by Tim Noakes
Keys has been heavily criticised for the way in which he selected the countries for his analyses, in particular his choice of just six or seven when data was available for 22 countries.19 A subsequent analysis of 21 of the 22 countries showed that they fell into two distinct groups: a group of six countries with low rates of both fat intake and heart disease, and another group of 15 countries with higher rates of fat intake but very variable rates of heart disease.20 Keys’s inclusion of mostly those countries with low rates of fat intake into the final analysis dramatically increased the probability that he would detect a significant relationship; a relationship that was absent when almost any combination of countries with higher rates of fat consumption was analysed.
Accordingly, P.D.P. Woods concluded: ‘It seems probable therefore that the Seven Countries study suffered from a selection effect of some magnitude. It is prudent whenever very small samples are offered as evidence in support of a hypothesis to make sure that they are fully representative of the population from which they were drawn.’21
For there to be definitive evidence that cholesterol is the direct cause of heart disease, all the strands of evidence must point in the same direction. We now know that all the evidence does indeed point in one direction, but it is the opposite of what is required for Keys’s hypothesis. This evidence was superbly presented by Nina Teicholz and Dr Zoë Harcombe in my trial (see Chapter 13), and was not contested by the complainant’s legal team.
I now briefly review some of the key pieces of evidence that the three of us presented at the HPCSA hearing.
1. Epidemiological evidence
Modern studies show that there is no relationship between the amount of fat, including saturated fat, eaten in different countries and their respective rates of coronary heart disease. In contrast, the intake of polyunsaturated fat is directly related to CHD rates.
As I argued in the Centenary Debate,22 modern data shows that total cardiovascular-disease mortality in a large number of European countries is an inverse function of total animal fat and protein intake, but is linearly related to energy intake from carbohydrates and alcohol, disproving the diet-heart hypothesis (see Figure 17.6 on page 344).23
Figure 17.5
An analysis of 21 countries that Ancel Keys could have included in the Seven Countries Study shows they cluster into two groups: six with low rates of animal fat intake and the remaining 15 with higher rates of fat intake. By selecting four countries with low fat intakes, Keys ensured that he would achieve a significant relationship between fat intake and CHD death rates. Redrawn from P.D.P. Wood, ‘A possible selection effect in medical science’24
Furthermore, those European countries with the greatest prevalence of raised blood cholesterol concentrations have the lowest rates of cardiovascular-disease mortality, disproving the lipid hypothesis. The prevalence of raised blood cholesterol concentrations is a linear function of (increasing) animal fat and protein intake; conversely, countries with the lowest prevalence of raised blood cholesterol concentrations have higher intakes of potatoes and cereals. Interestingly, the prevalence of high blood pressure in these countries is an inverse function of their prevalence of raised blood cholesterol concentrations.
P. Grasgruber et al. reported two other findings that are the converse of what the diet-heart hypothesis predicts:
The prevalence of elevated blood pressures in these countries is an inverse function of total animal fat and protein intake, but is linearly related to energy intake from carbohydrates and alcohol.
The prevalence of raised blood glucose concentrations in these countries is an inverse function of total animal fat and protein intake, but is linearly related to energy intake from carbohydrates and alcohol.
Figure 17.6
Total cardiovascular disease mortality in women in 42 European countries is an inverse function of mean daily consumption of total animal fat and protein (Panel A) and a linear function of carbohydrate and alcohol intake (Panel B). Cardiovascular disease mortality is an inverse function of raised blood cholesterol concentrations in women (Panel C), whereas prevalence of raised blood cholesterol concentrations in men is a linear function of total animal fat and protein intake (Panel D) and an inverse function of potato and cereal consumption (Panel E). The prevalence of raised blood pressure is an inverse function of raised blood cholesterol concentrations in women (Panel F). Redrawn from P. Grasgruber et al., ‘Food consumption and the actual statistics of cardiovascular diseases: an epidemiological comparison of 42 European countries’25
Accordingly, the authors concluded: ‘Irrespective of the possible limitations of the ecological study design, the undisputed finding of our paper is the fact that the highest CVD [cardiovascular disease] prevalence can be found in countries with the highest carbohydrate consumption whereas the lowest CVD prevalence is typical of countries with the highest intake of fat and protein.’26 Thus: ‘In the absence of any scientific evidence connecting saturated fat with CVDs, these findings show that current dietary recommendations regarding CVDs should be seriously reconsidered.’27
Additional data from some of these European countries shows a direct linear relationship between increasing heart-disease mortality rates and the intake of polyunsaturated fats – the ‘vegetable’ oils we are told to eat in place of saturated fats because they are supposedly more healthy (Chapter 7).28
Were Keys to repeat his study today, he would have to conclude (according to his flawed logic) that it is the consumption of ‘vegetable’ oils rich in omega-6 polyunsaturated fats (and perhaps other fats, the toxic effects of which have still to be identified) that causes arterial disease leading to heart attacks.
Figure 17.7
Linear relationship between polyunsaturated fat consumption and cardiovascular disease deaths in men and women in 19 European countries. Redrawn from Credit Suisse, ‘Fat: The New Health Paradigm’29
In Chapter 7, I explained how Keys and his colleagues chose to bury the findings of the Minnesota Coronary Experiment, which found that substituting vegetable oils for saturated fats provided no health benefits, and may even have been harmful, especially for males over 65. The negative findings from that study and from Rossouw’s WHI trial (Chapter 4) should have been sufficient to bury the diet-heart hypothesis. But this did not happen, because the authors of both studies, or their funders, did not see any value in accepting the null hypothesis.
2. It is not a high-fat diet that causes ‘dangerous’ cholesterol to rise
Ultimately, Keys acknowledged that cholesterol in the diet does not cause the blood cholesterol concentration to rise.30 Interestingly, dietary cholesterol comes only from animal products, since plants do not contain cholesterol.
If dietary cholesterol does not cause blood cholesterol concentrations to rise, then logically, according to the diet-heart and lipid hypotheses, animal fats cannot cause heart disease (since cholesterol-containing animal fats do not cause blood cholesterol concentrations to rise). This obvious conclusion was first brought to our attention by Dr Zoë Harcombe.31
It is now clear that the dangerous atherogenic dyslipidaemia that produces arterial disease is caused by a high-carbohydrate diet that produces NAFLD. Heart attack is therefore a disease of carbohydrate, not fat, metabolism. Thus we need to understand the way in which carbohydrates, and not fats, alter blood ‘cholesterol’ (lipoprotein) values.*
3. There is no evidence that blood cholesterol concentrations can predict the risk of future heart attack more effectively than measures of IR/pre-diabetes/T2DM
The lipid hypothesis proposes that the key driver of arterial disease is an elevated blood cholesterol concentration. But evidence shows that the blood cholesterol concentration is such a poor predictor of heart-attack risk that it can, or should be, ignored. Rather, it is T2DM and other markers of IR and metabolic syndrome that are the better predictors.
Table 17.1 Predictors of cardiovascular disease risk
Predictor
Hazard ratio
T
2DM
2.04
Smoking
1.79
Total cholesterol/HDL-C ratio
1.32
Systolic blood pressure (mmHg)
1.31
Apolipoprotein B:A-1 ratio
1.30
Apolipoprotein B (mmol/L)
1.24
Total cholesterol (mmol/L)
1.22
Triglyceride (mmol/L)
1.19
HDL-C (mmol/L)
0.83
Note that this is associational data. Independent scientists agree that only a hazard ratio of well over 2.0 might suggest a causal (not associational) relationship between two variables. Thus, in this table, the only variable that might be causally linked to heart-attack risk is T2DM. A hazard ratio of only 1.22 indicates that cholesterol can be exonerated of any role in the causation of heart attack. Data from E. di Angelantonio et al., ‘Lipid-related markers and cardiovascular disease prediction’32
But perhaps the best evidence that cholesterol has little to do with heart disease comes from the very condition, familial hypercholesterolaemia, that cardiologists like to promote as definitive evidence that cholesterol drives the process of coronary atherosclerosis.
As I first suggested in Chapter 7, the evidence is the opposite; blood cholesterol concentration is unable to predict who with FH will suffer from heart disease and who will not. The evidence shows that the longer people with FH survive, the more likely it is that they will have a normal life expectancy.33 Even more perplexing for the cholesterol advocates is the recent finding that blood cholesterol concentration has essentially no predictive value for heart attack or stroke in the next 5–10 years in people with FH.34 Rather, it is the usual suspects – age, previous history of atherosclerotic cardiovascular disease, obesity, diabetes, premature familial atherosclerotic CVD (all markers of IR) – that best predict risk of future heart attack in those with FH.
Table 17.2 Predictors of CVD risk in persons with FH
Variable
Hazard ratio
p values
Age
> 60 years
12.81
<0.001
History of ASCVD
6.64
<0.001
Body mass index
Obesity
6.12
<0.001
Age
30–59 years
5.88
<0.001
Body mass index
Overweight
4.69
<0.001
Diabetes mellitus
3.45
<0.001
Patient on ezetimibe
3.42
<0.001
Patient on maximum combined therapy
2.97
<0.001
Patient on maximum lipid-lowering therapy
2.88
<0.001
Male
2.76
<0.001
Premature familial ASCVD history
2.66
<0.001
Calculated pre-treatment LDL-C
> 160 mg/dl (4.5 mmol/L)
2.47
0.40
Patient on maximum statin dose
2.08
<0.001
Active smoking
1.77
0.004
ASCVD = atherosclerotic cardiovascular disease
Note that a hazard ratio of greater than 2.0 suggests an increasing probability that the risk predictor acts as a directly causal, rather than as an associational (but not causative), factor. Diabetes and obesity, both markers of IR, are therefore strongly implicated as directly causal factors for heart disease in these persons with FH. Note that p values <0.05 are not considered to be significant. A blood cholesterol concentration of >160mg/dl (4.5mmol/L) is the only variable on this table that is not a significant predictor of future cardiovascular disease. This dislodges a cornerstone of the lipid hypothesis
There are a number of unexpected observations on this table. First, that the predictors of future CVD in persons with FH are not greatly different from those in persons without the condition (Table 17.1), so that diabetes and obesity are again key risk factors.
Second, that the use of cholesterol-lowering statin drugs (ezetimibe; maximum combined therapy; maximum lipid-lowering therapy; maximum statin dose) are all significant predictors of future CVD risk in these patients. This probably does not mean that these drugs are the direct cause of future heart disease, but rather that those at greatest risk of future CVD are prescribed the most drugs. It might also mean that these drugs are rather less effective in preventing future heart disease in these patients than is usually acknowledged.
It’s the fatty liver disease, stupid
We have known for some time that the added risks associated with obesity depend, in part, on where that extra fat is stored in the body. Fat that accumulates under the skin (subcutaneous fat) appears to be far less unhealthy than fat that accumulates within and between the organs in the abdomen – so-called visceral obesity. A group of hepatologists (liver specialists) have now gone one step further to show that the real killer in visceral obesity is the fat that accumulates within the liver causing NAFLD, a disease that is now also reaching epidemic proportions among people eating the ‘displacing foods of modern commerce’.
By measuring the amount of fat in the livers of people with different metabolic conditions, including T2DM and NAFLD, Fernando Brill and his colleagues at the University of Florida have established that it is the presence of NAFLD, and not the overall level of body fatness, that predicts the presence of atherogenic dyslipidaemia.35
Their work therefore establishes that it is IR and NAFLD, and not obesity per se, that produces the abnormal metabolic state (atherogenic dyslipidaemia) that causes heart disease in those with IR and the related metabolic syndrome. The following are the metabolic features of atherogenic dyslipidaemia present in those with NAFLD and IR who eat highcarbohydrate (>25 grams per day) diets with high omega-6 to omega-3 ratios:
Elevated blood HbA1c levels;
Elevated fasting blood insulin levels;
Elevated fasting blood glucose levels;
Hyperinsulinaemia and hyperglycaemia (elevated blood glucose levels) in response to carbohydrate ingestion;
Low blood HDL-cholesterol concentrations;
High blood triglyceride concentrations;
Elevated numbers of small dense LDL-particles;
Elevated blood Apolipoprotein B concentrations; and
Elevated blood gamma-glutamyl transferase (GGT) activity (indicating the presence of NAFLD).
But the absolutely key point is that it is dietary carbohydrates, especially fructose36 present in sucrose, high-fructose corn syrup and fruit, allied to high intakes of unsaturated fats (oleate),37 and not dietary saturated fats, that cause NAFLD. In part, this is because the major pathway by which fructose is cleared from the bloodstream is its conversion to liver fat (triglyceride). When the liver, skeletal muscles and adipose tissue are insulin resistant, a continuing high-carbohydrate diet causes liver fat to accumulate, worsening the IR. Thus, a vicious cycle results: as NAFLD develops, IR worsens, hyperinsulinaemia increases, atherogenic dyslipidaemia deteriorates, and the seeds for the chronic diseases of obesity, diabetes, heart disease and perhaps cancer and dementia are sown.
Importantly, caloric restriction, whether on a high- or low-carbohydrate diet, reduces liver fat in those with IR,38 as can removing sugar, high-fructose corn syrup and refined carbohydrates from the diet.39 Dietary carbohydrate restriction produces a significantly greater effect than caloric restriction.40 Importantly, it is carbohydrate overfeeding that causes NAFLD, and this effect can be rapid, causing a more than tenfold greater increase in liver fat than in body weight within just three weeks.41 In contrast, eating an LCHF diet reverses all the metabolic abnormalities of atherogenic dyslipidaemia,42 presumably also by reversing NAFLD.
Thus, dietary carbohydrates, and not
dietary fats, are the direct cause of this group of chronic diseases in people with IR. In contrast, a high-fat diet combined with carbohydrate restriction can reverse many of these conditions.43
Third, the one factor that does not predict future heart disease is an elevated blood cholesterol concentration. Remarkably, in both those with FH (Table 17.2) and without FH (Table 17.1), the least effective predictor of future risk is a raised blood ‘cholesterol’ concentration.
These findings in patients with FH effectively disprove the lipid hypothesis and the ‘cholesterol-load’ theory, which holds that the more cholesterol in the blood and the longer that cholesterol has had a chance to damage arteries, the greater is the individual’s risk for future CVD. Rather, it is the conventional risk factors, specifically the presence of the markers of IR and perhaps abnormal blood clotting on a genetic basis,44 that predict the development of arterial disease in those with FH.
So what is it about T2DM and IR that increases the risk for the development of disseminated obstructive arterial disease?
First, we have to suspect it is not blood cholesterol concentration, because these levels are ‘elevated’ in persons with T2DM whether or not they have or will develop CHD.45 Ignored in that study are the much higher blood triglyceride concentrations (3.1 mmol/L) in people with T2DM who suffered heart attacks. This suggests that it is the presence of a fatty liver – the condition of NAFLD in people with IR who persevere in eating high-carbohydrate diets – that produces the atherogenic dyslipidaemia that causes the disseminated obstructive arterial disease that characterises T2DM.