta lipoprotein in all but the oldest subjects.
For cholesterol in men a strong residual effect remains after accounting for the level of S_f20-400 lipoprotein and the other factors In younger, but not older, women distinctly higher cholesterol values may also contribute independently to risk
Of all the identified host factors associated with increased susceptibility to coronary heart disease, the blood lipids are among the strongest.
If there is in fact a single common denominator through which the multiple interrelated predisposing factors in coronary heart disease operate, an abnormal accumulation or handling of blood lipids would appear the most likely candidate.
These investigators have found significant differences in the blood content of a variety of lipids and have implicated the total fasting triglyceride, beta and prebeta lipoprotein, alpha to beta cholesterol ratio, cholesterol to phospholipid ratio, and fatty acids
There were 6 xanthomatous individuals in the Framingham population sample of 5,127 men and women. All had serum cholesterol values exceeding 400 mg/100 ml and a strong family history of coronary heart disease, and within the follow-up period all six died of coronary heart disease before their fiftieth birthdays.
Any one of the lipids or lipoproteins examined, and, by inference, a triglyceride as well, can be used effectively for assessing vulnerability to coronary heart disease. None, however, would appear superior to the more convenient serum cholesterol determination for this purpose.
In women, however, the picture appears to be somewhat different. In women under the age of 50, as in men, high cholesterol values and not S_f20-400 prebeta lipoprotein appear to be associated with an increased risk. In older women, on the other hand, cholesterol appears to have no predictive value, and S_f20-400 prebeta lipoprotein actually appears to be superior to cholesterol for estimating risk.
The data presented suggest that in men the moderately elevated cholesterol values commonly encountered in the general population, regardless of the metabolic aberration responsible or how it is transported or partitioned among the lipoproteins, are associated with increased risk of coronary heart disease.

In case you were wondering, pre-beta lipoprotein is now called very low density lipoprotein (VLDL) which is similar to LDL and HDL in that it transports triglycerides and cholesterol through the blood. How anyone can read that study and only come away with cholesterol values in women over 50 doesn’t make much difference is really missing the point of the whole damn study. Especially when you consider what is very predictive of heart disease in women over 50 is a protein that transports lipids through the blood.

Here’s a simple chart from the article that shows the risk of heart disease on those with high serum cholesterol. FYI: Anything above a 1.0 means an increased risk. A 1.0 means no risk. Less than 1.0 means reduced risk. So a 2.0 would indicate a two-fold risk increase, 3.0 would be three times the risk, etc.

* * *

On page 27 Taubes mentions the Western Electric study but again leaves out important data that runs contrary to his thesis.

Two decades later, Jeremiah Stamler and his colleague Richard Shekelle from Rush–Presbyterian–St. Luke’s Medical Center in Chicago revisited Western Electric to see how these men had fared. They assessed the health of the employees, or the cause of death of those who had died, and then considered the diets each subject had reportedly consumed in the late 1950s. Those who had reportedly eaten large amounts of polyunsaturated fats, according to this new analysis, had slightly lower rates of coronary heart disease, but “the amount of saturated fatty acids in the diet was not significantly associated with the risk of death from [coronary heart disease],” they reported.

Immediately prior to the quoted phrase is a sentence that Taubes probably doesn’t want in his narrative so it is left out. I have reproduced it below.⁸ Bolding is mine.

When the risk of death from CHD was analyzed in terms of the component dietary variables, it was inversely related to intake of polyunsaturated fatty acids and positively related to intake of dietary cholesterol.

* * *

Taubes spills some ink on discussing a major dietary trial called the Anti-Coronary Club Trial on page 36:

The first and most highly publicized was the Anti-Coronary Club Trial, launched in the late 1950s by New York City Health Department Director Norman Jolliffe. The eleven hundred middle-aged members of Jolliffe’s Anti-Coronary Club were prescribed what he called the “prudent diet,” which included at least one ounce of polyunsaturated vegetable oil every day. The participants could eat poultry or fish anytime, but were limited to four meals a week containing beef, lamb, or pork. This made Jolliffe’s prudent diet a model for future health-conscious Americans. Corn-oil margarines, with a high ratio of polyunsaturated to saturated fat, replaced butter and hydrogenated margarines, which were high in saturated fats. In total, the prudent diet was barely 30 percent fat calories, and the proportion of polyunsaturated to saturated fat was four times greater than that of typical American diets. Overweight Anti-Coronary Club members were prescribed a sixteen-hundred-calorie diet that consisted of less than 20 percent fat. Jolliffe then recruited a control group to use as a comparison.

I’m gonna talk about what some might consider a minor point, but to me is significant. It is about rhetoric and word choice, so if you think this is petty and boring please skip ahead. What struck me was this: “In total, the prudent diet was barely 30 percent fat calories…” Barely? 30% of total calories as fat is smack dab in the middle of normal for most people. It’s nowhere near what anyone would consider a low fat diet. Yet Taubes uses the word “barely” to modify the 30% figure. Webster defines “barely” as “in a meager manner,” yet there is really nothing meager about eating a diet of 30-33% fat as it was stated in the actual study.⁹

As I have stated before and will no doubt state again, this is an example of when, if Taubes doesn’t blatantly misrepresent the results of a study, he will editorialize the figures in misleading rhetoric.

Let’s keep going with the Anti-Coronary Club Trial.

Jolliffe died in 1961, before the results were in. His colleagues, led by George Christakis, began reporting interim results a year later. “Diet Linked to Cut in Heart Attacks,” reported the New York Times in May 1962. “Special Diet Cuts Heart Cases Here,” the Times reported two years later. Christakis was so confident of the prudent-diet benefits, reported Newsweek, that he “urged the government to heed the club results and launch an educational and food-labeling campaign to change U.S. diet habits.”

The actual data, however, were considerably less encouraging. Christakis and his colleagues reported in February 1966 that the diet protected against heart disease. Anti-Coronary Club members who remained on the prudent diet had only one-third the heart disease of the controls. The longer you stayed on the diet, the more you benefited, it was said. But in November 1966, just nine months later, the Anti-Coronary Club investigators published a second article, revealing that twenty-six members of the club had died during the trial, compared with only six of the men whose diet had not been prudent. Eight members of the club died from heart attacks, but none of the controls. This appeared “somewhat unusual,” Christakis and his colleagues acknowledged. They discussed the improvements in heart-disease risk factors (cholesterol, weight, and blood pressure decreased) and the significant reduction in debilitating illness “from new coronary heart disease,” but omitted further discussion of mortality.

The “somewhat unusual” quote occurs in the first report, and not the second report as Taubes implies. What’s more is that the figures Taubes gives are not correct. The second report discusses deaths not related to heart disease:

Out of the 814 experimental group subjects, there have been 18 known deaths from causes other than coronary heart disease among individuals who had not experienced a new coronary event. This is in comparison to six such deaths out of the 463 individuals in the control group.¹⁰

The first report details deaths specifically due to heart disease, which was three in the experimental and zero in the control.⁹ This is not unusual considering the experimental group began with more risk factors for heart disease compared to the controls, namely greater rates of obesity, hypertension, hypercholesterolemia, etc. The results of the other endpoints (which Taubes surprisingly states, but also kind of minimizes) was that by the end of the trial the non-prudent dieters had significantly higher cholesterol, significantly higher blood pressure, significantly more obesity, and about triple the incidence of heart disease. All of this despite having less risk factors than the prudent group at the beginning of the trial. But I suppose these things aren’t important; what is important, according to Taubes, is the made-up death numbers.

Taubes states that further discussion of mortality was omitted which is clearly false because both reports obviously discuss the deaths in the trial. How do you pull numbers on mortality from the reports and then turn around and claim no discussion on mortality? Moreover, I don’t know what would make the data “less encouraging.” Considerably less encouraging for Taubes’s meat-heavy narrative, perhaps. The results are pretty consistent with the other similar dietary trials that Taubes tries to spin.

* * *

Pg 37 Taubes discusses another dietary study:

In July 1969, Seymour Dayton, a professor of medicine at the University of California, Los Angeles, reported the results of the largest diet-heart trial to that date. Dayton gave half of nearly 850 veterans residing at a local Veterans Administration hospital a diet in which corn, soybean, safflower, and cottonseed oils replaced the saturated fats in butter, milk, ice cream, and cheeses. The other half, the controls, were served a placebo diet in which the fat quantity and type hadn’t been changed. The first group saw their cholesterol drop 13 percent lower than the controls; only sixty-six died from heart disease during the study, compared with ninety-six of the vets on the placebo diet.

Only 66 died on the unsaturated fat diet while 96 died on the saturated fat diet? That’s only a difference of 30 lives! It’s 31% fewer deaths. Practically nothing, right? Barely a difference at all. By the way, in case you’re interested there are a few pertinent facts that Taubes left out (yet again). I have reproduced it below. Remember the control group was high in saturated animal fat, and the experimental group was high in unsaturated fats from plants.¹¹

The number of men sustaining events in major categories, in the control and experimental groups, respectively, was: definite silent myocardial infarction, 4 and 9; definite overt myocardial infarction, 40 and 27; sudden death due to coronary heart disease, 27 and 18; definite cerebral infarction, 22 and 13. The difference in the primary end point of the study-sudden death or myocardial infarction was not statistically significant. However, when these data were pooled with those for cerebral infarction and other secondary end points, the totals were 96 in the control group and 66 in the experimental group; P = 0.01. Fatal atherosclerotic events numbered 70 in the control group and 48 in the experimental group; P < 0.05. Life-table analysis in general confirmed these conclusions. For all primary and secondary end points combined, eight year incidence rates were 47.7% and 31.3% for the control and experimental groups, respectively; P value for the difference between the two incidence curves was 0.02.

If you don’t want to read the above block quote, I’ll summarize it for you: in all but one endpoint that was measured the experimental diet of unsaturated fats had less overt myocardial infarction, sudden death, cerebral infarction, fatal atherosclerotic events, etc. And not by a tiny margin – a significant margin.

The next paragraph Taubes states:

Thirty-one of the men eating Dayton’s experimental cholesterol-lowering diet, however, died of cancer, compared with only seventeen of the controls. The risk of death was effectively equal on the two diets. “Was it not possible,” Dayton asked, “that a diet high in unsaturated fat…might have noxious effects when consumed over a period of many years? Such diets are, after all, rarities among the self-selected diets of human population groups.” Because the cholesterol-lowering diet failed to increase longevity, he added, it could not provide a “final answer concerning dietary prevention of heart disease.”

What’s interesting is the the authors’ statement Taubes quotes from the paper. I want to make something clear: Taubes introduces the author and gives a brief background of the trial. Then he relates the conditions and methods of the study, then he cherry-picks one of the myriad results. In the next paragraph Taubes then gives his interpretation of the results, and tells you why he thinks the results were pretty much a wash. ONLY AFTER ALL THAT Taubes then reproduces a couple sentences from the journal article questioning whether a diet of unsaturated fat might have “noxious effects” presumably because of the study results. What Taubes likely wants the reader of GCBC to think is that after the results of the study are in and the numbers have been crunched and the data has been analyzed, Dr. Seymour Dayton is sitting at his desk and ruminating on what could have produced these results. As if he is asking a rhetorical question or providing a hypothesis for a future dietary trial.

What the reader of GCBC does not know is that Dayton asks this question in the introduction of the text, then later will explicitly answer this very question with his own data. Do you want to know if the experimental diet has noxious effects? Well there’s a section in the results portion of the study titled “Does the Experimental Diet Have Noxious Effects?” where Dr. Dayton states:

As indicated in table 29 and discussed in some detail above, the excess mortality in nonatherosclerotic categories was not sufficiently impressive to justify the conclusion that harmful effects had been demonstrated.

AND

One may also wonder whether the experimental diet may have exerted its effect on mortality data primarily by accelerating nonatherosclerotic deaths (see table 28), decreasing the atherosclerotic mortality by inducing early death due to other cause. Such a mode of action would be associated with higher numbers of deaths in the experimental group compared with the controls, whereas the reverse was true in this trial (fig. 13).

AND

The other observation which raised some question of a possible toxic effect was the low arachidonic acid concentrations in atheromata of long-term, high-adherence subjects on the experimental diet (tables 37 to 40). For reasons already cited, this may be more appropriately viewed as evidence of a salutary rather than a toxic effect.

I’d like to take this opportunity to mention that this is one of the many examples that shows that Taubes is not merely mistaken, but he is working very hard to quite literally deceive his readers. The numbers he cites as deaths from cancer is likely an honest citation error – no one could possibly know. But when you think about the information he presents it is evident that he read this text very carefully and thoroughly, yet only included the bits that he could spin to fit his narrative. With this text and many, many others he explicitly does not mention the very relevant information that runs contra to his nutritional ideology. He is essentially in the marketing business; marketing the low-carb religion rather than informing readers. It’s as if a movie critic were to review a film and write “Whoever was involved in making this film should be put in a gas chamber,” and the film’s marketer took the review and wrote “A gas!” on the movie poster.

Okay, I think I’ve made my point. Moving on…

Go to Good Calories, Bad Calories: A Critical Review; Chapter 3 – Creation of Consensus