Briefly, the cholesterol hypothesis proposes that by getting into arterial walls and causing atherosclerosis (plaque buildup in arteries), cholesterol increases the risk of cardiovascular disease (CVD) and cerebrovascular disease (strokes). However, according to a perhaps unlikely source – the American Heart Association: “Exactly how atherosclerosis starts or what causes it isn’t known.” That’s an admission that has to be made, because atherosclerosis is a heterogeneous disorder, involving vascular, metabolic and immune systems, exacerbated by genetic, environmental and lifestyle factors and, potentially, by metabolic or other co-morbidities. The susceptibility for atherosclerosis may well be set in childhood.
Nevertheless, it is cholesterol that authorities strongly associate with atherosclerotic CVD risk and, conveniently, there are drugs that can lower cholesterol concentration in the blood. With increasing numbers of people prescribed them, the cholesterol message is reinforced. Furthermore, even if cholesterol doesn’t cause atherosclerosis, it’s accepted that it contributes to plaque progression in a causal way, and therefore that lowering the availability of cholesterol in the blood is a good thing.
This concept was proposed in 1856 by Rudolf Virchow, a German pathologist. It used to be called the Filtration Theory – the notion that cholesterol moves into arterial walls by filtration across cell membranes that have been damaged in some way, and the higher the blood concentration the more this occurs. A role for inflammation was also known at this time.
The early years
Cholesterol itself was first identified in human gallstones around 1784 (our liver eliminates cholesterol via. the gall bladder, where it can crystallise under certain circumstances). However, it wasn’t until 1815 that it was isolated, purified and given the name ‘cholesterol’. Subsequently, cholesterol was identified in human blood and, sometime thereafter, in atherosclerotic arteries. This was the first circumstantial evidence that associated the newly discovered blood cholesterol with arterial plaque, although it did not show causality.
Over time, it would be recognised that there was more than just cholesterol in arterial plaques – e.g. necrotic cores, calcium deposits, accumulated oxidised lipids and phospholipids, lipid droplets, inflamed smooth muscle cells, cells of the immune system, foam cells, platelets and fibrin. However, cholesterol filtered into the collective consciousness as the principal causative agent.
The concept received a boost in 1913 when a Russian (Nikolai Anitschkow) fed cholesterol to rabbits, and they went on to develop atherosclerotic plaques containing cholesterol and inflammatory cells. This suggested that cholesterol caused atherosclerosis. However, serious research into cholesterol and atherosclerosis didn’t really get underway until around the 1940s, partly because it was not a widespread clinical problem, and because it was thought to be a simple consequence of ageing in some individuals.
The rabbit model of atherosclerosis
Anitschkow added around 2% cholesterol to normal rabbit diets. He reported that this increased blood cholesterol concentration within days – from the normal range of ~1 mmol/L to an astonishing ~50 mmol/L (a 5,000% increase). It took only a few weeks for the rabbits to develop atherosclerosis under this extreme condition. For comparison, high cholesterol in the human starts around 6 mmol/L.
Using rabbits to model humans in dietary studies overlooks an important difference between these mammals. Rabbits are herbivores and, there is no cholesterol in any plant matter. Therefore, dietary cholesterol is unknown to the evolutionary biology of rabbits, and they have not evolved an efficient mechanism to eliminate dietary sources. The result is that blood cholesterol in the cholesterol-fed rabbit reaches uncontrollable levels, and it should be no surprise that problems would arise from that. Surely, increasing anything else by 5,000% would too.
Which makes the obvious point that all animals (including us) would be healthier if they ate a diet appropriate to their evolutionary biology. The controversy for humans is that we have become so displaced from our evolutionary environment that we no longer know what our natural diet is. We can expect, however, that it wouldn’t include man-made processed food, refined grains, plant oils and purified sugar – the ‘foods of commerce’.
As it’s been noted – the human is the only animal clever enough to make its own food, and foolish enough to eat it.
Does cholesterol cause atherosclerosis and CVD in humans?
If it is accepted that the most definitive evidence for causality comes from a randomised controlled trial (RCT), then we don’t know for certain. This is because the ideal trial design would compare two groups of healthy people, matched for CVD risk-factors and with moderate-low levels of cholesterol, in which cholesterol concentration was increased in one group to see if they develop more CVD than the control group. Given the widespread presumption that cholesterol causes CVD, no Human Ethics Committee would approve such a study. Ironically, the hypothesis that cholesterol causes CVD stops the hypothesis from being tested.
Therefore, the hypothesis has been addressed indirectly and the evidence supporting it is circumstantial. There is also evidence not supporting it. I will start with the statin RCTs.
A mainstay of the cholesterol hypothesis comes from a RCT design that reduces the frequently-called ‘bad’ cholesterol (low-density lipoprotein cholesterol, LDL-C) with statins in one group, and tests whether there are fewer CVD events than in a placebo group. Mostly, but certainly not always, primary prevention trials show that statins can have a small effect on CVD events, usually about 1% or less (in absolute terms, see Lipitor), although there remains much controversy.
These data can be understood to support the cholesterol hypothesis, although the small effect size indicates that LDL-C does not strongly modulate clinical outcome. It is also noted that statin trials combine a number of types of CVD events, some of which are relatively minor (e.g. stents), in order to show a benefit, and they rarely show a benefit for all-cause mortality or even CVD mortality.
As well, we cannot be certain that statins exert a clinical effect by lowering LDL-C. Statins do more than reduce cholesterol. For example, they are known to be mildly anti-inflammatory, and we know that inflammation is central to CVD. An anti-inflammatory drug (Canakinumab) can somewhat reduce CVD events without any effect on LDL-C.
Also, statins act by inhibiting cholesterol synthesis early in a long and complex pathway, and they reduce not only cholesterol, but other important substances produced by that pathway. One of these are the isoprenoids, which are precursors for ubiquinone and heme-A (involved in energy metabolism), and prenylated proteins (attachments to proteins). Some of this may explain certain adverse effects of statins (muscle pain and fatigue), however, there is also some evidence that a modest reduction in protein prenylation may be cardioprotective.
Hence, we cannot definitively draw causality for the cholesterol hypothesis from these RCT designs, the reduction in LDL-C could still be only an association because multiple mechanisms are in action.
The Bradford-Hill criteria
There are some circumstances in which it is statistically acceptable to draw causation from association. This is how we have concluded that smoking causes lung cancer, without a definitive causally-designed RCT. They are known as the Bradford Hill criteria, and they were proposed in 1965 by Sir Austin Bradford-Hill, a statistician who had an interest in the smoking-cancer connection. There are 9 criteria, and the cholesterol-CVD data fails nearly all of them.
I have written about concluding causation from association before. As a reminder, there are two errors that can be made. First, causation the wrong way around – ambulances associate with road accidents, but it would be wrong to conclude that ambulances cause road accidents. Likewise – is cholesterol in arterial plaques because it caused them, or because it was sent to try and help with them? Second, the presence of a co-factor that associates with both variables – if C causes A and C also causes B, then A and B will be associated, but not causally in either direction. I gave a real example (from the literature) of this error being made in the red meat post.
Cholesteryl ester transport protein (CETP) inhibitors are a new class of drug, not related to statins and that do not act by inhibiting the cholesterol synthesis pathway. This drug raises the ‘good’ cholesterol (high-density lipoprotein cholesterol, HDL-C) while lowering LDL-C, resulting in a pattern much advocated by leading medical authorities for reducing CVD.
CETP facilitates the transfer of cholesterol (and triglycerides) between HDL and LDL particles. Normally, the LDL accept cholesterol from HDL, donating a triglyceride in return. The overall effect is to reduce HDL-C and increase LDL-C (i.e. contrary to medical authority objectives). However, drugs that inhibit CETP stop this from happening, leaving the cholesterol in the HDL particles and withholding it from LDL particles, thus keeping HDL-C high and LDL-C low.
While multiple CETP inhibitor trials have shown that HDL-C does go up and LDL-C down with these drugs, often substantially, they have consistently failed to reduce CVD events or all-cause mortality. The most recent trial, of the CETP inhibitor Anacetrapib, did show a small but statistically significant decrease in CVD events, however, it was too small to be clinically meaningful. Consequently, Merck, the manufacturer of Anacetrapib, announced just two weeks after the trial was published that it would not seek regulatory approval for Anacetrapib and in doing so abandoned the drug and their investment. After a series of costly failures, it seems that drug companies are set to give up on CETP inhibitors as a drug class.
However, the CETP studies have highlighted a crucial point – high HDL-C and low LDL-C is not cardioprotective. This may be an example of the second error of causation that I just mentioned – there is a separate factor that is cardioprotective, and that incidentally results in high HDL-C and low LDL-C. Employing drugs to simply redistribute cholesterol between HDL and LDL is therefore not going to be beneficial. It also supports the notion that statins do not exert their effects by lowering LDL-C, but rather through some unrelated mechanism.
In a condition known as familial hypercholesterolemia (FH), there are heritable genetic mutations of LDL receptors (through which cells would normally take in LDL-C from the circulation). This results in a deficiency of receptor numbers, meaning cells cannot take up as much LDL-C and more is left in the circulation. These individuals are of interest because, according to the cholesterol hypothesis, a higher circulating LDL-C should increase their risk of CVD. However, the literature is inconsistent – there are studies showing that individuals with FH have higher mortality, or have the same mortality, or even that they have reduced mortality compared to controls, that mortality is only higher when young and normalises later in life, and that FH individuals on statins have about the same outcome as those not.
Another approach has been to use genetic screening of the wider population to identify people with naturally low or high LDL-C due to a difference in genetic makeup. This is known as Mendelian Randomisation. Again, interpreting the literature in the presence of many confounds is a challenge.
Further, with these types of studies, there are uncertainties around extrapolating from individuals with special genetic-makeups to the general population. As well, the groups can be heterogeneous, with many genes and mutations – for example, at least 4 genes and over a thousand different mutations have been identified in FH and we cannot say what all of these are going to do.
It is a measure of the weakness of the cholesterol hypothesis that advocates need to reach for studies of this kind for support.
Is there an association between blood cholesterol and arterial plaque?
The cholesterol hypothesis predicts that the more LDL-C there is in the blood, the more likely it is to get into arterial walls and cause atherosclerosis. Therefore, we should expect an association between cholesterol in the circulation and cholesterol in the arterial walls. This is a crucial test because, while we cannot conclude causation from an association, we can reject causation if there isn’t an association.
This relationship was studied (for the aorta) as far back as 1936 in 123 cases of sudden death (mostly automobile accidents). The degree of aortic plaque (total fat content) and blood cholesterol (total) were directly measured at autopsy. There was no association between plaque and blood cholesterol, in any age group. The figure shows what they found overall (cholesterol in blood x-axis, aorta fat y-axis).
In 1960, the study was replicated and expanded to include other coronary and cerebral arteries by a group that reported on 200 autopsy cases. Their conclusion: ”No correlation could be observed between the serum cholesterol level and the amount and severity of atherosclerosis in the arteries”.
These studies were criticised because of uncertainty about the accuracy of post-mortem measures of blood cholesterol. This was addressed by measuring cholesterol during life, and atherosclerosis after death, in war veterans who were institutionalised for mental health or domiciliary care. Cholesterol levels were taken regularly in 800 patients monitored for 6 years, over which time 191 died. Cholesterol levels preceding death could be compared to atherosclerosis at autopsy. Again, there was no association – atherosclerosis was just as common in people with high or low cholesterol concentrations. This study, published in 1963, has been cited just twice since then – one example of establishment science marginalising discordant data, by ignoring it.
While it’s a different question, we also know that people hospitalised after a non-fatal heart attack are about equally likely to have high or low cholesterol.
There are techniques to estimate the degree of atherosclerosis in living individuals, although these methods have their limitations, and we can relate these measures to different lipoproteins. The studies variously do, or do not, support an association between cholesterol and atherosclerosis in living people. Again, we do not get clarity, but rather inconsistency.
Cholesterol in the elderly
The cholesterol hypothesis predicts that the degree of atherosclerosis should accumulate over a lifetime, so that elderly people with high LDL-C should experience greater CVD and all-cause mortality. The opposite is the case – most elderly people (> 60 years of age) with high LDL-C have higher survival rates than those with low LDL-C, while in the remainder it is about the same.
This was alluded to in the Framingham heart study that has been following people for decades (and that formed the basis for the first CVD risk calculator). In the original report, they only found a relationship between cholesterol (total) and CVD in men (not women) under 50 years of age. In a 30-year follow-up study, they found that as cholesterol levels fell, so did survival.
Because this is only an association between LDL-C and mortality, we cannot say that higher LDL-C is protective. Although it could be – LDL itself can be an anti-viral and some consider it to be part of our immune system, and cholesterol is anti-oxidant and vital to our biology. However, it could also be that an increasing frailty and susceptibility to disease is associated with a decline in cholesterol levels.
The curious thing is why has this become a problem now – we, as a species, have lived with cholesterol for about 2.5 million years, but atherosclerosis only became a medical concern around the mid-20th century. Animal fats are not to blame, because the hypothesis linking saturated fat to CVD via cholesterol (the diet-heart hypothesis) is not supported by clinical data and, besides, we were eating plenty of animal fats prior to the modern era. We need to look for something attributable to the 20th century and thereafter. It would make more sense to address whatever the triggers have been, or at least to identify them, rather than medically treat blood cholesterol that may only associate with CVD. It’s akin to treating diabetes with glucose-lowering drugs, which gets blood-glucose down but doesn’t improve the underlying cause – insulin resistance.
I’ve endeavoured to survey the evidence for and against the cholesterol hypothesis. There are those who accept the hypothesis on this evidence, and who will make medical and lifestyle decisions on that basis.
However, the marginality of the evidence, especially for an hypothesis as long-standing as this one, is interesting. Were this hypothesis true, we should have seen an accumulation of compelling evidence by now. Instead, the evidence seems just enough to prop up the hypothesis. Further, we have evidence that is not consistent with the hypothesis. Normally in science, just one piece of evidence contrary to an hypothesis would be sufficient to reject that hypothesis.
There is more going on here than science though. We have statin drugs that lower cholesterol, creating the need for a market, and creating an income stream for industry, the research they support and the medical associations they donate to. Reputations have been built around this hypothesis, at both scientific and institutional levels, making it difficult to back down from now. Pragmatically, the cholesterol hypothesis cannot be allowed to fail, and that might explain why this level of evidence is considered sufficient.
To assess this hypothesis with a fresh approach, we may need to wait for a new generation of medical and scientific researchers. To paraphrase the Nobel laureate and physicist Max Planck: Science advances one funeral at a time.
In the meantime, the trouble with accepting the cholesterol hypothesis is that it blinds us to the need for a better hypothesis.
Print This Post
I am not a medical doctor. Nothing herein is, nor should be taken to be, medical advice.
The evolutionary foundation of cholesterol, and how important it is to our biology ⇒
Cholesterol and lipoprotein transport ⇒
The marketing science of Lipitor ⇒
CVD risk calculators, including the Framingham risk calculator ⇒
Cholesterol clinical guidelines ⇒