Saturated Fats And Bad Science
The UK’s Scientific Advisory Committee on Nutrition (SACN) released its Report this week on ‘Saturated Fats and Health’. It recommended retaining the current guideline that dietary contribution of saturated fats be limited to no more than ten percent of energy intake. It did so on the basis of a ‘review’ of the available evidence, and as such offers us a useful opportunity to examine how the nutrition mainstream carries out its ‘science’, how they arrive at such poorly reasoned and erroneous conclusions.
The SACN represents one of the pillars of the nutrition establishment. Its members include some of the foremost academics from the United Kingdom, Ireland, and Europe. Chair of the committee is one Professor Paul Haggarty, from the Rowett Institute of Nutrition and Health at the University of Aberdeen. His research interests centre on the relationship between nutrition, genetics and epigenetics, one of the cutting edges of mainstream inquiry and far removed from the long settled question of whether saturated fats are bad for our health. This may account for some of the absence in rigour and general sloppiness of thinking that runs throughout the Report.
The motivation behind its publication in 2019 is also revealing. We are living in the midst of an obesity epidemic, the population is getting fatter and sicker by the day, the official dietary guidelines have presided over this developing health disaster, none of which has any connection with the Committee’s decision to conduct a review of the recommendation on fat intake. Instead, it is purely self-serving, its aim is to shore up the current position in the face of a growing recognition that its fundamental premise – the diet-heart hypothesis – is deeply flawed, and that its entire approach to the field of nutrition is built on a shonky foundation.
This, however, is never stated explicitly. The real context to this Report, the fact that the diet-heart hypothesis is currently under siege, is carefully hidden from view. At no stage is there any mention of the controversy surrounding this idea, even less that there exist competing hypotheses as to the cause of cardio-vascular disease, that the field is highly contested. In place of such an admission, the SACN puts itself forward as a selfless defender of scientific truth – the original guideline originated from the Committee on the Medical Aspects of Food Policy (COMA) back in 1991, was last reviewed in 1994, since then ‘a considerable body of research has been published’ (p.xvi), perhaps it is time to take another look, just to be sure…
The Committee’s deliberations take place, therefore, in a self-imposed vacuum. As I intend to show, this is the FUNDAMENTAL FLAW in their method, it is THE main factor why their ‘science’ turns out to be of such poor quality. For us this is helpful, because it is a consistent feature of the mainstream position (see here for another example) that it is both unwilling and unable to engage in genuine debate, choosing instead only to engage with one side of the argument (see here too). This failing does not only cast a shadow over their personal integrity as scientists, it also fatally undermines their science. This is the essential point I wish to demonstrate here.
In this respect, the situation as it applies to the SACN is worse than it was with COMA. For at least that committee was open in its declaration that the basis of its recommendation lay in the notion that, “increasing or decreasing the contribution of saturated fats to dietary energy is followed by a rise or fall in serum low density lipoprotein (LDL) cholesterol and in the commensurate risk of coronary heart disease” (p.xvi). The current committee simply presumes this, but in doing so it commits its most basic error – because it is precisely this theory, the diet-heart hypothesis – that needs to be placed under review. In other words, the Report ASSUMES precisely what it has to PROVE.
In the light of this, it is hardly surprising that the SACN ends up concluding that COMA was correct and the guideline should stand. It is the product of circular reasoning.
Evidence of this can be seen in the points of evidence the Committee cites in support of its final recommendations, the contents of its own ‘review’. These DO NOT support its final position, they directly contradict it. This is Bart Kay’s criticism too, in his excellent video on the Report (see here). This becomes apparent once we provide a few examples.
There was adequate evidence from RCTs that reducing intake of saturated fats had no effect on CVD mortality.
There was adequate evidence from PCS that intake of saturated fats was not associated with CVD mortality or CVD events.
There was adequate evidence from RCTs that reducing intake of saturated fats had no effect on CHD mortality.
There was moderate evidence from RCTs that reducing intake of saturated fats reduced risk of CHD events.
There was moderate evidence from PCS that lower intake of saturated fats was associated with a lower risk of CHD mortality and events.
Moderate in this context means a failure to find ‘adequate’ evidence, it is not a positive finding but a negative one.
There was moderate evidence from PCS that lower intake of saturated fats was associated with a lower risk of CHD mortality and events.
There was adequate evidence from RCTs of no effect between intake of saturated fats and total strokes. There was limited evidence from PCS that lower intake of saturated fats was associated with the increased risk of total stroke primarily in East Asian populations living in East Asia. There was adequate evidence from PCS for no association between lower intake of saturated fats and ischaemic strokes. However, there was limited evidence from PCS that lower intake of saturated fats increased the risk of haemorrhagic strokes in Japanese Asian populations living in Japan.
There was insufficient or no evidence from RCTs of any effect of substituting saturated fats with PUFA or MUFA on strokes. There was no evidence from PCS.
There was no or insufficient evidence from RCTs to determine any effect of reducing intake of saturated fats and risk of type 2 diabetes, or the effect of specific substitutions for saturated fats. (p.187)
Of course, I’ve cherry picked these examples to prove my point. There are other recommendations that do support their final conclusion. Nevertheless, given the list shown here, they would have to be pretty strong to offer a sufficient counterweight. In reality, however, they boil down to one,
There was adequate evidence from RCTs that reducing intake of saturated fats reduced the risk of CVD events. (p.187)
This was the finding that swung it, in the words of the overall conclusion,
“There were significant relationships between intake of saturated fats and CVD and CHD events, but not CVD and CHD mortality. SACN noted that, irrespective of the lack of evidence for an effect on mortality, non-fatal CVD and CHD events have a serious adverse impact on health and quality of life.” (p.196)
Heart attacks are bad news, even if you survive them, so the idea is ‘adequate’ evidence that reducing saturated fat intake could help you avoid one is enough to justify this recommendation. This in turn rests on some assumptions that the Committee acknowledge at the end of their Report,
“The evidence indicates that reducing saturated fats reduces the risk of CVD and CHD events, lowers total, LDL and HDL cholesterol and improves indicators of glycaemic control. The evidence also indicates that reducing saturated fats is unlikely to increase health risks for the general UK population. SACN concluded that reducing population average intake of saturated fats from current levels of intake to no more than about 10% of [total] dietary energy would result in health benefits to the population.” (p.196)
Here the diet-heart hypothesis makes a reappearance, but only as an assumption, at no point has the evidence submitted even CONSIDERED its validity, one way or another. Neither has any case been made for the 10% figure, all that seems to matter is that it is lower than the current intake.
The argument ultimately rests on an association the Committee considers to have been established, between lower saturated fat intake in the diet and fewer CVD or CHD events. When we investigate this relationship, we shall see that it is tenuous at best. This is not the main problem, however, the real issue lies in the method being applied. A correlation of this kind is MEANINGLESS, unless it is connected to a HYPOTHESIS, an idea of the causal mechanism that links one input – saturated fat – to one output – a CVD or CHD event. It is not a question of whether there is an association between the two, but what it MEANS if there is, how we are to understand it. This is why the SACN’s method is so wrong. In their case, there IS a mechanism in play, it is the one outlined in the diet-heart hypothesis, but it is THIS theoretical postulate that the evidence needs to confirm, NOT the association between input and output. The Committee are answering the WRONG question in their review.
The problem is that the diet-heart hypothesis has TWO steps, not one. These are –
Eating saturated fat raises blood serum cholesterol
Raised cholesterol causes atherosclerosis
We can add a third unstated premise, which is that atherosclerosis is what causes heart attacks. It turns out that this is contested too. More on this in a moment.
The point is that it is not enough to establish a link between point 1 and heart disease – we also have to establish the connection between points 1 and 2. At no stage in the Report is this question even ADDRESSED. In fact, when we get to the sections on cholesterol, the picture descends into complete confusion, reflecting the current disarray in the diet-heart hypothesis camp. If you remember, the original claim made by Ancel Keys was for total cholesterol, as it was for fat in general, both saturated and unsaturated. Over time, overwhelming evidence against this theory has forced its proponents to reduce its scope to saturated fats and to LDL. One of the results of this shift is that the literature is all over the place, with some studies examining total fat intake and others looking at cholesterol in general, while others specifically concentrating on LDL. This is one reason why meta-analyses struggle so much to build a coherent picture of where the evidence is pointing. The SACN review really draws on little more than four main studies for the bulk of its findings, and even these have all kinds of challenges that their authors acknowledge when you read them directly.
This incoherence can be seen in the Report. The authors are aware of the problem, here is how they tackle it,
“The Committee on Medical Aspects of Food Policy (COMA) based their recommendations for saturated fat intakes on the effect on LDL cholesterol concentrations and the link with morbidity and mortality. SACN endorsed the COMA conclusions that there is strong evidence that LDL cholesterol and other blood lipids are causally related to morbidity and mortality (Ference et al, 2017). The use of the serum total cholesterol:HDL cholesterol ratio has been proposed as a more sensitive and specific coronary heart disease (CHD) risk predictor than other individual cholesterol measures (Assmann et al, 1996; Kinosian et al, 1995; Stampfer et al, 1991). It has been found to be a predictor of CHD at all ages in women and the only lipid predictor independently related to CHD in men 65 to 80 years old (Castelli et al, 1992). This ratio is currently routinely used as the most discriminatory lipid risk marker to predict long-term CVD risk (QRISK 2 assessment)17 in the clinical setting (Hippisley-Cox et al, 2017). SACN considered the evidence on all lipid markers.” (p.62)
They considered it, and then what ?
The Report does not say.
At first I thought this was an editing error, a paragraph missing, but it seems not. Rather than resolving the issue, they leave it hanging. This is what I mean by sloppiness.
The result can be seen later, when it comes to evaluating the evidence on HDL and total cholesterol – does it want to be higher or lower ? The evidence seems to point towards lower (p.114), that is, reducing saturated fat intake is associated with a decrease in HDL, both in absolute and relative terms. HDL is considered to be ‘the good cholesterol’ among most mainstream thinkers, connected to its role in ‘reverse transport’, that is, the removal of cholesterol from the body’s tissues back to the liver. This interpretation appears to be endorsed by the Committee’s own diagram on page 46, and yet in their summary they point to a reduction in HDL as a good thing and supportive of their recommendation (p.196).
This gives rise to a certain suspicion, that any association of whatever kind between fat and cholesterol was always going to be placed in a bad light whatever, no matter how little sense it made. That uneasy feeling is reinforced when we turn to the sections on diabetes, cancer, and Alzheimers. None of these find any traceable connection between fat intake but that is hardly surprising. The real question is – why are these even discussed ? It comes across as the Committee going on a fishing expedition among the known health ‘risk factors’ to see what associations it could uncover in support of a ‘guilty’ verdict on saturated fat. Evidence suggesting the opposite, as in the case of stroke (see above), is simply ignored when it comes to the final reckoning on the benefits or otherwise of reducing dietary fats. The problem boils down to this – what is the hypothesis being investigated ? When it comes to these diseases, there is none, there is no mechanism being postulated that places saturated fat as a causal factor. These sections of the Report appear within a methodological vacuum, they serve no scientific purpose, they are evidence of NOTHING whatsoever, one way or another.
One possible exception is blood pressure, which we could expect to be elevated if the diet-heart hypothesis is correct. The Report finds no such association, which is, if we follow the logic, evidence AGAINST this idea. The SACN, however, make no comment on this and simply move on.
This oversight is either dishonest or else a travesty of the scientific method. The Committee behave as if all they need to do is find correlations and their job is done, any will do, so long as they link saturated fat with something bad. This is not science, it is simply an institution closing ranks to protect its own.
what would a genuine review look like ?
What would a genuine review of the evidence look like ? The first thing would be to state accurately what it is that is under investigation, that is, the diet-heart hypothesis that provides the theoretical underpinning for the current guideline.
The second thing would be to acknowledge that the terrain is contested, alternative hypotheses exist, and what is required is an examination of the evidence to see which one comes out strongest. This involves an entirely different standard of evidence, and it also means looking for different kinds of evidence – it means different questions have to be answered than those covered in this Report.
We can see this if we take the example of LDL, and its association with heart disease. The stumbling block for the mainstream position lies in the concept of reverse causality, and its inclusion in at least one of the rival theories. This means that the cause effect relationship may be the opposite, namely that it is CVD that leads to an elevated LDL-C count, not the other way round. In the face of this hypothesis, what is needed is evidence that will differentiate between the two explanations for this association, so we can see which one is confirmed. This calls for some thought and care, as often the evidence is ambiguous. A case in point is the mechanism described by the European Atherosclerosis Society, one of the key supporters of the diet-heart hypothesis, in their account for why PCSK-9 inhibitors and statins are effective (see here p.7). This highlights the enhanced role of LDL receptors in the liver, which take LDL particles out of circulation. This idea, although by no means generally accepted, is consistent with a position that gives LDL a role in reverse transport but sees the problem of atherosclerosis as one of oxidative damage to LDL particles causing them to not be recognised in the liver. The evidence supports both views. In any case, one of the contradictions for the SACN in the EAS view is that it renders the amount of saturated fat in the diet largely irrelevant.
When we look at some of the competing hypotheses, we get an idea of what kind of evidence would suit our purposes. One of the current theories focuses on LDL particle size, Pattern A versus Pattern B, and sees this as more relevant than LDL-C levels in blood serum. It would not be difficult to devise measures to test this idea for its association with atherosclerosis. Another theory, put forward by Malcolm Kendrick, redefines heart disease not as a problem of sclerosis, but thrombosis, a problem of blood clotting (see here). This could be detected by looking closely at the progression of the disease to see which model fits best. Yet another hypothesis alters the causality even further, arguing that heart failure comes first, and it is the drop in blood flow that leads to blockages inside the coronary arteries. Consistent with several of these theories is the idea that atherosclerosis starts from the outside of the artery and moves in rather than inside out, in which case the level of LDL-C in the serum is neither here nor there.
a null hypothesis
It is only in the light of such hypotheses that the available evidence can be assessed properly. A good example also comes from Malcolm Kendrick, who has argued that heart disease might have NO connection at all with diet (see here). This idea is important to bear in mind when we come to the evidence contained in the SACN Report. When we look at it closely, the overriding impression it leaves is that not only is it very weak, it is highly contradictory. Here for instance,
“In summary, the most comprehensive systematic review with meta-analysis by Hooper et al (2015) concluded that there was no effect of saturated fats on CVD mortality using both random-effects and fixed-effect models and this was confirmed by other sensitivity analyses. This evidence was considered as adequate due to the high number of studies and reported CVD cases. However, there was adequate evidence from 11 RCTs for a significant reduction in CVD events and the effect was observed using both random-effects (17%) and fixed-effect models.” (p.70)
How are we to deal with such conflicting results ? This is not an isolated case, similar findings run right through the Report. The response of the Committee seems to be to try and weigh up these different results, and decide on the balance of evidence, but this is an incorrect method. It does not take into account that opposing results from well-conducted studies do not have equal weight, their combined effect is to NULL the hypothesis. We see this in the Hooper et al (2015) meta-analysis, which is heavily relied on by the SACN. Included among their studies, all of which were rigorously selected, is the Sydney Heart Study of 1978 (see here), in which lowered saturated fat intake was associated with a HIGHER mortality rate from heart disease. If the study design and execution is not at fault here, then these results REQUIRE EXPLANATION. It is simply not adequate to discount them with other studies that say the opposite. Unless some convincing account for the Sydney study can be provided, we are left with strong support for Kendrick’s NULL HYPOTHESIS, as this is the BEST way to explain such wildly inconsistent results.
The same applies to the Sachdeva et al study in 2009, of 137,000 patients hospitalised with a CVD event, who were found to have LOWER than average levels of LDL. (see here) This alone is enough to sink the diet-heart hypothesis. A major study used by the SACN, the Siri-Tarino meta-analysis, contradicts the Report’s pivotal finding on CHD events (see here). This too is devastating. In the face of this, the EAS has tried to drown such evidence out by weight of numbers (see here), throwing no less than two million participants from its favoured studies into the ring. The SACN uses this to argue ‘case closed’, but this is false – a better explanation for the discrepancies between these large studies is that LDL just does not play that important a role in heart disease, one way or another.
This matters when we start to bring some other metrics into consideration. One of these is all-cause mortality. A key finding of the de Souza meta-study, also critical to the SACN Report (see here), is that saturated fat has no effect on the overall rate of death. This stands in contrast to trans-fats for which an association could be found. The Committee makes no mention of all-cause mortality anywhere in its findings, an omission which raises eyebrows in light of its cavalier assertion that there are no risks from lowering saturated fat intake, only benefits. Widespread evidence exists, including the famous Framingham study, that fats are essential for longevity.
Other measures of health do appear in the Report, but their significance is not addressed. One of these is obesity, which the Committee acknowledges has risen sharply since the COMA recommendation was made (p.65). This raises the obvious question as to whether the guideline has contributed to weight gain among the population. The SACN concede that obesity is a risk factor, but go no further in tackling this issue. Furthermore, they dismiss it as purely a ‘lifestyle’ matter (p.65), again ignoring alternative hypotheses as if they don’t exist, the insulin model for example.
The Report runs into another problem when we look at cancer. It admits that the aetiology of cancer is ‘complex’ (p.65), and lists some of the known risk factors such as smoking. It then argues that reducing these risk factors would be beneficial, following the same line of reasoning as it does for saturated fats throughout. The difficulty with this lies in the non-linear causality that applies in complex systems – reducing any one factor in isolation has unpredictable consequences which can be negative. This is a problem with the whole ‘risk factors’ approach and links in to the point about obesity above, and all-cause mortality. Even if the diet-heart hypothesis was correct, reducing a single risk factor such as saturated fat intake may not be a good thing. A reduction in CVD might be accompanied by an increase in deaths from other diseases, and this is what the evidence seems to suggest when it comes to cancer. Before renewing the 1994 recommendation, this would certainly be an issue that would need some resolution, if the SACN were serious in their intent.
PUFAs given a free pass
The Committee’s methodology comes into question from yet another angle when we look at its second key recommendation, that saturated fat intake be substituted by polyunsaturated fats (PUFAs) (p.202). This appears to be justified on the grounds of the same diet-heart hypothesis, with evidence showing that PUFAs reduce serum cholesterol levels. At no point are any of the alternative theories of heart disease, including those that posit vegetable seed oils as a key factor, taken into consideration, they are simply ignored. The closest that the SACN come to an acknowledgement that there might be an issue is when they recognise a limitation in the research on PUFAs is the failure to distinguish between the effects of Omega 3s and Omega 6s (p.40). The implications of this for its recommendation are not taken up beyond this passing reference. PUFAs are given a pass purely on the basis of their ability to reduce LDL.
Underlying this is a genuine imbalance in the way the Report is structured, given its second of only two dietary recommendations on substituting saturated fats with PUFAs. The overwhelming focus throughout is on saturated fats, and yet the entire thrust of the SACN position is to promote PUFAs. These do not undergo anything like the same examination, no evidence is offered or evaluated on their potential impact in relation to any of the risk factors. This is a serious shortcoming of the Report.
Another weakness consists of its treatment of dietary fats, and their combination with carbohydrates. When the Committee lists the foods that make up saturated fat intake, they include cereals (21%), sugars (5%), nuts (2%), and vegetables (6%). The significance lies in another competing hypothesis, namely that it is when fats are combined with carbohydrates, especially in processed products like biscuits, or cereal grains, that they do their greatest damage, in particular by promoting insulin resistance. Because the Report does not take this into account, it is open to the charge that all of its data is compromised by this confounding factor, that none of its claims for the effects of reduced saturated fat have any validity. This too is a serious deficiency that the SACN would have been well to address, rather than pretend it did not exist. This is relevant for Hooper et al’s findings, in many ways central to the Report, which compare reduced fat intake against a ‘normal diet’, similar to the one described above. This leaves the real question open, which is how does the recommended low fat diet measure up against other ways of eating that exclude processed foods high in both fat and carbs ? This would be a REAL test of the guideline, not the vacuous comparisons actually being made in the studies looked at.
This is the guts of my argument – that the problem with the Report is that it chooses to operate in a vacuum, it ignores the fact that there are competing hypotheses, this is contested terrain, and that therefore the task of genuine science is to seek evidence that can allow us to choose between these hypotheses. Instead, like our mainstream textbook authors (see here) they conduct an artificial narrative, failing to address any of the real issues at stake in the debate. Worst of all, they presume what they have to prove, and as a result the ‘scientific’ evidence base they generate is totally worthless. Those of us who are genuinely interested in the truth will have to look elsewhere.
Finally, I want to draw attention to the nature of my critique here. I have not gone down the usual path of criticising epidemiological studies, arguing correlation is not causation, or that such studies are not conducted in a rigorous manner. I do not share the view that RCTs which isolate a single variable have any more validity than observational studies of the kind the SACN cite, both have extremely limited value for reasons I have argued elsewhere (see here and here). I think this line of argument is more worthwhile, in part because it gives us an idea as to what we need to do in this space, namely succeed in cracking the debate wide open, in raising a general understanding that the mainstream position is under challenge, that viable alternatives exist, and that real science at this point in time requires a proper and full debate, and a search for the critical evidence that can settle the truth of the matter. I hope you find that conception convincing, if so lets spread the idea around and push to make it happen.
If you like this blog and want to support it, I have a Patreon account (link here)