Where I Stand on Cholesterol

I believe the data

That's the position. Not a position about cholesterol, a position about evidence, and everything else in this piece follows from it rather than sitting beside it.

The data says apoB-carrying lipoproteins are causal. It says the risk tracks the level. It says other factors play a role and that we're still working out how much and how badly. And it says outliers are real, which is a different statement from saying they overturn anything.

None of those are my opinions and I'd rather they weren't. If the data moved, I'd move. I'd sooner tell you that in advance than defend a position I had already quietly stopped holding.

That distinction matters more than it sounds. Most of this argument online is people trading beliefs, which is unwinnable, because a belief can absorb any amount of evidence and carry on. I'm not offering you my belief about cholesterol. I'm showing you what the evidence says and telling you I'll follow it.

I want to be careful about one thing before I start. I'm not interested in the war. There are people I respect on the other side of this, some of them doing careful work, and the tone of this argument online has done more to confuse ordinary readers than either camp's evidence has done to inform them. I'd rather show you what the evidence is and let you decide than tell you who to be angry at.

What the case rests on, and it is not one study

In 2017 the European Atherosclerosis Society put a consensus panel on this question and did something unusual. Rather than run another trial, they asked whether the existing evidence met the criteria for causation.

What they assembled was meta-analyses of over 200 prospective cohort studies, plus Mendelian randomization studies, plus randomized trials. More than two million participants. Over twenty million person-years of follow-up. More than 150,000 cardiovascular events. Across all of it, the same dose-dependent relationship kept appearing between how much LDL a person's arteries had been exposed to and their risk of disease, and the effect grew with the duration of exposure.

Their conclusion was that the evidence "unequivocally establishes that LDL causes ASCVD."

That is a strong word and consensus panels don't use it casually.

Three kinds of evidence, and only two of them answer the question

It's worth separating what those three bodies of evidence actually do, because they are not interchangeable and the difference is the whole reason this question is settled.

The 200-plus cohort studies establish that the relationship exists and that it tracks by dose. That is a great deal of information and it is not causation. Observational data has a permanent weakness. People with high cholesterol differ from people with low cholesterol in a hundred ways, and no amount of statistical adjustment fully fixes it. Cohorts can show you a relationship. They cannot tell you which direction it runs or what else is riding along with it.

The other two got around that, separately, and by completely different means.

Some people are born carrying rare variants that impair the LDL receptor. They run high LDL from birth, through nothing they did, and they get more cardiovascular disease in proportion to how high it runs. Other people carry variants that lower it, and they get less. Nobody chose their variant, nobody's diet caused it, and there's no lifestyle confounder hiding underneath it. It's as close to a randomized trial as nature runs.

Then the drug trials arrived from the other direction. Statins, ezetimibe and PCSK9 inhibitors lower LDL by three entirely different mechanisms, and each one reduces events roughly in proportion to how far it lowered the number. The mechanism didn't matter. The size of the reduction did.

Two independent routes to the causal question, arriving from opposite directions, giving the same answer. That's what a settled question looks like, and it's rare enough in nutrition and longevity that I think it's worth saying out loud.

Why apoB and not just cholesterol

You will see me write apoB rather than LDL cholesterol in most places, and there's a reason.

ApoB is a protein, and every particle that can lodge in an artery wall carries exactly one copy of it. So counting apoB counts the particles, while LDL cholesterol measures the cargo they happen to be carrying that day. Two people can have identical LDL cholesterol and quite different particle counts, and it's the particles that do the damage.

I'll be honest that this is the less settled half. The causal evidence above was built on LDL cholesterol. That apoB is the better measurement of the same causal thing is well supported and increasingly standard, but it doesn't rest on the same mountain, and I'm not going to pretend otherwise.

There is one more thing about apoB worth knowing, and it's the reason I keep coming back to it. Among everything on the risk factor list, apoB-carrying lipoproteins appear to be the only one that can drive atherosclerosis even in the absence of other risk factors. Blood pressure matters. Inflammation matters. Insulin resistance matters. But those seem to need something to work with, and this is what they work with.

I'd flag that the mechanistic work behind that statement is largely animal and laboratory evidence, which sits low on my scale. I'm not resting my position on it. The human data already carries the argument, and the mechanism is simply consistent with it.

The outliers are real, and I'm not going to pretend otherwise

Some people put themselves on a ketogenic diet and their LDL goes up enormously. Not modestly. Several times over. It happens most in lean, metabolically healthy people with low triglycerides and high HDL, and the pattern is well enough described to have a name.

One of them has been documented in detail. A man in his thirties adopted a ketogenic diet for an inflammatory bowel condition and his LDL went from 95 to 574 mg/dL, with total cholesterol reaching 705. He held those numbers for nearly seven years, and his CT angiogram showed no coronary plaque at all.

That's a real person and a real result and it's genuinely interesting. A larger group with the same phenotype, 80 people averaging LDL of 272 for about five years, was compared against matched controls averaging 123, and the plaque scores were not significantly different between them.

I'm not waving that away. If the story were simple, those findings wouldn't exist.

What a case can do, and what it cannot

Here's where I part company with how this evidence gets used.

The paper describing that man says in its own conclusion that a single case can't redefine clinical practice. The researchers were careful. What happened after publication wasn't, and the case travelled a long way from the caveat its authors attached to it.

A single case can prove something is possible. It can't tell you how often it happens, who it happens to, or whether it will happen to you. Two million people with twenty million person-years of follow-up can. Those are different instruments answering different questions, and putting them on the same set of scales is the mistake I am actually arguing against here.

The group study is worth reading carefully too. It compared plaque at one moment in time. It couldn't tell you what happened to those arteries afterward, because it never followed them. A study that ran the longitudinal version of that question, and reported that apoB did not predict plaque progression, was retracted in May 2026 at the request of its own authors and the journal, after concerns about the methodology that the authors agreed were too large to correct. I mention it because that paper is still circulating as though it stands, and because withdrawing your own work when it won't hold is the system doing its job rather than a scandal.

So the honest state of the question: whether a large diet-driven LDL elevation carries the same risk as the same number arrived at another way is not settled. It's more open than it looked a year ago, and it's more open than either camp tends to admit.

The thing almost nobody checks

Here's the part I care about most, and it has nothing to do with the argument.

About one person in 300 carries familial hypercholesterolemia, an inherited condition that keeps LDL high from birth no matter how they eat or train. It's one of the most common genetic disorders there is, it carries very high cardiovascular risk, and it's badly underdiagnosed and undertreated worldwide. Even in wealthy countries, and even among patients already attending specialist lipid clinics, only a minority reach the targets their physicians set for them.

For the rarer and more severe form, a US health database query found people whose lipid profiles matched the condition, and 40% of them were on no lipid-lowering treatment at all. A formal diagnosis was uncommon.

In a room of 300 people, one of them has this. They probably don't know. And no camp in the cholesterol argument disputes any of it, which is exactly why I think it belongs at the end of a piece like this one.

Where this sits in the system

This is a Molecules piece, and it connects outward in every direction. What you eat moves these numbers. Training moves them. Bodyweight moves them. The Meals and Movement work in the rest of my writing isn't separate from this, it's the part you control without a prescription.

It also connects to how I read evidence generally, which is the actual reason this piece exists. Grades exist so that a case report and a two-million-person meta-analysis do not get quoted with equal confidence. That's not gatekeeping, it's the difference between knowing something and having heard something.

The Honest Take

The industry on one side sells you certainty about a number. The industry on the other side sells you the story of the man who beat the number. Both are selling, and the second is more fun to watch, which is most of why it travels.

Here's the honest version. The evidence that apoB-carrying lipoproteins cause atherosclerosis is the strongest evidence in this field, built from two independent routes to causation that agree, and I'm not going to soften it because it's unfashionable in certain corners. That isn't loyalty to a position. It's what following the data looks like when the data is inconvenient for somebody. Other factors are real and we're still working out how much they contribute. Genuine outliers exist and they should be studied rather than dismissed. And none of that changes the useful move, which is the same whichever camp you land in.

Find out where yours sits. Do it with a physician who will read it in the context of your history, and with a coach who'll work on the food and the training that move it. Then decide what to do with the answer.

Not knowing has never protected anyone.

Evidence

Highest evidence grade in this article: A · Proven. A is reserved for the strongest convergent causal evidence, genetic and randomized arriving at the same answer, and almost nothing in nutrition or longevity clears it. Each claim below is graded on its own, and a grade for one does not carry to the others.

• LDL causes atherosclerotic cardiovascular disease. A · Proven. The rare case where genetic and randomized evidence converge on the same causal answer. Meta-analyses of over 200 prospective cohorts, Mendelian randomization studies and randomized trials, over two million participants, more than twenty million person-years and over 150,000 cardiovascular events, with a dose-dependent log-linear relationship that strengthens with duration of exposure (Ference 2017, EAS Consensus Panel). A is reserved for exactly this pattern of evidence and almost nothing else clears it.
• ApoB is the better measure of the causal particle. Not graded here. Well supported and increasingly standard practice, but it rests on discordance analyses and guideline adoption rather than on the convergence above. I have deliberately not carried the A across to it.
• ApoB lipoproteins can drive atherosclerosis without other risk factors present, and their retention in the artery wall is an early step. E · Frontier. Mechanistic and preclinical work, including a transgenic mouse model in which binding-defective LDL produced significantly less atherosclerosis (Skålén 2002). Preclinical evidence grades E on my scale regardless of the journal, and my position does not rest on it.
• Large LDL elevations occur on carbohydrate restriction, particularly in lean metabolically healthy people, and some show no plaque. C · Moderate at best. One cross-sectional comparison of 80 such individuals against matched controls found no significant difference in plaque burden (Budoff 2024). Cross-sectional and observational, so capped at C, and it cannot answer a progression question. The individual case reports are single cases and are not graded.
• Whether diet-induced LDL elevation carries the same risk as other causes. Open, and not graded. The longitudinal study addressing it was retracted in 2026 at the request of its authors and the journal, for methodology errors the parties agreed were too large to correct. No finding is attributed to a retracted paper in either direction.
• Familial hypercholesterolemia affects roughly 1 in 300 and is widely underdiagnosed. C · Moderate. Epidemiological review evidence and registry data (Polychronopoulos 2021; Cuchel 2023, CASCADE FH Registry). Observational, which caps it at C, and prevalence estimates vary by population and ascertainment method.

Sources

• Ference BA, Ginsberg HN, Graham I, et al. Low-density lipoproteins cause atherosclerotic cardiovascular disease. 1. Evidence from genetic, epidemiologic, and clinical studies. A consensus statement from the European Atherosclerosis Society Consensus Panel. European Heart Journal. 2017;38(32):2459-2472. https://doi.org/10.1093/eurheartj/ehx144
• Skålén K, Gustafsson M, Rydberg EK, Hultén LM, Wiklund O, Innerarity TL, Borén J. Subendothelial retention of atherogenic lipoproteins in early atherosclerosis. Nature. 2002;417(6890):750-754. https://doi.org/10.1038/nature00804
• Norwitz NG, Feldman D, Soto-Mota A. Seven years of 700 cholesterol without coronary atherosclerosis: a lean mass hyper-responder case report. Diseases. 2026;14(5):168. https://doi.org/10.3390/diseases14050168
• Budoff M, Manubolu VS, Kinninger A, et al. Carbohydrate restriction-induced elevations in LDL-cholesterol and atherosclerosis: the KETO trial. JACC Advances. 2024;3(8):101109. https://doi.org/10.1016/j.jacadv.2024.101109
• Retraction notice: Longitudinal data from the KETO-CTA study: plaque predicts plaque, ApoB does not. JACC Advances. 2026;5(5):102824. https://doi.org/10.1016/j.jacadv.2026.102824
• Polychronopoulos G, Tzavelas M, Tziomalos K. Heterozygous familial hypercholesterolemia: prevalence and control rates. Expert Review of Endocrinology and Metabolism. 2021;16(4):175-179. https://doi.org/10.1080/17446651.2021.1929175
• Cuchel M, Lee PC, Hudgins LC, et al. Contemporary homozygous familial hypercholesterolemia in the United States: insights from the CASCADE FH Registry. Journal of the American Heart Association. 2023;12(9):e029175. https://doi.org/10.1161/JAHA.122.029175