The cost of misguided faith in end points – Prof. Noakes on Ancel Keys’ Cholesterol Con

In Part Six of Professor Tim Noakes’ ‘Ancel Keys’ Cholesterol Con’ series, Noakes recounted the developments that occurred between 1960 and 1967, which exonerated sugar and pointed the finger of CHD causation squarely at dietary fat. Part seven of the series – written in scrupulous detail only someone of Noakes’ calibre and comprehensive understanding can generate – proceeds with the timeline of the ever-increasing concretisation of Keys’ Twin Hypotheses as fact, picking up in 1968 when the initial results from the Oslo Secondary Prevention (Diet-Heart) Trial were reported. Once again, the interpretation of the results was wholly dependent on Keys’ Twin Hypotheses being correct and unchallengeable. – Nadya Swart

Ancel Keys’ Cholesterol Con. Part 7. 1968–1970

  1. 1968. The initial results from the Oslo Secondary Prevention (Diet-Heart) Trial are reported. 

The first results of this study initiated at five medical departments in Oslo, Norway between 19561958 were published in 1968 (1). The goal of the study was to determine if a diet ‘rich’ in polyunsaturated fats from soybean oil would reduce blood cholesterol concentrations and as a result, according to the Keys’ Twin Hypotheses, the progression of CHD in survivors of a first heart attack. 

Indeed the initial five-year results appeared very impressive. First, blood cholesterol concentrations fell steeply on the low-fat intervention diet during the first five years of the trial (Figure 1). 


Legend to figure 1. Mean serum cholesterol concentrations during the first five years of the Oslo Diet-Heart Study. Blood cholesterol concentrations fell sharply on the low-fat diet rich in polyunsaturated fats and remained low for the five years of the trial. Note, however, that the scale on the Y-axis (serum cholesterol concentrations) is truncated – it begins not at zero but at 220mg/dL – so that the magnitude of the effect is greatly magnified. Reproduced from figure 1 in reference 1. 

Second, the relapse rate for further CHD events was significantly lower in the diet intervention group (figure 2). 


Legend to figure 2. Cumulative incidence of total CHD relapses in the Oslo Diet-Heart Study. The reduction in total relapses in the dietary intervention group over the five-year period (64 vs 90 in the control group) was statistically significant (p<0.011). Reproduced from figure 2 in reference 1. 

So everything was looking hopeful for a favourable outcome as the trial progressed. But there was trouble in paradise. Perhaps the initial results were simply too good to be true.  

First, although the recurrence rate was significantly reduced in the total group in the intervention trial, the benefit was enjoyed only by those below the age of 60 (1). Above that age, there was no benefit, even in the presence of lowered blood cholesterol concentrations. If a particular mechanism explains an effect in those below 60 years of age, then that same mechanism should be even more effective in older persons who are at an even greater CHD risk. 

Second, as the trial progressed, it became clear the sole remaining significant difference in outcomes was in rates of fatal myocardial infarction (Figure 3) without any reduction in rates of sudden death (Figure 4). Thus, the author concluded: “It seems as if sudden death in survivors of myocardial infarction is uninfluenced by diet. The reason for this is unclear.” (2, p.941).


Legend to figure 3: Deaths from fatal myocardial infarction during the 11-year follow-up were significantly reduced (p=0.004) in the diet intervention group in the Oslo Diet-Heart Study. Reproduced from figure 1 in reference 2.

Legend to figure 4: Overall mortality over 11 years was not significantly different (p=0.35) between the two diet groups in the Oslo Diet-Heart study. Nor was overall mortality (p=0.097 – data not shown). Reproduced from figure 4 in reference 2.  


The author’s explanation was that lowering the blood cholesterol concentration might well be reversing atherosclerosis (their supposition) but that the development of fatal cardiac arrhythmias might be as a result of residual factors of prior damage to the heart caused by the previous heart attack. 

Once again, the interpretation is wholly dependent on Keys’ Twin Hypotheses being correct and unchallengeable.

So, the great challenge posed by the study was that, whereas some measure of cardiovascular outcomes – heart attack relapse rates and fatal heart attack rates –  were significantly reduced by the diet, the truly important measures – overall (all-cause) mortality and total CHD mortality – were not improved by dietary intervention. 

The clear warning from the Oslo Diet-Heart Study was that while lowering the blood cholesterol concentration might well reduce the clinical expression of some forms of fatal CHD, in this case fatal heart attacks (figure 3), it must be increasing deaths from other causes since total (all-cause) mortality was not reduced by the dietary intervention (figure 4). 

This ability of interventions, including the use of cholesterol-lowering drugs, the statins, to modify only one measure of hard cardiovascular outcomes, like heart attacks or sudden deaths, without also reducing total mortality would become a recurring theme. 

It would be devoutly avoided in polite discussions by those promoting the alleged benefits of the cholesterol-lowering effects of drugs and low-fat dietary interventions. To this day, this approach continues unchanged and unchallenged in almost all medical schools across the world. 

  1. 1967. Donald Fredrickson MD, Robert Levy MD and Robert Lees MD develop a novel classification of blood lipid abnormalities predisposing to CHD.

In all the clinical trials then being considered for funding by the NHI in the 1960s, the singular focus had been on the measurement of blood cholesterol concentrations. 

However, in 1967, Donald Fredrickson MD and colleagues published a series of five articles (3–7) in the prestigious New England Journal of Medicine with an editorial in the Annals of Internal Medicine (8). 

Recall that John Gofman had used ultracentrifugation to delineate the different lipoproteins and had produced an atherosclerosis index based on the concentrations of different lipoproteins according to their flotation values (9,10). Remember also his warning that carbohydrates caused one of the specific abnormalities he had identified: “Dietary carbohydrate intake is a prime factor controlling the serum level of the Sf o 20–100 and Sf o 100–400 lipoprotein classes. Restriction of dietary carbohydrates can provoke marked falls in the serum level of these lipoproteins…” (11, p. 282-283). 

The problem for Gofman was there were so few ultracentrifuges in the world then, as now, that his classification method (9) was never promoted or adopted. So, instead Gofman left the field and began a highly successful second career investigating the risks of low level radiation on human health (12).

The advance in the science that Fredrickson and colleagues provided was to classify five different hyperlipoproteinemias based on the electrophoresis of blood samples. As the technique was more widely available than ultracentrifugation, for a short time, the Fredrickson classification became the global standard. But it too would ultimately be regarded as too complex; to be replaced by the reductionist measurement of just blood LDL- and HDL-cholesterol concentrations.

The authors stressed the detection of the hyperlipidemias/hyperlipoproteinemias begins with the measurement of blood cholesterol and triglyceride concentrations. If both are within the normal limits, “hyperlipoproteinemia is ruled out with a degree of precision that is quite adequate for current use” (3, p.148). 

Judging by modern practice and teaching, this rule has long been forgotten as, in current practice, lipid abnormalities are defined essentially exclusive on whether or not the blood cholesterol concentration is elevated above 200mg/dL (5.2mmol/L); nothing else is really of much interest. 

There are currently four reasons why so little attention is paid to the simultaneous measurement of both blood cholesterol and triglyceride concentrations and why medical practitioners have little understanding of the meaning of elevated blood triglyceride concentrations.

  • No drug has yet been found that can safely lower blood triglyceride concentrations.
  • The commonest factor driving elevated blood triglyceride concentrations is a high carbohydrate diet (13). 
  • The result is that a low carbohydrate diet, not prescription drugs, is the sole treatment required for the majority of persons with hypertriglyceridemia.  
  • Since there is no financial incentive for the pharmaceutical industry to promote a dietary intervention for the management of (carbohydrate-sensitive) hypertriglyceridemia, neither the medical profession nor their patients will ever likely be informed of this simple cure.

Fredrickson, Levy and Lees produced what became known as the Fredrickson Classification of Hyperlipoproteinemias. I have grossly simplified the classification in table 1 for a reason that will become apparent shortly.


In 1972, Levy, Fredrickson and eight other participants attended an NIH conference on the dietary and drug treatment of the hyperlipoproteinemias (22). The overall dietary advice they recommended was the following: 

“In summary, fasting chylomicronemia is treated by reducing the dietary fat intake, thereby reducing the rate of entry of chylomicrons into the bloodstream. Endogenous hypertriglyceridemia (very low density lipoproteins) is managed first by weight reduction and then a maintenance diet that is restricted in carbohydrates and alcohol. Elevated low-density lipoproteins and the resultant hypercholesterolemia are treated by a diet low in cholesterol and saturated fats, to which polyunsaturated fat may be added” (22, p.275). 

For the management of Type I, II, III and V Hyperlipoproteinemias, they advised low fat or low cholesterol diets, or both (Table 4, p.276). For Type IV Hyperlipoproteinemia, they advised: “Controlled carbohydrate (CHO) (approximately 40 to 45% calories); moderately restricted cholesterol” (p.276). Carbohydrate restriction required that “most concentrated sweets (should be) eliminated” (p.276).

Despite all the evidence that this form of Hyperlipoproteinemia responds extremely well to significant carbohydrate restriction (13,16-21), the best these experts could offer was that patients needed to eliminate “concentrated sweets” from their diets. Ignoring the “carbohydrate factor” in the causation of the hyperlipidemias, specifically carbohydrate-sensitive hypertriglyceridemia, would only intensify with time.

Indeed, these guidelines set the pattern that would be followed right up to today. They continue to ignore the published evidence and instead demonise fat as the sole dietary nutrient of concern in the prevention of CHD. Any dangers of carbohydrate ingestion in raising blood triglyceride and lowering blood HDL-cholesterol concentrations has simply been written out of history. 

  1. 1968. The first findings of the Los Angeles Veterans Trial of the “effects of a diet high in unsaturated fat in preventing complications of atherosclerosis” are reported.

In 1962, University of California at Los Angeles Professor of Medicine Seymour Dayton MD, initiated an eight-year long clinical trial among 846 middle-aged and elderly men living in a local Veterans Administration (VA) Hospital (23–27). The study would become known as the Los Angeles Veterans Trial. 

The 846 volunteers were randomly assigned in equal numbers to the control and experimental groups. The matching was so good, the two groups were indistinguishable at the start of the trial.

The persons in the control group received the standard American diet at that time providing 40% of calories as fat with a cholesterol intake of 650mg/day. The experimental diet provided 39% of calories as fat but was lower in saturated fat and higher in linoleic content (38% of total fatty acids compared to 10% in the control diet). This was achieved by replacing the saturated fats in butter, milk, cheese and ice cream with polyunsaturated fats from corn, soybean, safflower and cottonseed oils (28, p.75). Unnoticed until recently was the fact that the control diet included hard margarine and hydrogenated fish oils providing as much as half a cup of trans fats per day; the intervention diet contained no trans fats (28, p.78–79). In addition, the intervention group reduced their smoking rates by 45% more than did the control group. Both these changes biased the outcomes significantly in favour of the intervention group and it was no longer a fair trial. 

The experimental diet produced a prompt 13% reduction in blood cholesterol concentrations in the intervention group. Adipose tissue linoleic concentrations also rose more than three-fold in diet adherers. 

The final analysis after eight years showed that although the number of sudden deaths owing to CHD (27 vs 18; control vs experimental) and of acute heart attacks (44 vs 36) was lower in the intervention group (Table III in reference 24), the differences were not statistically significant. Thus according to the hypothesis being tested, specifically that the intervention would reduce the primary end points of the trial – sudden deaths and heart attacks – the study had failed. 

However, when data from other non-specified endpoints – specifically strokes and “other secondary endpoints” like ruptured aneurysms, amputations and “miscellaneous” – were included in the analysis, the results suddenly became significant. So that, all of a sudden, there were 96 “primary and secondary end-points” in the control group but only 66 in the experimental group, and this difference was now significant (p<0.01). And this is the number that is remembered (28, p.75).

However, the more important numbers were those for total all-cause mortality. For the first six years of the trial, all-cause mortality was not different between groups (figure 5 – left panel). But then something odd happened: “During the late part of the trial, there was a crossover of the curves for total death-rates due to excess nonatherosclerotic mortality among the experimental subjects … The difference in the nonatherosclerotic deaths in this period was due to trauma (0 controls, 4 experimental) and to carcinoma (2 controls, 7 experimental)” (24, p.1062). 


Legend to figure 5: Left panel: Cumulative death rate from all causes (top lines) and from fatal acute atherosclerotic (A.S.) events (heart attacks and strokes) in the experimental and control groups in the Los Angeles Veterans Trial. Right panel: Percent of survivors (who had not died from other causes) in the experimental and control groups in the Los Angeles Veterans Trial. Reproduced from figure 3 in reference 24 and figure 14 in reference 25). 

Figure 5 (left panel) shows that for the first seven years of the Los Angeles Veterans Trial, all-cause mortality in the experimental and control groups was identical. But in the eighth year, mortality in the experimental group suddenly increased and remained above that of the control group for the remainder of the trial.

The right panel of figure 5 shows the reasons for this. In the eighth year of the trial, there was a dramatic increase in deaths owing to causes other than those related to coronary atherosclerosis. 

And the main contributor to these excess deaths in the experimental group were deaths from cancer (26) with 31 men on the ‘vegetable’ oil diet dying from cancer compared to just 17 eating the control diet. Dayton and his colleagues had effectively answered in the positive, the question they had posed when they began the trial a decade earlier: “Was it not possible, too, that a diet high in unsaturated fat (which appeared the most attractive modality for a test (of the Diet-Heart hypothesis – my addition) 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. Thus there had been little practical experience with them and practically no long-term experimental experiences when we planned the study” (25, 1L-2). 

In addition, the trial reported the incidence of gallstone measured at autopsy was significantly increased in the group eating the vegetable oil diet (27) whereas the extent of atherosclerosis also measured at autopsy was not different between groups (25). 

Thus the true message from the Los Angeles Veterans Trial was that this dietary-induced reduction in blood cholesterol concentrations may have had a small influence on CHD events, either fatal or non-fatal, but that it failed to influence all-cause mortality, which should always be the key outcome measure. It helps not if a diet ‘prevents’ a few fatal heart attacks if it increases deaths from other causes, in this case from cancer. So that, in the end, mortality rates are unchanged.  

Unfortunately, in their rush to prove that a dietary change – particularly the replacement of saturated fats with polyunsaturated ‘vegetable’ oils – is beneficial, many scientists, especially those with close ties to the vegetable oil industry continue to include this study, as well as the Oslo Diet-Heart and the Minnesota Coronary Experiment, as definitive proof that diets rich in polyunsaturated ‘vegetable’ oils will improve CHD outcomes (29). 

Yet this is patently misleading since all these three studies found that ‘vegetable’ oils did not alter all-cause mortality, which is the key measurement. What is more, when all the appropriate studies are included, the evidence is that replacing dietary saturated fatty acids with omega-6 polyunsaturated fatty acids does not provide cardiovascular benefit, but may be harmful (30). 

Whether or not part of that harm is an increase in risk of developing specific cancers is a question that is seldom, if ever, posed. 

  1. 1969. The National Institutes of Health Review Panel meets to decide whether a test of the Diet-Heart hypothesis is feasible.

In June 1969, the NIH released the findings of its committee led by Edward ‘Pete’ Ahrens MD to review the evidence for and against the Diet-Heart Hypothesis and to provide recommendations of how the Institute should proceed (31, p.40). The report noted: “The essential reason for conducting a study is because it is not known whether dietary manipulation has any effect whatsoever on coronary heart disease.” Taubes makes the telling point this was despite the fact that, motivated by its funding from Procter and Gamble and other food companies, the AHA “had been recommending low-fat diets for almost a decade already” (31, p.40). 

However, as Ahrens recalls, the NIH committee members believed the conduct of a study, which would provide definitive answers “would be so expensive and so impractical that it would never be done” (31, p.40).

But it seems even this negative conclusion would not be sufficient to mitigate the influence of Keys and his Diet-Heart acolytes who would continue to pretend such a study was largely unnecessary since they already knew what results it would (or should) produce.

  1. 1970. The first results from Keys’ Seven Countries Study (SCS) are published.

Fourteen years after funding was granted to Keys to begin the SCS, the first findings and conclusions were published. I’ve reviewed those results in some detail in an earlier column (32); here I consolidate those ideas once more.

The SCS had some significant weaknesses, which were downplayed by Keys and his colleagues and which continue to be ignored by those who believe Keys produced ‘irrefutable evidence’ that prove his hypotheses (33,34). 

First, it was an associational study and associational studies cannot prove causation except in exceptional circumstances (Figure 1 in reference 32).

Second, the choice of study countries was not random and so could not exclude the presence of residual confounders in those countries that were studied. In other words, the associations that were found could have been a chance finding, a result purely of the specific countries chosen non-randomly. Those countries might have included undetected confounders, which were the true explanation for the associations found between saturated fat intake and heart disease rates.

Third, many of the countries in which dietary analyses were done were recovering from the devastating effects of World War II: “The historical period of the Seven Countries study was also a problem. The years that it encompassed, from 1958 to 1964, were a time of transition in the Mediterranean region: Greece, Italy and Yugoslavia were still recovering from World War II, which had brought about extreme poverty and near-starvation, and Italy was also emerging from 25 years of suffering under a fascist government. Hardship had led four million Italians to flee their country, and at least 150,000 Greeks to leave theirs” (28, p.37-38). 

Teicholz describes the effects of the war on the nutrition of those living in Crete and in Calabria, situated in the boot of Italy. Those living in Crete complained they were “miserable” with their post-war diet; 72% said that meat alone or with cereal was the favourite food that they missed (28, p.220). Likewise the Calabrians “considered the lack of food … almost entirely vegetarian, as the cause … of general mortality for cases linked to nutrition, the low stature of individuals, their physical weakness, their low ability to work and psychological debility” (28, p.221). Like the Cretans, the Calabrians desired one thing: “Meat is what the peasants craved, above all else … The robust man, tall and ‘erotic’, was the man who had eaten meat” (28, p.221).

In addition, the data for Crete, whose inhabitants were found to be at especially low risk for developing CHD, was collected during Lent when Orthodox Greeks abstain from “all foods of animal origin, including fish, cheese, eggs and butter” (28, p.40). Subsequent studies found the intake of saturated fat falls by half during Lent in Crete (28, p.40). 

Fourth, unlike the excellent collection of dietary data in the Framingham Heart Study (35), the collection of the dietary data in the SCS was abysmal (28, p. 40–43). 

The dietary data were hopelessly incomplete as it was collected in less than 4% of the total study population (28, p.40). The classification of some foodstuffs was also incomprehensible. As a result, cake and ice cream were classified as saturated fats instead of refined carbohydrates whereas meat and eggs were classified as saturated fat when more than half their fat content is unsaturated. Similarly one-third of the fat content of butter and cream is unsaturated fat but this was also misclassified (36,37). 

So bad in fact were the dietary data that the authors eventually published the following disclaimer: “Obviously, an attempt to relate the 15-year risk of death of the individual to his diet is impractical … Those estimates of average contributions of the nutrients of interest should be reasonably good for the cohort but nothing can be said about the risks of individual men … The relationships shown here are statistically important but they are not claimed to be necessarily causal” (38, p.914 – my emphasis added). Yet this has not prevented others from concluding the SCS provided evidence of a causal link between diets high in saturated fat content and increased rates of CHD (34). 

Fifth, the SCS found a neatly significant correlation between the median blood cholesterol concentrations and the 10-year CHD death rates in the different populations/countries (39, p.183) (Figure 6).


Legend to Figure 6: The Seven Countries Study (SCS) found a significant relationship between the median blood cholesterol concentrations and the 10-year CHD mortality rates in the different populations/countries. Redrawn from figure 4 in reference 39, p.183. 

Sixth, the SCS found a significant relationship between the percentage of calories from saturated fat and the 10-year CHD mortality rates in the different countries/populations (Figure 7) (38, p.183).  

Legend to Figure 7: The Seven Countries Study (SCS) found a significant relationship between the 10-year CHD mortality rates and the percentage of dietary calories from saturated fat in those countries/populations. Recall that the manner in which these dietary data were collected was, by the authors’ own acknowledgement, far from ideal. Redrawn from figure 5 in reference 39, p.183.

Setting aside the concerns with the manner in which the dietary information was collected in these different countries, many have drawn attention to the anomalies in the data; anomalies that Keys glossed over. 

Teicholz has pointed out that the Eastern Finns “died of heart disease at rates more than three times higher than the Western Finns, yet their lifestyles and diets, according to Keys’ data, were virtually identical. The islanders of Corfu ate even less saturated fat than did their countrymen on Crete, yet on Corfu heart disease rates were far higher. Thus, within countries, the correlation between saturated fat and heart disease didn’t hold up at all” (28, p.39).

Smith and Pinckney (40) have re-analysed the original data used by Keys and I have reproduced their findings in Table 1. In Figures 8 and 9, I emphasise the anomalies they detected in those data. 

When the countries used to produce figure 7 above are ranked according to % calories from saturated fat (Table 1), it is clear it is not possible to predict CHD death rates and, especially, total mortality (all deaths) from the simple knowledge of the saturated fat intakes of the people living in the different countries.

Table 1: Comparison of CHD deaths and total death rates in 16 different population groups studied in Keys’ Seven Countries Study when ranked according to the % of dietary calories ingested as saturated fat. Reproduced from Table 3.11 in reference 41, p.54. 

The reason is simply because there are so many anomalies identified by Teicholz (28), Smith and Pinckney (40) as subsequently also highlighted by Enig (41). The most glaring of these anomalies are shown graphically in Figures 8 and 9. 

Legend to Figure 8. Comparison of CHD death rates and saturated fat consumption in pairs of countries from the Seven Countries Study having the identical saturated fat consumption but different CHD death rates. Redrawn from data of Smith and Pinckney (40) as reported by Enig (41).

Legend to Figure 9. Comparison of CHD death rates and saturated fat consumption in pairs of countries from the Seven Countries Study having only a 1% difference in saturated fat consumption but different CHD death rates. Redrawn from data of Smith and Pinckney (40) as reported by Enig (41). 

Clearly, the citizens of Rome, Crevalcore and Zrenjanin have higher CHD death rates than they should for their low % saturated fat intakes (Figure 9) whereas something else, either genetic or environmental, is protecting the citizens of Crete, Velika and Dalmatia and, to a lesser extent those of West Finland, from the supposedly toxic effects of their relatively high dietary saturated fat intakes (Figure 8).

But there is worse to come. 

For what Keys and his acolytes managed to achieve was to blur the evidence. They did this by referring only to any (associational) evidence that might support Keys’ Twin Hypotheses while never drawing any attention to any evidence that challenged those hypotheses.

So figure 6 shows what appears to be a very tight relationship between median blood cholesterol concentrations and CHD death rates in the different populations that were studied. The graph looks impressive and this is why it was chosen as the initial graph in Steinberg’s rapturous article (38). 

Figure 7, on the other hand, shows a rather more scattered relationship between % saturated fat intakes and CHD death rates, a relationship, which appears on closer examination, probably to be meaningless (Table 1; Figures 8 and 9).

But by showing Figures 6 and 7 together, Steinberg hoped he would fool us into not asking the really important question: Yes, but where is the graph showing the relationship between % saturated fat intake and blood cholesterol concentrations? For that is the absolutely fundamental basis of Keys’ Diet-Heart Hypothesis. Without that relationship, any other relationships become irrelevant.

Recall the figure I used in an earlier column (42) and which is reproduced here for convenience as Figure 10.

Legend to Figure 10: The ‘proof’ of Keys’ Diet-Heart and Lipid Hypotheses are based on a triangulation of interactions between dietary saturated fat and blood cholesterol concentrations in causing elevated blood cholesterol concentrations, which then directly cause coronary atherosclerosis and CHD. All three relationships rely almost entirely on epidemiological, associational studies. While associational studies cannot prove causation (Figure 1 in reference 32), the absence of this relationship in any study argues strongly against causation. Importantly, relationship 1 is critical. If it is absent, the hypothesis has no foundation.

The key takeaway point from figure 10 – the one that is always ignored by Keys and his supporters – is that each of the three associational relationships shown in that figure must always be true in all reported studies for the hypothesis to be supported (my emphasis). For whereas an associational relationship cannot prove causation except under exceptional circumstances, the absence of an associational relationship between two variables in any study is strong evidence that the two variables are almost certainly not causally related. 

What Keys and his acolytes have done so successfully over the past 67 years is to juggle the evidence. They have mixed and matched scientific studies that support only one or at best two of the three legs of the association as if that is sufficient evidence to prove the hypothesis. But the proof of Keys’ Twin hypotheses requires that all three legs of evidence must be present in all studies. When they are not, the hypothesis is disproved. As I will show, the SCS, like the FHS, provided definitive proof that one of the three legs is unsupported by any evidence. 

So figures 6 and 7 address only two of the three legs of this relationship. Specifically, the relationship Number 3 between dietary saturated fat and CHD mortality and relationship Number 2 between blood cholesterol concentrations and CHD mortality. But Steinberg’s article (39) includes no reference to the relationship Number 1 between dietary saturated fat and blood cholesterol concentration – the foundation on which Keys Diet-Heart hypothesis is based.

To correct that error, I have collected the data and prepared the appropriate slide to address this question (Figure 11).

Legend to Figure 11. The crucial missing evidence from the Seven Countries Study (SCS) that Ancel Keys and his team somehow forgot to publish. The SCS, just like the Framingham Diet Heart Study (35) and many, many others failed to find any specific dietary factors that might explain differences in blood cholesterol concentrations in individuals in that study (32; 43). Naturally, this disproves the foundational prediction on which the Keys’ Twin Hypotheses are based (see figure 10).

The reason why Figure 11 was not included in Steinberg’s review (39) is because it is the missing piece of evidence that disproves Keys’ Twin Hypotheses. By publishing only figures 6 and 7, Keys and Steinberg were able to hide the inconvenient evidence that disproved Keys’ Twin Hypotheses, as explained in Figure 10.

So, in the end, it was the key evidence, which Keys failed to report that is the most important. Which reminds one of the Sherlock Holmes story of the dog that did not bark in the night (44), a story I told with relish during my trial before the Health Professions Council of South Africa (45, p.265–267).

The final point from the SCS was that whereas elevated blood cholesterol concentrations were associated with higher rates of CHD, lower blood cholesterol concentrations were linked to higher rates of non-coronary events in the different populations/countries.

As a result, longevity was unrelated to either higher or lower blood cholesterol concentrations (32,46). 

If anything, the key conclusion from the SCS should have been that, in terms of how long you are going to live, the SCS found it matters not how much saturated fat you eat or what your blood cholesterol concentration is. Naturally, Keys and his acolytes ensured this conclusion was never made public.

Eight years before this report on the SCS findings was published, EA Law warned of the impossibility of determining causation from associational studies of poorly defined (unhomogenous) populations: “…it is not possible to tell from uncontrolled samples (of populations) whether an observed association between the incidence rate of (atherosclerotic coronary heart disease) and serum cholesterol is or is not, in substantial part, due to uncontrolled factors…. comparability of such (coronary heart disease) death rates between countries with widely different medical and cultural standards is obviously highly questionable” (47, 48 p. 871).

Too true.

  1. 1970. The NHLBI forms an expert panel on hyperlipidaemia and atherosclerosis. This became the 1971 NIH Task Force on Atherosclerosis. The Task Force agrees to spend $250 million on two trials – the LRC CPPT and MRFIT – neither of which tests the role of diet in CHD.  

In June 1970, the then Director of the NHLBI, Theodore Cooper, invited Donald S Fredrickson to convene an expert panel, the Panel on Hyperlipidemia and Atherosclerosis (also known as the Task Force on Atherosclerosis (49, p.40) to advise the Institute on two questions. 

It’s important to emphasise at this point that neither of these questions was directly related to the Diet-Heart hypothesis. It was as if the NIH and the NHLBI had accepted for the time being that a definitive trial of that hypothesis was not possible. So it redirected its funding to test the Lipid Hypothesis, which holds that blood lipids, cholesterol especially, drives the development of CHD. 

Fredrickson had been chosen largely because he had recently published that series of influential articles in the world’s most prestigious medical journal, the New England Journal of Medicine (3-7).  

The first question to be addressed was: should the NHLBI establish a network of lipid centres of excellence across the country that could standardise methods of lipid and lipoprotein analysis? (50).

Not unexpectedly, since most of the members of the expert panel would draw considerable personal and institutional benefit from their positive answers, all 21 panel members agreed. 

The second question was: do you believe the evidence is sufficient to warrant the detection of and some form of individual treatment of hyperlipidemia? Twenty panel members agreed. However, “they recognised that the evidence that such treatment would reduce heart attack rates, and by how much, was still limited. Therefore they went on to recommend that the program must include a randomised intervention trial to determine the effect of the treatment of hypercholesterolemia on atherosclerotic complications” (50, p.2).  

In the end, the Task Force agreed that while a definitive test of Keys’ Diet-Heart Hypothesis “in the general population is urgently needed” (49, p.40), they did not believe such a study was practical because of “formidable costs” estimated to be as much as $1 billion. 

So, instead of spending $1 billion on a single large study, the Task Force advised the NHLBI to spend $250 million on two smaller studies.

Importantly, neither of the studies was a direct test of Keys’ Diet-Heart Hypothesis. Instead one trial would evaluate the effects of a cholesterol-lowering drug on CHD outcomes. The trial would become known as the Coronary Primary Prevention Trial (CPPT) of the Lipid Research Clinics (LRC). The second, the Multiple Risk Factor Intervention Trial (MRFIT), would use multiple interventions to alter what are considered the key coronary risk factors – cigarette smoking, elevated blood pressure and elevated blood cholesterol concentrations – to measure any potential impact on CHD outcomes. 

As Taubes (49) explains: “Neither of these trials would actually constitute a test of Keys’ hypothesis or of the benefits of a low-fat diet. Moreover, the two trials would take a decade to complete, which was longer than the public, the press, or the government was willing to wait (p. 41). 

The end result was that within a short time, the NHLBI had established a new branch, the Lipid Research Branch, and had committed to funding 12 collaborating laboratories – the Lipid Research Clinics (LRC) – to answer the first question. In addition, within a year, the group began planning the drug intervention trial, the LRC CPPT.

Clearly no one at the time had the foresight to consider the deep scientific hole the NHLBI had dug for itself. For no one could be certain what would be the outcome of their research studies. Begun in blind faith, the ideal outcome for the 12 LRCs and their directors would be a positive outcome. Then the investment would have been justified and there would be a reason for the NHLBI to continue funding the clinics, further to advance understanding.

But what if the outcomes were negative? What would that mean to the directors of the 12 collaboration LRCs and the staff running the studies? 

The expectations of the 12 collaborating LRCs would have grown heavy on the years of assured income as the trial was being conducted. And a positive outcome would conveniently extend their capacity to extract yet more research funding from the NHLBI to undertake even more “confirmatory” trials. 

Perhaps no one foresaw that a negative result would require the dismantling of the extensive bureaucracy and all the high hopes and expectations that had been created.

The foundation had been laid onto which only the construction of a positive research finding would be sanctioned. That is the nature of the human condition; and of scientists who depend on the state for their funding and their careers. 

Fredrickson expressed his personal concerns about large scale clinical trials in humans: “The fourth is interpretation, the postgame critique that determines what we have learned and whether we shall begin all over again … For example, a number of field trials in our day that have dealt with the value of anticoagulants or the effect of various drugs and diets upon atherosclerosis can be said kindly to have ‘ended in equivocation.’ …. It is not that we must always have a positive result or that we abhor the thought of a negative (my emphasis). It is the drawing of neither that is so unsettling. What is the problem? Is man too complex for us to reduce to manipulation of single variables? Shall we abandon field trials and rely on intuition when laboratory animals will not do? The answers are obvious. Man is complicated, but he is also often unique as a test system. Field trials are an indispensable ordeal. The problem is to conduct better field trials. We can do this … We have learnt the cost of compromise, of cutting numbers of subjects too close in misleading attempts at economy, of neglecting randomisation and the double-blind … of failing to standardise methods or, worst of all, of having misguided faith in end points that perniciously soften with time”. 

“Learning by trial and error is part of the experimental process. One cannot expect each experiment to be a success. But when the experiment is a field trial, the costs of an equivocal answer are on a different scale from that of the laboratory … The institute also has before it the report of a distinguished and hard-working panel of experts who concluded from their Diet-Heart Feasibility Study that the diet can be altered so as to lower cholesterol levels in free-living Americans. They further recommended strongly that such alterations be carried out in a population large enough and for a sufficient time to determine once and for all whether this decreases the mortality and morbidity from coronary artery disease … Costs are being analysed. It is conceivable they might exceed 10 or even 20 million dollars a year for up to 10 years. Unthinkable? No, eminently thinkable. The economic cost of premature coronary artery disease (deaths from myocardial infarction before age 65) easily exceeds one billion dollars a year. A positive step that reduces this toll will be proportionately worth it in cost (my emphasis)” (51, p.986).

As I describe subsequently, the findings of the LRC CPPT would have long-term consequences for human health. And most were dire.


  1. Leren P. The effect of plasma-cholesterol-lowering diet in male survivors of myocardial infarction. A controlled clinical trial. Bull NY Acad Sci 1968:44:1012-1020.
  2. Leren P. The Oslo Diet-Heart Study. Eleven-year report. Circulation 1970;42:935-942.
  3. Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins – an integrated approach to mechanisms and disorders. N Engl J Med 1967;276:34-44.
  4. Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins – an integrated approach to mechanisms and disorders (continued). N Engl J Med 1967;276:94-103.
  5. Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins – an integrated approach to mechanisms and disorders (continued). N Engl J Med 1967;276:148-156.
  6. Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins – an integrated approach to mechanisms and disorders (continued). N Engl J Med 1967;276:215-225.
  7. Fredrickson DS, Levy RI, Lees RS. Fat transport in lipoproteins – an integrated approach to mechanisms and disorders (continued). N Engl J Med 1967;276:273-281.
  8. Frederickson DS. An international classification of hyperlipidemias and hyperlipoproteinemias. Ann Intern Med 1971;75:471-472. 
  9. Goffman JW, Hanig M, Jones HB, et al. Evaluation of serum lipoproteins and cholesterol measurements as predictors of clinical complications of atherosclerosis. Report of a cooperative study of lipoproteins and atherosclerosis. Circulation 1956;14:691-741.
  10. Noakes TD. Ancel Keys’ Cholesterol Con: Part 4. Crossfit Health. Please complete. 
  11. Gofman JW. Diet in the prevention and treatment of myocardial infarction. Am J Cardiol 1958;1:271-283.
  12. The Right Livelihood Foundation. John Gofman 1992 Laureate.
  13. Noakes TD. It’s the insulin resistance, stupid: Part 1. Crossfit Essentials. Please complete. 
  14. Ahrens EH, Hirsch J, Oette K, et al. Carbohydrate- and fat-induced lipemia. Trans Assoc Am Physcns 1961;74:134-146.
  15. Fredrickson DS, Lees RS. Familial hyperlipoproteinemia. In: The Metabolic Basis of Inherited Disease, Second edition. Edited by JB Stanbury, JB Wyngaarden and DS Fredrickson. New York, McGraw-Hill, 1966. 
  16. Kinsell LW, Walker G, Michaels GD, et al. Dietary fats and the diabetic patient. N Engl J Med 1959;261:433-434.
  17. Kuo PT, Bassett DR. Primary hyperlipidemias and their management. Ann Intern Med 1963;59:495-507.
  18. Knittle JL, Ahrens EH. Carbohydrate metabolism in two types of hyperglyceridemia. J Clin Invest 1964;43:485-495.
  19. Kuo PT, Bassett DR. Dietary sugar in the production of hyperglyceridemia. Ann Intern Med 1965;62:1199-1212.
  20. Kuo PT, Feng L, Cohen NN, et al. Dietary carbohydrates in hyperlipidemia (hyperglyceridemia): Hepatic and adipose tissue lipogenic activities. Am J Clin Nutr 1967;20:116-125.
  21. Kuo PT. Hyperglyceridemia in coronary artery disease and its management. JAMA 1967:201:87-94.
  22. Levy RL, Fredrickson DS, Shulman R, et al. Dietary and drug threatment of primary hyperlipoproteinemia. Ann Intern Med 1972;77:267-294.
  23. Dayton S, Pearce LM, Hashimoto S, et al. A controlled clinical trial of a diet high in unsaturated fat – preliminary observations. N Engl J Med 1962;266:1017-1023.
  24. Dayton S, Pearce LM, Goldman H, et al. Controlled trial of a diet high in unsaturated fat for prevention of atherosclerotic complications. Lancet 1968;2:1060-1062. 
  25. Dayton S, Pearce LM, Hashimoto S, et al. A controlled clinical trial of a diet high in unsaturated fat in preventing complications of atherosclerosis. Circulation 1969;40:II-I-II-63. 
  26. Pearce ML, Dayton S. Incidence of cancer in men on a diet high in polyunsaturated fat. Lancet 1971;1:464-467.
  27. Sturdevant RAL, Pearce ML, Dayton S. Increased prevalence of cholelithiasis in men ingesting a serum-cholesterol-lowering diet. N Engl J Med 1973;288:24-27.
  28. Teicholz N. The Big Fat Surprise. Why butter, meat and cheese belong in a heathy diet. Simon and Schuster, New York, NY. 2014. 
  29. Mozaffarian D, Micha R, Wallace S. Effects on coronary heart disease of increasing polyunsaturated fat in place of saturated fat: A systematic review of meta-analysis of randomized controlled trials. Plos Medicine March 23, 2010.
  30. Ramsden CE, Zamora D, Leelarthaepin B, et al. Use of dietary linoleic acid for secondary prevention of coronary heart disease and death. Evaluation of recovered data from the Sydney Diet Heart Study and updated meta-analysis. BMJ 2013 Feb 4;346:e8707.
  31. Taubes G. Good calories bad calories. Fats, carbs, and the controversial science of diet and health. Anchor Books, New York, NY. 2008.
  32. Noakes TD. It’s the insulin resistance, stupid: Part 8. Crossfit Essentials. Please complete     
  33. Mayosi BM, Forrester T. Commentary: ‘Serum cholesterol, diet, and coronary heart-disease in Africans and Asians in Uganda’ by AG Shaper and KW Jones. Int J Epidemiol 2012:41:1233-1235.  
  34. Pett KD, Kahn J, Willett WC, Katz DL. Ancel Keys and the Seven Countries Study: An evidence-based response to revisionist histories. White Paper commissioned by The True Health Initiative. August 1 2017. Available at:
  35. Noakes TD. It’s the insulin resistance, stupid: Part 9. Crossfit Essentials. Please complete 
  36. Harcombe Z. The Seven Countries Study – Part 1.  Response to The True Health Initiative White Paper. August 6th 2017.
  37. Harcombe Z. The Seven Countries Study – Part 2. Response to The True Health Initiative White Paper. August 14th 2017.
  38. Keys A, Menotti A, Karvonen MJ, et al. The diet and 15-year death rate in the Seven Countries Study. Am J Epidemiol. 1986;124: 903-915.
  39. Steinberg D. An interpretive history of the cholesterol controversy: part II: the early evidence linking hypercholesterolemia to coronary disease in humans. J Lipid Res 2005;46:179-190. 
  40. Smith RL, Pinckney ER. The Cholesterol Conspiracy. St Louis, MI: Warren H Green Inc, 1991.
  41. Enig MG. Diet, serum cholesterol, and coronary heart disease. Chapter 3: In: Coronary Heart Disease. The Dietary Sense and Nonsense. An evaluation by Scientists. Mann GV, Ed. Janus Publishing Company, London. 1993; p.36-60
  42. Noakes TD. Ancel Keys’ Cholesterol Con: Part 5. Crossfit Health. Please complete. 
  43. Blackburn H, Menotti A. Major results of the Seven Countries Study. In: Kromhout D, Menotti A, Blackburn H. The Seven Countries Study. A scientific adventure in cardiovascular disease epidemiology. Marjan Nijssen-Kramer, Studio RIVM, Bilthoven, The Netherlands. 1993., pp.159-174.
  44. Skotnicki M. “The dog that didn’t bark:” What we can learn from Sir Arthur Conan Doyle about using the absence of expected facts. Briefly Writing July 25th, 2012.  
  45. Noakes TD, Sboros M. Nutrition on Trial. How the diet dictators tried to destroy a top scientist. Columbus Publishing Ltd, UK. 2019, p.1-488.
  46. Keys A, Menotti A, Aravanis C, et al. The seven countries study: 2,289 deaths in 15 years. Prev Med 1984;13:141-154.
  47. Law EA, Biostatistical pitfalls in studies of atherosclerotic heart disease. Fed Proc 1962;21(suppl11):62-70.
  48. Reisser R. Oversimplification of diet:coronary heart disease relationships and exaggerated diet recommendations. Am J Clin Nutr 1978;31:865-875.
  49. Taubes G. Good calories bad calories. Fats, carbs, and the controversial science of diet and health. Anchor Books, New York, NY. 2008.
  50. Steinberg D. An interpretive history of the cholesterol controversy, part IV: The 1984 Coronary Primary Prevention Trial ends it – almost. J Lipid Res 2006;47:1-14.
  51. Fredrickson DS. The Field Trial: Some thoughts on the indispensable ordeal. Bull NY Acad Sci  1968;44;985-993.

Read Also: