10th October 2026
ADDENDUM: 16th October 2026. I stated the following. ‘Bradford Hill stated anything less than a doubling, or halving in risk, can be safely ignored in any observational study.’ A reader pointed out that Bradford Hill did not say this. I have checked, and he did not. Or if he did, I cannot find it. I got this figure from another source. Bradford Hill made only general statements about observational studies and their inherent bias(es). It is, however, generally agreed that an Odds Ratio of >3.0 would be considered strong evidence in an observational study. It is not an absolute figures, but can be checked if you wish to do so.
The ethical issues preventing black and white clinical studies
Medical research is very difficult. Always has been, always will. One of the main reasons is that designing a study to prove a hypothesis, beyond any doubt, could well be completely unethical. You would be stopped.
To use a historical example. You believe that smoking causes lung cancer? Well, find a few thousand non-smokers, start half of them smoking, and tell the other half not to smoke. Wait a few years and then see what happens?
Bong!
Clearly you are not allowed to ask people to do something you suspect to be harmful. Because of this ethical dimension, there has never been randomised, controlled study (RCT) done on smoking
Nor did anyone try the reverse. By which I mean, find a group of people who are already smoking. Then convince half of them to stop while the other half carry on. Wait a few years, then see what happens.
Why was this not allowed? Well, apart from the fact that smoking is addictive, and stopping is tough, so you are very unlikely to get half of them to stop, if you believe smoking is damaging, you cannot let people smoke away for the sake of your study. You should be trying to dissuade them, not let them continue.
Because of such ethical considerations, Hill and Doll used an ‘observational’ study design. Which means they didn’t ‘intervene’ in any way. Instead, they watched doctors who smoked vs. doctors who didn’t. Then found out who got lung cancer. As it turned out, the difference was massive. Smokers were twenty-five times more likely to suffer the condition.
Observational studies are often criticised for their potential bias(es), and confounding factors. Sticking to the smoking theme here, smokers may differ from non-smokers in many other ways. They could take less exercise, or drink more, or do other unhealthy things. Therefore, any difference you find between the groups could be due to these other confounding factors. And there may be others, you never even thought of at the time.
This ‘confounding factor ‘problem is always a major hurdle in medical research, especially in observational studies. Because of this, Bradford Hill stated anything less than a doubling, or halving in risk, can be safely ignored in any observational study. Too much noise, not enough signal.
However, once you reach a certain point, you overwhelm all possible confounding factors. With smoking, Hill and Doll found a difference of twenty-five. Which means they were looking at something as close to absolute proof as you are ever going to get. Some epidemiologists set the observational ‘proof limit’ at six. I am not entirely sure why this figure was chosen.
The confounding problem also explains why Hill and Doll studied doctors. They reasoned doctors lived pretty similar lives to each other – other than their smoking habits. They then did a lot of research on the matter, by surveying doctors about their personal habits. As it turned out, they really didn’t need to.
Smoking is not alone – there is also LDL/Cholesterol
Smoking is a relatively extreme example when it comes to the level of harm caused. But there are very similar ethical problems in many other areas. Looking at LDL/cholesterol for instance. The best study you could possibly do, leaving ethics to one side, would be to find a way to increase LDL, and see if this leads to an increase in cardiovascular disease.
Obviously, you would never be allowed to do this. What researchers were allowed to do, in the past, was lower LDL to see if this reduced the risk of cardiovascular disease (using statins mainly).
However, the ethical target does not stand still. For example, it is no longer ethical to do a cholesterol lowering study today without giving statins to everyone, in both groups. Once a drug has been shown to be beneficial, you cannot withhold it.
Ergo, the study cannot be:
New drug vs placebo
Instead, the model has to be
New drug and statin vs. placebo and statin
This can also be referred to as ‘new regime vs. standard regime.’
Confounding still stalks the land of randomised controlled trials (RCTS)
Moving onto the ‘gold standard’ of randomised controlled trails (RCTs). Even here, you can still run straight into confounding problems. For example, does a drug that lowers LDL levels have additional effects that could explain any benefit? Effects that you had no idea about at the time.
This is a long-standing argument I have about statins. They lower LDL … fact. They also reduce the risk of CVD, although I believe this relatively small benefit to have been massively overhyped. But statins do not just lower LDL. They have many other additional ’off target’ actions. In medical jargon these are known as pleiotropic actions.
One of the most important of these pleiotropic actions is that statins help to maintain the health of endothelial cells (the cells lining all blood vessels). This is important because endothelial dysfunction is now recognised to be a key trigger for cardiovascular disease.
Statins help to promote endothelial cell health by activating a process called Nitric Oxide signalling. Without getting hyper-technical, here is the key conclusion from a paper on that issue:
‘Statins are widely used to treat hyperlipidaemia and provide pleiotropic effects on the vascular system by regulating eNOS*-NO signalling. Existing evidence suggests that statins activate eNOS-NO signalling to improve endothelial dysfunction and provide a potential therapeutic target in preventing cardiovascular event.’ 1
*Endothelial nitric oxide synthase (the enzyme that creates Nitric Oxide)
There is a very big confounding factor in all statin trials. One that was not known to exist when the trials were done. Does this mean that the benefits of statins are due to increasing nitric oxide levels, not lowering LDL levels. How can you now know?
Or, to put it another way. What would be the best way to prove, or disprove the LDL hypothesis today? As you cannot do trials on statins anymore. There is a way. It is being done.
The best study ever?
I should probably call this, the best study possible – from an objective scientific perspective.
Returning to my original premise. If you want to find out if factor x increases the risk of disease y. You increase factor x This represents your ‘test to destruction’ experiment.
However, I just said you are not allowed to do this. Or are you? You certainly can’t ask for volunteers to have their LDL raised. Unethical would be the immediate cry.
Sometimes, though, fate intervenes in ways that you could never have predicted. As it turns out there are people whose LDL level shoots up. If, that is, they don’t eat carbohydrates. There are not that many of them, but they do exist. These people have to be thin, or lean – and there must be some genetic component here as well.
They have been labelled as lean mass hyper-responders (LMHR).
They began to turn up in higher numbers when the keto/paleo movement began to take off. This is a group who do not eat vegetables, or bread, or rice, or potatoes … at least only in very small amounts.
Instead, they eat steaks, and eggs, and cheese and bacon, and … animal fat. I have to say that I try to be keto. But it is tough Those croissants are calling, as is the gin and tonic. But vigilant Ketos do live among us. Eschewing carbohydrates in an attempt to achieve … something. Avoid dying, I suppose.
In LMHRs LDL levels shoot up. And I mean shoots up. Doubling, trebling, in some cases quadrupling, even more. Why? Let’s leave that to one side here. It gets complicated. You can read about the possible mechanism in the paper ‘Elevated LDL Cholesterol with a Carbohydrate-Restricted Diet: Evidence for a “Lean Mass Hyper-Responder” Phenotype.’ 2
So here, by chance, we have a group of people who have greatly increased their LDL levels, voluntarily. They don’t want to change their diets, and many of them do not want to take LDL lowering medications either. Does this mean they will be more likely to have heart attacks and strokes. That would be the mainstream view.
But do they? Here we have, potentially, the perfect experiment to prove, or disprove, the LDL hypothesis.
There are a couple of immediate problems that crop up. First, getting money to study this from conventional sources is almost impossible. No pharmaceutical company is going to cough up a bean – for obvious reasons.
As for other potential funders. Universities, Governmental organisations? ‘You’ want to do a study that might disprove the LDL/Cholesterol hypothesis…Next!’ No University would want their reputation tainted by such an association.
Another problem is that many lean mass hyper responders are relatively young, and also generally fit and healthy. They are unlikely to have serious health problems any time soon. Waiting for a heart attack, or stroke, could take many years. Decades even.
With regard to the first issue … money, or the lack of. This means the study had to be relatively small. As it was funded by enthusiasts, out of their own pockets. Maor pharmaceutical trials can cost hundreds of millions of dollars.
With regard to the second issue. Which is that you may have to wait a very long time to see a change in hard clinical endpoints e.g. heart attacks. You can, instead study their coronary arteries – arteries of the heart.
These are usually the worst affected by atherosclerosis (thickenings and narrowings). The most commonly used test to look for atherosclerosis is a coronary artery calcium scan (CAC scan). This is, at present, the most accurate predictor of a future cardiovascular event.
No calcium – good
Lots of calcium – bad
Calcium building up fast – very bad.
Like all such tests it is not perfect, but it is pretty good, and widely accepted by most researchers. It has been validated in many different studies.
Bringing all this together. We have a group of people eating a diet that raises their LDL levels dramatically. They don’t want to change their diet, nor take statins (or other LDL lowering medications).
At this point it becomes ethical to study what has happened to them – what is happening to them. Have they, or have they not, developed calcified more atherosclerotic plaques – compared to a well-matched group who do not have raised LDL levels.
Yes, this study is not randomised – it cannot be. It is not placebo controlled – it cannot be. It is not, directly, interventional. Although you could argue it is. In that the LMHR ‘intervened’ with their own very low carb diet. It is an interventional diet – just not one provided for them as part of a trial. Perhaps.
Given these provisos, it is pretty much as good as you are ever going to get.
Next – what happened?
1: https://www.sciencedirect.com/science/article/pii/S0753332224000738
Thank you for this and the ever increasing light at the end of the tunnel. This is purely anecdotal but I’m 74 and in my mid sixties went on a ketogenic diet because at the time there was some thinking that it helped migraines. I loved that way of eating and was on it for about 3 years. The only reason I stopped was because as I was getting older I was worried about the interaction between being in ketosis and any medications I needed as I got older. I’m pretty sure during that time I reset my metabolic health. Before I would become weak at times if I didn’t eat and very hangry if I didn’t get food on time. That stopped has completely. Since coming off the keto diet I’ve still stuck to about 80 gms carbs per day. My LDL is always high but my TG/HDL ratio is really good.
Thank you for your continued resistance.
Best wishes, Teresa
I’ve been following Dave Feldman via low carb forums and his website for years. He certainly grabbed my attention when he could successfully manipulate and predict his cholesterol levels via his diet.
Didn’t you find it very strange that one of the companies that were paid to analyse the LMHR scans reported calcium progression when other analyses of the same scans showed regression.
How curious, I thought. It got me thinking about who might have some interest and influence in calcium scanning, surely not. 😉
You think the system may be corrupt, or corrupted? You would most likely be absolutely correct. The cholesterol lowering market is worth trillions. The best way to dismiss findings you don’t like it to spread doubt.
Corruption you say.
When some of the LMHRs had their (trial blinded) scans analysed privately by the same company and got opposing results to those reported in the trial – what else can you call it? Rounding error?
Hope I’m not treading on the toes of your next article
Well my first comment has gone into moderation – touchy subject obviously. 😉
It is me that moderates, and I do a few other things in my life. Off to play golf at royal Marrakesh this afternoon
Well that sounds pretty amazing and no doubt a smidge warmer than the UK at the moment – enjoy.
Good to have you back, Dr K. I am one of those who have followed a very low carb diet for 15 years now, rather skinny by today’s standards but normal by those of 40 or 50 years ago. My cholesterol levels shot up from around 5 to 7 but with an excellent HDL/trig ration of around .6. All was fine until I began an extremely stressful period of my life which has lasted for some years now, including a severe heart failure diagnosis (successfully treated) and a husband of 63 years with mixed dementia. It all sent my blood glucose levels shooting up, (diabetic for 58 years) although in January the cardiac nurse congratulated me on my HDL/trig ratio. “Excellent.” The lovely, wise cardiologist told me my blocked arteries were caused by a combination of genes, age (I’m 84) and bad luck and to carry on what I’m doing, lifestyle-wise.) I thought of your wise books about the part stress plays in coronary problems and took great comfort from that. As part of a welter of medications I have to take every day, I am on a low dose of Rosuvastatin which I can just about tolerate.
Why am I wittering on like this? Because I recognised myself in your article and it’s nice to get a bit of recognition. Thank you so much
An example of unethical research back in the days when rules where different. See Wikipedia for reference.
Vipeholm Study (Swedish: Vipeholmsexperimenten) were a series of human experiments where patients of Vipeholm Hospitalfor the intellectually disabled in Lund, Sweden, were given various amounts of sweets to eat, including “extra sticky toffee” (that was specifically manufactured for the study without certain ingredients so that the toffee would stick longer to tooth surfaces) to provoke dental caries.
I just love the way you always manage to think around a problem and then come up with a solution – pure magic, tadaaaaaaa!
That was a great discussion and explanation about the LMHRs and their LDL. I hope you can do a blog post about the failed HORIZON trial (Novartis) and why Lp(a) particles have nothing to do with “CAUSING” atherosclerosis.
I also would love to hear some follow up about Lp(a).
I’m looking forward to the denouement! 😊
I was lying in my tiny tent in the Lake District with no internet thinking deeply about 4S and pleiotropic effects and I had the eureka moment about LDL and statins and what you write about. I’m a minor keto guy, I’m not convinced I’m lean but I do have strong arms and legs and my LDL is high. I’ve had MRI and carotid/aortic ultrasound and nothing remarkable was found.
So I’m claiming LMHR??
I love what you do and try and pass it on to anyone who will listen. One thing I have noticed is that GPs were LDL sceptical 15 years ago (I used to talk to them about Ezetimibe). But I think it’s been beaten out of them.
Best regards
Andy
Sent from Outlook for iOShttps://aka.ms/o0ukef ________________________________
Confused of Calstock, does this mean we should be taking statins even though cholesterol isn’t the problem? Speaking as someone with cholesterol of around 10 and PAD but refusing statins.
Beetroot juice – organic of course – to boost nitric oxide.
Same, can I ask Doctor Kendrick do you take a statin ?
No
No, He can’t ask, or no you don’t take one? sorry, i am not sure 😐
I dont take one.Nor do I know what my LDL level is.
Great Question!
So we should take the statin for the endothelium benefit?
You could, I don’t. The benefit is too small to outweigh the downsides, in my opinion
The way I see it, the body was designed to run on water and food and not drugs. Theoretically, a person can live into their 90’s and never need drugs. When I see (in the US) the expected life span slipping even after 40-50 years of endless drugs, something has gone bonkers. Drugs are not helping us live longer. And the US is one of the most drugged up countries in the world. Something else is going on and that is what needs to be addressed. Not whether we need more silly statins.
As always, an interesting read. As CAC scans have been mentioned here before I thought I’d mention that I had a private CAC scan a couple of weeks ago more out of interest than anything else, because I’m an ex smoker, (gave up in 09), my cholesterol especially TriGlycerides has always been pushing the recommended limits (now reduced through a few dietary tweaks), and I wasn’t greatly active until I retired. My Doctor is always trying to push statins on me & at 69 I wanted to see what the damage was. Pleased to say, a score of 6 on the Agatston scale (one minor fleck of calcium in the LAD). I won’t be complacent but I was pleasantly surprised.
Your doctor may not be pushing statins on you. Should you have a heart attack he needs to be able to say – I advised a statin but the patient refused. My GP always suggests a statin. Rather nicely when I last saw him about 3 years ago (already well past 80) as I was leaving he said – well I guess at your age it wouldn’t make much difference anyway. Spot on !
I’m also confused. My last blood test showed an LDL level of 8.5 but I paid to get a cardiac calcium test 4 years ago because of pressure from my GP to take statins. My score was 0% which was pretty amazing considering I was 61 at the time and have had raised LDL for many years. Now I’m wondering if I should consider statins particularly in light of the fact that I take a PPI daily.
Interesting to see this over in The Guardian today: “A controversial research trial that will withhold hepatitis B vaccines from some African newborns was given approval to move forward this week.”
I can’t find where Bradford Hill stated that “anything less than a doubling, or halving in risk, can be safely ignored in any observational study.” Do you have a reference for this? Thanks.
You are right. I got this figure from Zoe Harcombe, and her blog – and she normally checks her information very thoroughly. I have checked, in various places, and find that I am wrong.
Bradford Hill did not provide an absolute figure. At least I have been unable to find one. I shall pop up an addendum shortly. Other epidemiologists did contact me about my book Doctoring Data, where I stated that observational studies could not prove causality. I was told that an Odds Ratio of six(ish) would normally be considered proof of casualty. Although like most things in this area, this figure is not an absolute.
I have seen articles suggesting we need to go for 5 Sigma, which, I think is a little tough. A 5-sigma result corresponds to a p-value of about 3 × 10⁻⁷ (roughly 1 in 3.5 million).
Thank you for commenting on my blog, and correcting me.
Hi Malcolm
Surely there are other ways to improve NO signalling than taking statins or eating beetroot every five minutes?
Let’s have them
I second that request !
To improve NO levels, surely it’s best to live in a climate where the sun shines for 2,500 to 3,000 hours a year? (Many people in the USA do.)
I’ve just bought a sun lamp. I don’t think vit.D pills are really enough at latitude 52 degrees north.
Looking forward to the next part!
Hello,
Great! Can’t wait to read what happens next!
Always had low LDL, don’t know why, even when low carb as a body builder – and I don’t eat much beetroot! Waiting for part 2
So these ethical limitations are why new vaccines are never tested against a placebo, but rather the previous version of the vaccine (flu, for example)?
Yes
I am keen to learn to your views on statements made about the apparent benefit of the Shingles vaccine on the future risk of developing dementia, stroke and heart attack. Is this benefit simply the result of people seeking the Shingles vaccine simply taking more care of their health in general? Have you interrogated the available data?
Ah, and there is a difference between smoking commercial products (with over 100 additives, including pyrazines, flavor enhancers) and nicotine itself (which we actually find in some foods, like eggplants, tomatoes, and white potatoes).
Dear Dr Kendrick, last month I foolishly posted a comment on your column of 10 December 2023 titled “Very high LDL no impact on plaque progression” about lean-mass hyper-responders. This was three years after your column, which was clearly a daft thing to do and for which I apologise. You have now returned to the subject, so I wonder if I could submit my question about your response to the actual science in two columns on the same subject by Dr Nicola Guess, both of which appeared on Substack: the first titled “‘Lean Mass Hyper-Responders’ and Atherosclerosis” (posted on 11 April 2025) and the second “Nick Norwitz could have been great. He isn’t” (posted on 16 May 2026).
https://drguess.substack.com/p/lean-mass-hyper-responders-and-atherosclerosis
https://drguess.substack.com/p/nick-norwitz-could-have-been-great
Dr Guess, according to Google AI, believes that “elevated LDL cholesterol is a direct causal factor in atherosclerotic cardiovascular disease (ASCVD)”.
Thank you.
Vicky
The timing of this article is interesting! My husband who is 48 has recently been on a big health kick. He started fasting about a year ago, then a few months ago he went strictly carnivore for about a month. He’s also been going to the gym and has lost a lot of weight. He had a flu about a month ago and went to the GP as he was quite sick They did the usual tests and blood tests and then called him back in. Apparently his cholesterol was very high (I can’t remember the numbers) and I wondered how with the carnivore diet it had shot up so much, so this sounds like an explanation! His GP then recommended he go on a vegan diet haha, but luckily he wasn’t interested in that advice. He is supposed to get a CT scan I think which we haven’t got around to yet due to costs. But I’m thinking after reading this that perhaps his increased levels are not a huge concern?
They are not
Cliff hanger
“Existing evidence suggests that statins activate eNOS-NO signalling to improve endothelial dysfunction and provide a potential therapeutic target in preventing cardiovascular event.’
Were there (are there?) drugs that lower cholesterol without having this effect on the endothelium? If so that could shed light on whether the benefit of statins is from this effect rather than from lowering cholesterol.
That is a very good question. IN fact, i think I will address this next, as it is highly relevant to the entire disucssion.
Dear Dr Kendrick
I am wondering if you could link me up to a UK based longitudinal study of medical benefits of a low carb high fat ketogenic diet
My email is
Kimnlockyer@icloud.com
Thank you for your insight and relentless energy in the field!
Best wishes
Kim
“it is no longer ethical to do a cholesterol lowering study today without giving statins to everyone, in both groups. Once a drug has been shown to be beneficial, you cannot withhold it.”
How can a drug ethically be considered to be beneficial when the risks of its harms have been insufficiently studied?
Dr Kendrick sets out the problem himself: “There is a very big confounding factor in all statin trials. One that was not known to exist when the trials were done. Does this mean that the benefits of statins are due to increasing nitric oxide levels, not lowering LDL levels. How can you now know?”
“Or, to put it another way. What would be the best way to prove, or disprove the LDL hypothesis today? As you cannot do trials on statins anymore. There is a way. It is being done.”
IMPROVE-IT added ezetimibe on top of simvastatin, so the pleiotropic effect sat in both arms: “The Kaplan-Meier event rate for the primary end point at 7 years was 32.7% in the simvastatin-ezetimibe group, as compared with 34.7% in the simvastatin-monotherapy group (absolute risk difference, 2.0 percentage points; hazard ratio, 0.936; 95% confidence interval, 0.89 to 0.99; P=0.016).”
“When added to statin therapy, ezetimibe resulted in incremental lowering of LDL cholesterol levels and improved cardiovascular outcomes. Moreover, lowering LDL cholesterol to levels below previous targets provided additional benefit. (Funded by Merck; IMPROVE-IT ClinicalTrials.gov number, NCT00202878.)”
FOURIER, evolocumab against placebo on a statin background: “Relative to placebo, evolocumab treatment significantly reduced the risk of the primary end point (1344 patients [9.8%] vs. 1563 patients [11.3%]; hazard ratio, 0.85; 95% confidence interval [CI], 0.79 to 0.92; P<0.001)"
"In our trial, inhibition of PCSK9 with evolocumab on a background of statin therapy lowered LDL cholesterol levels to a median of 30 mg per deciliter (0.78 mmol per liter) and reduced the risk of cardiovascular events. These findings show that patients with atherosclerotic cardiovascular disease benefit from lowering of LDL cholesterol levels below current targets. (Funded by Amgen; FOURIER ClinicalTrials.gov number, NCT01764633 .)"
For the sourced version of this question: https://openneedle.ai/q/non-statin-ldl-lowering-drugs-reduce/
It might help to put the risk numbers in context:
in the Women’s Health Study (over 28,000 women followed for 21
years), the hazard ratios for future coronary heart disease
were striking:
Risk Factor
Hazard Ratio for CHD
Type 2
Diabetes
10.7
Metabolic
Syndrome
6.0
Insulin
Resistance (LPIR test)
6.4
LDL
Cholesterol
1.6
All of the larger hazard ratios involve elevated insulin – which is not normally measured.
A1C can be misleading,
delaying the detection of the underlying hyperinsulinemia that
drives vascular disease. This hyperinsulinemia, through complex
and tissue-specific effects on cell survival, contributes to an
environment where clots form and are not cleanly resolved. The
result is the layered, “paved-over” plaque described by the
thrombogenic hypothesis, a model that has more explanatory power than the traditional LDL narrative.
So why the focus on LDL?
Lp(a) seems to rely on the type – the kringles vary with genetics. But once again – it is a clotting factor:
“The LPA gene that codes for apo(a) evolved from plasminogen in Old World monkeys: high homology between apo(a) and plasminogen”
I also ask why they don’t measure Lp(a) in nattokinase studies – seems an obvious omission