Estimated read time: ~12 minutes
Introduction: the "Cholesterol Paradox"
One of the most common angles of argument from Cholesterol Denialism is that low levels of low-density lipoprotein cholesterol (LDL-C) increases risk for mortality from other, non-cardiovascular diseases.
This is commonly heard in relation to all-cause mortality and cancer, while also suggesting that there is little benefit to targeting lower LDL-C in the elderly in order to prevent cardiovascular disease (CVD).
Video above: Aseem Malhotra claims about LDL-C & Mortality Risk
Originally from this interview on May 26 2021
Especially interesting to look at age 60-74 with high cholesterol and low BP/non-smoking vs those with low cholesterol. pic.twitter.com/Z9m60EUXow
— Dave Feldman (@DaveKeto) November 16, 2018
From: Malhotra, BMJ 2013;347:f6340
Like many Quackery arguments, there is often a sliver of evidence on which to serve a feast of fantasy. And like such arguments, reconciling seemingly discrepant data against the totality of evidence can bring us closer to an approximate truth of that issue. In this Sigma Statement, we examine the supposed "Cholesterol Paradox."
The theory of a "cholesterol paradox" posits that low levels of LDL-C is associated with worse non-CVD health outcomes in the elderly, specifically in populations aged over 60 years of age.
The apparent "paradox" stretches back to observational research in the 1980's which suggested a link with low LDL-C levels and subsequent cancer risk.
A 2016 systematic review by prominent Cholesterol Denialists purported to show an inverse relationship between higher LDL-C and all-cause mortality, i.e., higher LDL-C lowered risk of death from any cause.
To reconcile this apparent "paradox", there are several lines of thinking we can consider:
- Is this supposed risk evident with deliberately (exogenously) lowering LDL-C, e.g., through drugs?
- Is this supposed risk evident with naturally (endogenously) lower LDL-C levels, compared to higher?
- What could contribute to low LDL-C in the elderly and/or increased mortality risk?
- Does this observation hold true across the lifespan?
Exogenously Lower LDL-C
One aspect of the argument in relation to low LDL-C and increased non-CVD mortality risk is that deliberately lowering LDL-C poses a risk for developing cancer or dying from any other cause.
This is then used to argue against deliberately intervening to lower LDL-C in order to reduce CVD risk. Essentially, this is the basis for the "avoid statins because low LDL is bad" advice.
In this regard, it is important to distinguish between:
- Exogenously lowering of LDL-C: Deliberately lowering LDL-C through pharmacological intervention.
- Endogenously lower LDL-C: Internal mechanisms which may result in naturally lower cholesterol levels.
Let's first deal with the evidence for deliberate pharmacological lowering of LDL-C.
Armitage et al., for the Cholesterol Treatment Trialists' Collaboration (CTTC), conducted a meta-analysis of 28 statin randomised controlled trials (RCTs) to examine safety and efficacy in relation to both CVD and non-CVD mortality outcomes in the elderly. The established benefit to lowering CVD risk was clearly evident. However, in addition there was no increased risk of any non-CVD cause of death, irrespective of age group, based on different LDL-C lowering. Meaning that for every 1mmol/L (38.6mg/dL) that LDL-C was lowered, there no difference in non-CVD mortality risk.
If we consider cancer specifically, as this is commonly a point raised by those discussing the "paradox", we see a similar story; i.e. there was no risk observed per 1mmol/L reduction in LDL-C for either cancer incidence or death from cancer. Then if we factor in both CVD and non-CVD mortality together, the risk of all-cause mortality was 9% lower (RR 0.91, 95% CI 0.88 - 0.93) in patients treated with statins.
Thus, the evidence does not support that deliberate lowering of LDL-C using statins increases risk of death from any cause or cancer.
Effects on cancer incidence and cancer death per mmol/L reduction in LDL-cholesterol.
Taken from: Armitage et al., Lancet. 2019 Feb 2;393(10170):407-415
Crown Copyright © 2018 Published by Elsevier Ltd.
There is also evidence from non-statin pharmacological interventions, including:
- Ezetimibe: an NPC1L1-inhibitor, which inhibits intestinal cholesterol uptake
- PCSK9-inhibitors: inhibit a pathway resulting in upregulated cholesterol clearance from the circulation
- CETP-inhibitors: inhibit cholesterol transfer between HDL and atherogenic lipoproteins, like LDL
Sabatine et al. conducted a meta-analysis of non-statin randomized-controlled trials (RCTs), and compared these data to the evidence from the CTTC meta-analysis discussed above. The most important characteristic of this analysis was that it focused on patients with already low LDL-C levels from statin therapy, and specifically investigated whether further lowering to very low LDL-C levels was associated with any adverse effect on death from other causes.
In these trials, the control groups were also treated with statins and all control groups had baseline LDL-C levels of less than 1.8 mmol/L (70 mg/dL). Further lowering of LDL-C (which went as low as 0.5 mmol/L, or 19 mg/dL, with PCSK9-inhibitors in the FOURIER trial) was not associated with risk of cancer in participants treated with other non-statin pharmacological interventions. Nor was there any association with serious adverse events, myalgia and/or myositis, new-onset diabetes, or hemorrhagic stroke.
From: Sabatine et al., JAMA Cardiol. 2018;3(9):823-828
Copyright © 2018, American Medical Association
The important point in relation to these different pharmacological interventions for lowering LDL-C is that they all lower LDL-C through different mechanisms, namely:
- HMGCR (statins)
- NPC1L1 (ezetimibe)
- PCSK9 (PCSK9i)
Thus, independent of the mechanism of lowering LDL-C, there is no observable safety concerns in relation to risk of non-CVD incidence or mortality.
However, it should be noted that there may still be some risk evident in the research. A meta-analysis of studies which compared levels of LDL-C in patients treated with statins to levels in patients in the control group (i.e., not treated with statins). This meta-analyis found similar risk for cancer, whether treated with statins or not, at any level of LDL-C. However, this analysis also indicated that both groups had a higher risk for cancer with lower LDL-C levels. Nevertheless, the similar rates in the intervention and control group are critical, as it corroborates the above research demonstrating that lowering LDL-C through statin intervention is not the cause of increased cancer risk. For example, in the CTTC meta-analysis the rate of cancer diagnosis was identical (6.4%) in both intervention and control groups.
The overwhelming evidence demonstrates that deliberate pharmacological lowering of LDL-C does not increase cancer risk or all-cause mortality risk.
However, despite no difference between pharmacological treatment and control group risk of cancer, there remains in the data some evidence of a putative association between low measured LDL-C and cancer risk. What could explain this? This is where the concept of endogenously lower LDL-C requires examination.
Endogenously Lower LDL-C
To consider endogenously lower LDL-C levels, we need to come back to a couple of the questions posited in the introduction, specifically:
- Is there a risk of mortality with naturally lower LDL-C levels?
- Is there a risk across the lifespan?
Let's consider these questions.
Bear in mind that the associations between low LDL-C and mortality risk are primarily observed in the over 60's age group. However, we know that exposure to LDL-C influences health and disease risk from the second decade of life. This is important because claims from studies like the 2016 BMJ Open paper, which suggested high LDL-C lowered mortality risk in the over 60's age group, do not account for LDL-C exposure over the course of the lifespan. Because many people will have died already, this type of analysis excludes the effects of cumulative exposure; i.e. this is an error of survivorship bias.
With Mendelian Randomisation (MR) studies, it is possible to account for this by examining lifelong exposure to different levels of LDL-C. Meaning, we can look specifcally at people who have different lifelong levels of LDL-C, based on genetics. Postmus et al. conducted an MR study using data from three cohorts, by calculating genetic scores for genes related to LDL-C levels, across age groups (see figure below). The combined analysis of the three studies showed:
- A significant association between a higher LDL genetic score and increased all-cause mortality.
- Individuals in the over 90-years age group with the highest LDL-C genetic score had a 13% increased risk of all-cause mortality (HR 1.13, 95% CI 1.00-1.26), compared with individuals in the lowest LDL-C genetic score.
- This relationship was evident across the lifespan. Meaning, at every age group examined, genetically higher LDL-C levels was associated with higher all-cause mortality risk.
- Importantly, the participants with genetic predispositions for longevity were those with significantly lower LDL-C levels compared to the general population.
From: Postmus et al., Int J Epidemiol. 2015 Apr;44(2):604-12
Copyright © 2015, Oxford University Press
The above figure from Postmus et al. illustrates the associations between LDL-C and all-cause mortality across six different age brackets according to genetic predisposition to:
- Low LDL-C (light grey bars)
- Moderate LDL-C levels (grey bars)
- High LDL-C levels (dark grey bars)
Consistent with the totality of evidence in support of risk related to LDL-C as a cumulative exposure, higher compared to lower LDL-C was associated with increased risk.
Thus, if lifelong exposure to naturally (i.e., genetically) lower LDL-C is associated with lower all-cause mortality risk at every age group, could there be other factors which result in blood cholesterol levels decreasing as a natural response to other physiological processes in the body?
One explanation for the relationship between low LDL-C and cancer is what Rose and Shipley termed the “unsuspected sickness phenomenon”. This is where the metabolic consequences of an underlying, undiagnosed disease (e.g. cancer) causes a drop in LDL-C levels.
In a case-control study of the Framingham Heart Study Offspring Cohort, LDL-C values were lower in participants with cancer compared to matched controls at each point of assessment over an average of 18.7 years prior to diagnosis, indicating that low LDL-C levels predate diagnosis over periods of years.
One way to tease this out is to exclude the first few years of follow-up from the study (i.e. conduct a sensitivity analysis). When this approach is used, if the association between an exposure and outcome is weakened, it indicates that potential confounders may have already been present at the start of the study, but were latent or undetected. Such confounders may have influenced the association. So by eliminating the early follow-up period data, one can prevent such confounders (like undiagnosed/early-stage cancer) from erroneously influecing the association. When one takes studies that show a relationship between low LDL-C levels and cancer risk, and eliminate the early follow-up period, the majority of the association is abolished.
In the UK Whitehall Study, low cholesterol levels were associated with increased risk in the initial analysis. However, by excluding the first two years of follow-up from the analysis, there was no longer any association between low cholesterol and cancer risk. Tornberg et al. examined both cancer incidence and mortality relative to blood cholesterol levels over 18-20yrs, and found that the relationship between low cholesterol levels and cancer incidence and mortality was strongest with the first two years of follow-up. This data corroborates a pre-clinical, cholesterol-lowering effect of undiagnosed cancer. The study found that cholesterol levels measured closer to the time of death from cancer (i.e., mortality as an outcome) were lower again than cholesterol levels measured before the time of a diagnosis (i.e., incidence as an outcome). This finding is important, as the stronger association of lower LDL-C with cancer mortality rather than cancer incidence confirms a pre-clinical metabolic effect of underlying cancer; i.e. LDL-C begins to lower prior to diagnosis and continues to be suppressed with active disease. This time-course relationship eradicates the potential for a causal relationship between lower cholesterol and cancer incidence. Meaning, low LDL-C does not cause cancer.
Thus, the apparent relationship between low cholesterol levels and increased mortality risk is primarily one of reverse causality. Therefore it is underlying metabolic dysfunction related to latent disease which explains the risk of death, not the low blood cholesterol levels. This is confirmed by the evidence that naturally lower LDL-C levels are associated with lower all-cause mortality risk across the lifespan in genetic studies. And it is corroborated by the lack of evidence from deliberate, pharmacological lowering of LDL-C of any increased cancer or all-cause mortality risk compared to untreated patients. Which brings us to a final, but important, consideration.
LDL-C as a Cumulative Exposure: Relevance for Risk Reduction in the Elderly
It is important to consider this concept of a cumulative exposure to LDL-C over time as it relates to deliberate intervention to lower LDL-C in the elderly. By the 6th and 7th decades of life, atherosclerosis may already be advanced. Thus, the benefit to lower LDL-C relates to the absolute magnitude of exposure to lower LDL-C , i.e., both the duration of lower exposure and the actual levels of LDL-C achieved.
These data are observed prospectively in cohort studies over the long-term. When scaling out follow-up periods of up to 40 years, low baseline cholesterol levels were associated with 25% lower all-cause mortality risk, in men aged 30-45 years old at baseline. When follow-up was extended to 46 years, these associations persisted: men with the lowest baseline cholesterol levels had the longest survival time, i.e., longevity. However, given that the same voices who argue that 'lower cholesterol increases all-cause mortality' will cite observational studies which support their claim, but then dismiss all epidemiology when it does not, let us thus turn our focus to genetic studies and randomised controlled trials to illustrate this point.
It is possible to compare the risk reduction from genetically lower LDL-C to the risk reduction from pharmacologically lower LDL-C, as evident in the graph below from Brian Ference's work. You can see there is a stark difference in the slopes of the two lines:
- The top line representing genetically lower LDL-C (endogenously lower)
- The bottom line representing trials in which LDL-C was lowered pharmacologically (exogenously lower).
From: Ference et al., J Am Coll Cardiol. 2015 Apr 21;65(15):1552-61
The difference in the angles of the lines shows that the relative risk reduction at a given magnitude of lower LDL-C is not as pronounced for drug interventions as it is for the genetic variants causing the lower LDL-C. This is explained by considering the duration and magnitude of lower LDL-C. Genetic predisposition means lower LDL-C from birth, with the greatest effect seen for a PCSK9 variant which results in the greatest magnitude of lower LDL-C. However, the average age of participants in drug intervention trials is 63 years old, and at this point in life there has been a greater (lifelong) duration of exposure to higher LDL-C. And so the lower relative risk reduction of pharmaceutical intervention in the elderly reflects their higher baseline absolute risk due to lifelong cumulative exposure.
Elderly patients are complex due to multiple potential factors:
- competing mortality risk from other conditions
- age itself
These factors together add up to a population which is high-risk prior to any intervention to lower LDL-C. Proportional reductions in risk with statin intervention in the elderly do decrease with age, however, this trend is not significant.
An important point to grasp here is that relative risk reduction to lowering LDL-C is independent of baseline LDL-C levels, and independent of other risk factors like hypertension treatment, and smoking. The relative risk of CVD is generally constant over a lifespan when other risk factors are held constant, except age. Therefore (reiterating the point above) the absolute risk is higher in patients over 75 years old.
If relative risk is constant, and absolute risk is higher in the over 75 years age group, there should still be a net benefit to intervening and lowering LDL-C with statin treatment in this age bracket, even though the magnitude of risk reduction may be smaller. This is precisely what the data shows.
In the figure below from Collins et al., you can see that:
- Per 1 mmol/L (38 mg/dL) reduction in LDL-C, the relative risk reduction is similar irrespective of the pre-treatment (i.e., baseline) LDL-C levels.
- Below that, you can see this relative risk reduction per 1 mmol/L (38 mg/dL) lower LDL-C stratified by age bracket:
- You can see that up to 75 years, the magnitude of risk reduction is practically identical at ~21-22% lower risk (RR of 0.78 and 0.79 respectively).
- You can also see that the risk reduction is 13% in the over 75's age group (RR of 0.87). This shows that the proportional risk reduction is smaller, but nevertheless there is a statistically significant and clinically meaningful reduction in risk for CVD events in this age group.
- However, it should be noted that this benefit is primarily evident in secondary prevention, in participants with existing heart failure and renal disease.
From: Collins et al., Lancet. 2016 Nov 19;388(10059):2532-2561
© 2016 Elsevier Ltd. All rights reserved.
Overall, these data are consistent with the competing mortality risk which makes assessments of risk:benefit more challenging in the elderly.
Nonetheless, the proportional benefit does appear to persist once excluding trials of heart failure or renal disease: in a meta-analysis of 14,483 participants older than 75 years and with existing CVD, statin treatment was associated with a 15% (HR 0.85, 95% CI 0.73–0.98) reduced rate of major vascular events. Although the benefit in primary prevention in patients over 75 years is less clear, the data supports a proportional risk lowering by intervening to lower LDL-C in secondary prevention and in patients with existing CVD.
- There is no evidence that deliberate pharmacological lowering of LDL-C is associated with increased all-cause mortality risk.
- There is clear and unequivocal evidence of lower cardiovascular disease event and mortality risk with deliberate pharmacological lowering of LDL-C.
- Thus, there is no evidence of a competing risk or "trade-off" in mortality risk from intervening to lower LDL-C resulting in higher mortality risk from other disease.
- Lifelong exposure to naturally lower LDL-C levels show only one direction of effect, which is to lower CVD and all-cause mortality risk compared to higher lifelong LDL-C levels.
- Thus, natural endogenously lower LDL-C levels from non-pathological processes are not associated with mortality risk of any kind.
- However, endogenously lower LDL-C from underlying latent disease pathology may result in a natural lowering of LDL-C as a metabolic consequence of the disease.
- In relation specifically to cancer, LDL-C appears to lower pre-clinically before a diagnosis, and may be suppressed even further if measured in closer proximity to death. The pre-clinical effect of cancer on lowering LDL-C levels may predate a diagnosis by years.
- Overall, this time-course relationship and reverse causality largely explain the supposed relationship between low LDL-C levels and cancer risk.
- Thus, there is little scope for arguing that interventions to lower LDL-C in the elderly will result in higher competing mortality risk from other disease, including cancer.
- It is important to note that elderly patients, particularly in the over 75 years age group, already have a higher competing mortality risk. The absolute risk of mortality, and of CVD mortality, is higher in this population group.
- This often translates into a lower proportional risk reduction in this age group when treated with statins. However, the net effect is still a benefit, i.e., a lower risk of CVD mortality.
- Ultimately, these data support that interventions with statins and/or other pharmacological lowering of LDL-C is more effective when started earlier in life.
What stops us from recommending statins to all adults (barring those with contraindications) regardless of lipid levels?
I think this is actually something some have debated. And there is certainly a good argument to be made that statins could be prescribed to more people that are at lower-risk and/or are younger. Some groups have interpreted the data that way: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(12)60367-5/fulltext
I think the counter arguments would probably centre around the fact that taking any drug is not risk-free, and therefore whether prophalyactic prescribing to someone that is (and may remain at) low-risk is a route we should go down. This aligns with a view that in an “ideal” situation, we want to have someone on the minimum amount of medications.
But there are certainly others who can talk more authoritatively about this issue, and who are more closely connected to first-line patient care. I’ll perhaps ask some of them to chime in here with some thoughts.
See comment from Dr. Austin Baraki below.
Here is a link to the 2019 ACC/AHA clinical guideline summary that is used to guide decision making at the moment, specifically the section describing risk assessment (e569) and primary prevention (e572). https://www.ahajournals.org/doi/pdf/10.1161/CIR.0000000000000677
All medical interventions come with potential benefits that must be weighed against potential risks, costs, and pill burden. In my opinion, current guidelines that focus on making decisions based on 10-year risk are a bit shortsighted, particularly for younger patients who have much longer left to live. This has prompted calls to use longer-term (e.g., 30+ years) risk assessments to guide prescribing decisions, as described here:
Quoting from this second paper:
“… pharmacological therapy will produce clinical benefit only in those in whom that specific cause is present. Moreover, the degree of benefit will relate to the degree of abnormality, that is the absolute risk and the absolute benefit from pharmacological therapy depend on the absolute levels of the cause(s) of vascular disease. Thus patients with normal LDL levels but high BP do not require LDL-lowering therapy, whereas those with very high LDL levels require more intensive therapy than those with only moderately elevated LDL levels. Targeted therapy would be more cost-effective than population- based single dose therapy precisely because it is limited to those who need it most.”
There are certainly valid arguments from a purely biological standpoint that more aggressive population lipid lowering starting early in life would dramatically lower the incidence of CV events, but this comes up against many more practical issues in the real world. With that said, the field of CV prevention is very hot right now and it will be interesting to see how guidelines and practice evolve in the coming years.
I would be interested more discussion on evidence based optimal ranges for dietary and blood cholesterol values based on an individual’s context/goals.
A blanket avoidance of all dietary cholesterol as the primary dietary goal can lead to the elimination of many nutritious high satiety foods. But at the other extreme glorifying absurdly high cholesterol values is pretty dumb too.
What if we just targeted nutritious high satiety foods to achieve a healthy body composition…. could your body work the rest out for itself?
I’m not sure I entirely understand. First, I think targeting nutritious, high-satiety foods and aiming for a healthy body composition are both already consistently put forward as standard advice. So I think anyone aiming to eat for health is likely orientating towards these already.
As to whether the body can work out the rest for itself… for some people, yes, but for others the answer is clearly no. Some people eat nutrient-dense foods and have a healthy body composition yet have elevated LDL-C and/or ApoB. This may be driven by some aspects of diet (most notably saturated fat intake) or by genetics. In such cases, these people are still at elevated risk, relative to their LDL/ApoB being lower.
As for optimal ranges for blood cholesterol based on individual’s context/goals, in what possible context could elevated LDL-C/ApoB be optimal? While some people claim there is a benefit to this, it seems like nonsense to me. In terms of disease risk reduction, lower LDL is better, and there are already evidence-based cut-off points given at a population level. If anything, an individual’s “optimal” LDL would likely be lower than the limits already suggested at a population level.
I agree that dietary cholesterol is relatively irrelevant for the most part, and it only seems to be detrimental when combined with very high saturated fat intake as it has an additive effect on elevating serum LDL-C. But overall, it seems unlikely dietary cholesterol needs to be avoided. But given that it tends to come along with saturated fat, limiting high SFA foods can be beneficial.
So overall, I think a measure of LDL-C or ApoB over time is crucial in determining one’s risk, and I see no context or goal in which an elevation of either marker would be beneficial.
Great article! Reading this, a question came to my mind, for which I couldn`t find an answer to. When every cell of the body can produce its own cholesterol and is not dependant on absorbing cholesterol from the blood, why do we have this whole lipid transport system? Why does the liver produce VLDL and subsequently LDL to ship cholesterol to peripheral tissues, when the tissues could produce it on their own? Wouldn`t it be more efficient from an evolutionary perspective to get rid of all the cholesterol in the blood to reduce the risk of atherosclerosis, if the body is not dependent on cholesterol in circulation?
Thanks in advance 🙂
Thanks for the question. While yes, cells can synthesise cholesterol, the primary site of cholesterol synthesis remains the liver. Hence, the “forward cholesterol transport” pathway that you described remains important to ship cholesterol out from the liver to target tissues, and particularly because cholesterol will be recycled, cells will need a constant supply to maintain sufficient membrance levels. But, as we noted in the podcast accompanying this Statement, those actual membrane levels are relatively modest.