My Cholesterol Went Up and My Doctor Told Me to Stop

Potholes, delivery trucks, and what your lipid panel isn’t telling you

The fight nobody told you about

Somewhere in the world right now, there are two doctors arguing on the internet about cholesterol. One says high LDL is poisoning people and needs a drug yesterday. The other says LDL is a distraction — what matters is your triglycerides, your HDL, and the state of the highway, not the number of trucks on it.

Both of them are part right. Both are part wrong. And the piece of the argument almost nobody is talking about is the piece I have spent my career watching in real patients: inflammation. The fire. Not the smoke — the fire.

Here is my argument in one paragraph, and then I will spend the rest of this article showing you exactly which parts the science supports, which parts it does not, and which parts it simply has not answered yet.

Cholesterol is not poison. It is building material that your body moves through your blood in little delivery vehicles. The vehicles are not the problem. The problem is the road. When the inner lining of your artery is healthy — smooth, slick, unbroken — the delivery trucks drive past harmlessly, all day, every day, for decades. But when that lining is inflamed, it develops cracks: microscopic tears in the surface. Now the trucks dump their cargo into the cracks. The cracks fill up. The lumps grow. A lump tears open — and that is a heart attack.

So my thesis is simple: a man with sky-high LDL and a perfectly smooth, un-inflamed road is fundamentally different from a man with moderate LDL and a road full of cracks — and your doctor’s blood test cannot tell the difference.

For a long time, that thesis had almost no high-quality science behind it. That has changed. In the last decade, medicine ran the experiments I always wanted to see — and the results are genuinely beautiful. Some of them prove me right. Some of them hit my argument with a two-by-four. I am going to show you all of it, including the parts I did not want to find. There is also a scandal in here: a study at the very center of this fight was retracted in March 2026, and we are going to read the whole story, because it changes what you are allowed to believe.

By the end, you will not have permission to ignore your doctor. You will have something better: you will understand what the argument is actually about, and you will know exactly which questions to bring to your own physician.

That is a fair trade. Let’s go.

The words. All of them.

Nothing here assumes you know anything. If you already know this stuff, skip ahead — nobody is grading you.

Cholesterol. A waxy, fatty substance your body both eats and makes. It is not a poison; it is a construction material. Every cell wall in your body contains it. Your brain is full of it — about a quarter of the cholesterol in your body lives there. You would die without it.

Blood. Mostly water. Fat and water do not mix — the same way butter will not dissolve in a glass of water. So your body cannot just “pour” cholesterol into your bloodstream and hope. It has to package it for transport.

Lipoprotein. The delivery vehicle. Think of a waterproof delivery van: a shell, cargo inside, a label on the side. These vans travel your bloodstream the way trucks travel a highway system.

LDL (low-density lipoprotein). The van that carries cholesterol out to the body. When your doctor says “your cholesterol is high,” he almost always means this number. (Here’s the confusing part: LDL is technically the vehicle, but when people say “LDL,” they usually mean the amount of cholesterol riding inside all the vans.)

HDL (high-density lipoprotein). The van that carries cholesterol back to the liver for recycling or disposal — the return trip. Higher is generally considered better, because it means your body has a functioning pickup-and-return system.

Triglycerides. The plain word is fat. Fat traveling in your blood, mostly on its way to storage. High triglycerides usually mean too much energy arriving all at once — the port is backed up and the tanker ships are circling.

ApoB (apolipoprotein B). The label on the side of every “outgoing” delivery van. Every single one of those vans carries exactly one ApoB label — one van, one label. So if you could count ApoB, you would be counting vans, not cargo. Standard cholesterol tests weigh the cargo instead. Two people can carry the same cargo in ten large vans or thirty small ones. Thirty is worse: more vans means more chances for one to be small enough to wedge into a crack. (This is why cholesterol-literate doctors and researchers increasingly measure ApoB — it counts the traffic, not the freight.)

Lp(a) (lipoprotein little-a). A particular kind of van that is unusually sticky. It is set mostly by the genes you were born with — diet barely moves it. Most people never get it measured. Most people should, once, ever. If yours is high, it changes how seriously you take everything else on this page.

Artery. The blood vessel that carries blood away from the heart. Think garden hose: the inside surface is smooth and slick. Smooth is the whole point — smooth is how flow happens.

Intima. The name for that inner lining. One cell thick in places. Beautifully smooth when it is healthy. It is the surface of the road.

Plaque. When that lining gets damaged, cholesterol works its way into the wall, underneath the surface. It builds up. That lump is plaque. Enough plaque, or a lump tearing open, and you get a blood clot and a heart attack. Important: plaque is inside the wall, like a pothole filling with water under the asphalt — it is not sitting in the tube of the hose narrowing it (that part comes much later). That’s why you can have a “widow maker” heart attack with normal-looking arteries on a plain stress test.

Inflammation. Your body’s repair-and-alarm system. Useful in short bursts — it’s why a splinter goes red. Harmful when it never switches off — then the repair crew never leaves, and the repair crew itself starts damaging the road it came to fix. It is the fire. Throughout this article, when I say “the fire,” I mean inflammation.

CRP / hs-CRP (C-reactive protein). A blood test that measures how much fire is burning right now. The body’s liver makes CRP when inflammation signals are loud. “hs” means high-sensitivity — a version sensitive enough to measure low levels, which is what you want for predicting heart disease.

ESR (erythrocyte sedimentation rate). An older, cruder fire gauge. Red blood cells settle through a tube of blood; inflammation makes them clump and settle faster. High ESR = fire somewhere.

Ferritin. The body’s iron warehouse protein. Most people know it as an iron test. But ferritin is also an acute-phase reactant — a fire alarm. It rises when the body is inflamed, even when iron is fine. So when I order it, I’m checking two things at once.

CT angiogram. A specialized CT scan that takes an actual picture of your coronary arteries and measures how much plaque is really there. Not a guess from a blood test — a photograph. This matters later: it is the only way to see the road itself.

Mitochondria. The power plants inside your cells — the furnaces that burn fuel (glucose and fat) to make the energy currency your body runs on (ATP). If the furnaces are overloaded and spilling sparks, you get a specific kind of cellular smoke called oxidative stress, which fans inflammation.

Ketosis / ketogenic diet (keto). When you cut carbohydrates low enough that the liver’s sugar reserves run dry, your body switches fuels and starts burning fat, producing molecules called ketones as its new fuel. Some people (including me, right now, as I write this — more on that later) run their bodies this way deliberately, for months at a time.

Statins. The most prescribed cholesterol drugs on Earth. They work by slowing down your liver’s cholesterol factory (an enzyme called HMG-CoA reductase). They lower LDL substantially — and, interestingly, they also lower CRP. That last part matters for our story.

Right. Those are all the words. Now the argument.

The one picture that holds this whole fight

Your arteries are a road. The delivery vans (LDL) drive on it.

• If the road surface is perfect, vans drive over it all day and nothing happens. They deliver cargo, go home, repeat, forever.
• If the road is cracked, cargo works its way down into the cracks. It packs in. The crack becomes a bump. The bump is plaque.

Hold that picture. The entire cholesterol war — every expert shouting at every other expert — reduces to one question:

Is the problem the number of vans, or the state of the road?

Camp one (mainstream cardiology, roughly): fewer vans, less cargo in the cracks. Cut the van count and you are safe. This is why LDL dominates every lipid panel and why statins — the great van-reducers — are everywhere.

Camp two (keto/low-carb community, roughly): fix the road and the vans don’t matter. No cracks, nowhere for cargo to go. This is why they watch triglycerides, HDL, and inflammatory markers, and why some of them shrug at an LDL of 250.

My clinical experience — twenty-plus years of watching labs and people — put me firmly in camp two for a long time. Then the last ten years of randomized trials happened, and I had to grow up a little. The evidence says both camps are right, and each is wrong about the other. I will show you the proof. But first, we have to deal with the thing that brings most people to this article in the first place: your LDL just went up, and your doctor told you to stop.

Why your LDL went up — and the part almost nobody tells you

Here is the thing that will make you feel less crazy.

When you cut carbohydrates, some people’s LDL barely moves. Other people’s LDL rockets. Same diet, wildly different result. This is not you doing it wrong. There is a pattern to who rockets — and it surprised everyone who studied it.

It is not the heaviest people. It is the leanest.

A 2024 meta-analysis pooled the data from 41 randomized controlled trials of low-carbohydrate diets and found the pattern plainly: on a low-carb diet, LDL rises in adults of normal body weight — and does not rise (and often falls) in adults with high body weight. Same diet. Opposite result. The dividing line is how much of you there is. The same analysis found that simply having a BMI under 25 was more than five times as powerful a predictor of an LDL rise as eating the top quarter of saturated fat — a result that upends the simple “fat in the diet raises LDL” story most of us were taught. ¹

Why would leanness matter? Think of the vans as a fuel delivery system. A lean person on a very low-carb diet is burning fat around the clock. The body ramps up the fat-delivery fleet to feed the furnaces. More fuel being hauled = more vans on the road — and the cargo they’re carrying is cholesterol-rich LDL. It’s a metabolic response, not a dietary error. That is the current best explanation, and it fits the data remarkably well.

Researchers even named the extreme version: the lean mass hyper-responder, or LMHR — say it once, then just think of them as lean rocketers. The definition, from the team that first described it: on a ketogenic/carbohydrate-restricted diet, LDL of 200 mg/dL or higher, PLUS HDL of 80 or higher, PLUS triglycerides of 70 or lower. ² Look at that triad again. The “bad” number is terrible. The other two numbers — the ones that tell you about metabolic health — are spectacular. That is the entire controversy in one blood panel.

The “good panel” check — the way the keto world looks at it. I will tell you the version my side of the fence actually uses, because it is not the nonsense you sometimes hear on podcasts. The targets, in adults:

• Triglycerides under 150 mg/dL (the lower, the better; on low-carb, people routinely land under 100, sometimes under 70)
• HDL above 40 for men, above 50 for women (and on healthy low-carb, experienced practitioners like to see men in the 50s–70s and women in the 60s–80s)
• A triglycerides-to-HDL ratio under about 3.5 (in mg/dL units) — this ratio has been validated over and over as a marker of insulin resistance; the higher it climbs, the more insulin-resistant — “the cells don’t hear the insulin knock” — the more likely you have the small dense vans and the underlying metabolic disease driving the whole show. ³

Why do those two numbers matter so much? Because they are the road report: high triglycerides + low HDL is the classic fingerprint of insulin resistance and the inflamed, dysfunctional metabolism that cracks the road. When someone on keto shows you triglycerides of 68 and an HDL of 92 — the “lean rocketer” pattern — they are showing you a metabolic panel that most sedentary sick patients would sell a kidney for.

Why does that not settle the argument? Because the vans are still on the road, and we still have seventy years of accumulated evidence — from genetics, from clinical disasters, from drug trials — that vans matter too; the receipts are in Parts Six and Eleven. The “road report” tells you a lot. It does not tell you everything.

Now the honest part about me, because I am one of these people.

I started this experiment at a body mass index of about 31 to 32 — the “high body weight” group, the group whose LDL does not rise on carbohydrate restriction. Since then I have lost a great deal of weight. I am now in the lean group. The group whose LDL does rise. My lipid response to my own protocol may have shifted underneath me while I was busy congratulating myself, and I have been running on blood ketones of 4 to 6 millimoles per liter since day one of 180 — day 78 as I write (Sep 2, 2026). And here is the thing I want you to notice about the timing: I have not drawn my own lipid panel yet. I am publishing this anyway, on purpose.

If I waited until I had my results, you would have no way of knowing whether I wrote this article to fit them. Every argument in here would arrive pre-shaped by an answer I already had. So the argument goes out first, while I am still in the dark, and the numbers follow.

That is the same thing I did with my symptoms. I scored all ninety-nine of them on June 17th, before I knew how any of it would end, precisely so there would be something honest to be measured against later. This is that, applied to my arteries.

When the panel is drawn it gets published here in full — LDL, ApoB, Lp(a), triglycerides, HDL — whatever it says. I have failed this protocol twenty-three times and published every failure. This would just be the most expensive thing I have ever been wrong about. I have failed this protocol 23 times and published every single failure. This would just be the most expensive thing I’ve ever been wrong about.

This is the natural coffee break. Bookmark here if you need to. The next part is where things get strange.

The Oreo experiment. Yes, really.

I promised you one strange thing, and here it is.

A researcher named Nick Norwitz is a lean rocketer himself. His LDL on a ketogenic diet ran to extremely high levels. He wanted to test an idea: if this kind of LDL is a metabolic response — the fuel-delivery fleet ramping up because the body is running on fat — then adding carbohydrate back should drop it. Not slowly. Fast.

So he ate twelve Oreo cookies a day for sixteen days.

His LDL fell 71 percent. From 384 to 111 milligrams per deciliter — a drop of 273.

Then he ran the comparison. He took rosuvastatin — one of the strongest statins — at 20 milligrams a day. His LDL fell 32.5 percent — a drop of 137 points.

The cookies were roughly twice as powerful as the high-intensity statin. In him. This was published in a peer-reviewed journal — Metabolites, 2024 — and yes, the title literally says Oreo. ⁴

Before anyone takes this somewhere stupid, let me slow down, because this study gets misused constantly.

• It does not mean Oreos are good for you. They are not. He is not claiming they are.
• It does not mean eating cookies before a blood test fools your doctor. I was told this story by a man with an LDL over 400 who genuinely believed it — which brings me to a job interview I once had, because this is where the story starts for me. The man interviewing me worked in a medical office. He told me his LDL and total cholesterol were crazy high — four hundred plus — and his doctors were furious. Then he leaned in and told me his secret: if he ate a few Oreo cookies right before his blood test, his LDL would drop “drastically to normal.” I remember thinking: that is not how any of this works — and, more importantly, that this very smart, very stubborn man was about to make health decisions on a garbled version of a real study. The truth is that this took sixteen days of twelve cookies a day. Four cookies on the morning of a blood draw does essentially nothing. The man was right that the numbers moved and wrong about why, the same way a man who wins the lottery twice is right that the numbers came up — the mechanism is the whole story.

What the Oreo study actually shows is genuinely interesting. In this kind of person, LDL behaves like a thermostat, not a bank balance. It is responding, hour by hour and day by day, to how much energy is arriving and in what form — not slowly accumulating the consequences of what you ate in 2011. When carbohydrate comes back, hepatic (liver) glycogen — the liver’s sugar reserves — refills, and the body downgrades the fat-delivery fleet. That is a charge against the naive story that “high LDL on keto is just genetic bad luck.” It is a metabolic number in these people.

Now — one person. One. A heroic n-of-1 with a case of the giggles. And I have to tell you something else about him, which brings us to the hard part of this article: the study I most wanted to build a reassuring article around was retracted before I could finish writing it.

The study I wanted to build this article on was retracted

For several years, the lean rocketers had a hypothesis. It went like this: our high LDL is different. It comes from a healthy metabolism, not a sick one. So it should not cause the same damage that ordinary high LDL causes.

That is a testable idea. And to their enormous credit, they tested it properly — with a real prospective study, on 100 lean rocketers, with actual pictures of their coronary arteries taken by CT scan at the start and again one year later to see whether plaque had grown. Published on April 7th, 2025 in JACC: Advances, it was titled: ”Longitudinal Data From the KETO-CTA Study: Plaque Predicts Plaque, ApoB Does Not.” ⁵

The message the internet took from it was simple and enormously reassuring: your existing plaque predicts your future plaque. Your van count does not. If you are lean and metabolically healthy, high LDL does not appear to hurt you.

That paper became the foundation of nearly every reassuring article you will find on this subject — including the one I was planning.

In 2026, it was retracted. ⁵ The notice is short, and I will give it to you in the journal’s own words, because they matter: the paper was retracted ”at the request of the authors and the Editors.” Concerns had been raised about the methodology, concerns that went to ”the reliability of the data.” And the errors were ”too great to be corrected” by publishing a correction. The paper had to go entirely.

Not by an enemy. The authors themselves asked for it to be withdrawn, together with the editors. Norwitz — the Oreo man — published a piece titled “We Want to Retract Our Own Paper” explaining why. ⁶

Here is what actually happened. After publication, questions were raised about the methods — critics had formally argued the study did selective reporting, had no proper comparison group, and used too short a timeframe. ⁸ Then the deeper blow landed. The imaging analysis for the entire study — the thing the whole conclusion rested on — had been produced by an outside commercial company called Cleerly, which read the heart scans with AI. Dave Feldman, one of the researchers, dug into the raw data after publication and found what he called highly anomalous patterns. Worse: the before-and-after scans had been read unblinded — meaning the company knew which scan was the “before” and which was the “after,” a cardinal sin in a study measuring change over time. The authors asked Cleerly to redo the reading properly blinded. The company refused — repeatedly. Two independent re-analyses (one by HeartFlow, one by a second method called QAngio) agreed with each other and disagreed with Cleerly. And there was a conflict of interest that readers were never told about — though the failure here belongs to the journal, not to the man himself. One of the paper’s authors, James Earls, was Cleerly’s chief medical officer. He told the journal about his connection to the company when he submitted the manuscript. When the paper was accepted, he went further and told them he held equity in Cleerly as well. The journal printed his affiliation and left the equity out. ⁷

The other authors say they did not know he held that role at the company when the scans were being read. ⁷

Then came the part that settled it. A group of the participants, working through their own cardiologists, quietly sent their scans back to Cleerly a second time. Because the company had no idea these were the same scans, it read them blind. Eight people did this. In the original unblinded reading, those eight showed an average plaque increase of 20.9 cubic millimetres. Read blind, the same eight scans showed an average decrease of 5.1. Progression became regression. Same scans, same company, one difference: this time nobody knew which scan was which. ⁶

I want to say the fair thing here, because it matters. Asking to retract your own paper is the hardest thing in science, and almost nobody does it. They did. That is the behaviour I want from researchers, and they showed it. The journal attached a formal expression of concern to the paper in January, and the retraction notice followed on the eleventh of March. ⁷

Now let me give the other side its say, because I would be doing exactly what I accuse others of doing if I left it out.

Not everybody thinks this story ends with the authors as heroes. Brad Stanfield, a family doctor in New Zealand who has criticised the study from the start, says the new analysis ”raises new concerns rather than resolving the old ones,” and that presenting the same data as confirming the original story is ”more advocacy than science.” ⁷

His sharpest point is one I cannot answer. If the original methods were good enough to be promoted as revolutionary, they should not have needed retracting. If they needed retracting, then the publicity campaign that came with them was premature. The institute where the scans were done announced the work as a paradigm shift. Patients read that. Some of them made decisions about their own cholesterol on the strength of it.

In his words: ”The public, including patients making clinical decisions about their own LDL, deserved more humility from the start.” ⁷

He is right. And I will hold myself to the same standard in this article, which is why you are reading about a retraction instead of a revolution.

Another critic, the internist Michael Mindrum, put it harder still — he called two of the authors social-media influencers with an outsized footprint, and said the whole affair sits ”at the intersection of social media grifting and medical science.” ⁷ I think that is too harsh on people who did, in the end, pull their own paper. But you should know it was said, and by a doctor, not a stranger on the internet.

One more thing you deserve to know, since I have asked you to weigh who did what. Dave Feldman, the researcher whose deeper look at the data started all of this, is a software engineer, not a physician. ⁷ That cuts both ways, and I will let you decide which way it cuts. He is the one who found the anomalies. He is also not a doctor.

But the reassuring evidence is now gone. Not weakened. Withdrawn. It cannot be cited by me, or anyone, as support for anything.

And there is more you should know. In January 2026, the authors posted a preprint — a paper posted before other scientists have checked it; treat a preprint as a rumor with footnotes, not a finding — reporting a re-analysis of the same 100 participants using the two independent, blinded methods. ⁹ Here is the thing that will surprise you, because the internet has not caught up with it:

• The retracted paper’s own analysis reported roughly three times more plaque progression than the later blinded analysis would find — real, measurable progression. That is why independent commentators titled their write-ups “Lean Mass Hyper-Responders See Rapid Plaque Progression” back in April 2025.
• The re-analysis preprint reports median progression of +5.6 mm³ — far below the withdrawn figure — and describes it as “modest and heterogeneous.” Fifteen of the ninety-nine participants, 15 percent, actually went backwards: their plaque shrank, while their LDL sat at an average of 242 and their ApoB at 180.

But I promised you the parts I did not want to find, so here is one.

That same preprint reports something the headlines skipped. Using the study’s own pre-registered definition of rapid progression, 21 percent of participants met it. By the second method, the blinded HeartFlow reading, 29 percent did. So somewhere between a fifth and nearly a third of these lean, metabolically healthy people were progressing quickly by the study’s own yardstick. One needed a procedure to open an artery during the year.

That is not a reassuring number, and it sits in the same paper as the reassuring ones. Both are true. This is what “heterogeneous” actually means when you look underneath it: some people did well, some people did badly, and the group average hides both.

And the authors add a caution I want to repeat, because it is the honest kind. More than half of these participants had no measurable calcium in their arteries at the start. The software that measures plaque was never built to measure almost-nothing, so when you are looking at very small amounts, a good deal of what you see may be noise. There was also no comparison group. The authors say all of this themselves, in print, which is to their credit.

So the honest position today is not “high LDL on keto is fine” (that evidence was withdrawn) and not “high LDL on keto is proven harmful” (that evidence is not settled, and the replacement analysis shows much less progression than the original). The honest position is: we are back to not knowing — which is precisely where the argument stood before April 2025. The authors insist the headline finding (LDL didn’t predict which individuals progressed) survived every re-analysis. Leading critics — including cardiologists who have followed this closely — call the re-framing “more advocacy than science.” I am standing in the middle of that fight, in public, at 4 to 6 millimoles of ketones, on day 78 of 180, and I will tell you plainly: the strongest claim my side of the fence used to make is now running on fumes.

So is it the vans, or the road? Both. Here is the proof

This is the part I find genuinely beautiful, because the experiments are so clean. Two drugs, two opposite targetings, both tested in enormous randomized trials — the kind where nobody — not the patient, not the doctor, not the person reading the scans — knows who got the drug and who got the dummy.

Experiment one: fix the road, leave the vans alone.

There is a drug called canakinumab. It is an antibody that quiets the interleukin-1β alarm bell — one specific cry in the inflammation fire department. It does nothing to cholesterol. In the CANTOS trial, researchers gave it to over 10,000 people who had already had a heart attack. ¹⁰

Their cholesterol did not fall. Their heart attacks did. Stroke, heart attack, and cardiovascular death fell by about 15% — and the protection concentrated in the patients whose inflammation blood test (CRP) came down the most. Fix the road, leave the vans. Fewer events.

Experiment two: change the vans, leave the road alone.

There is a class of drugs called PCSK9 inhibitors (pronounced pee-see-es-kay-nine). They pull LDL out of the blood with unprecedented force. In the FOURIER trial, 27,564 people on statins got one of these (evolocumab) or a dummy. ¹¹

Their LDL fell by 59 percent — from 92 down to 30 milligrams per deciliter, on average. Their inflammatory marker (hs-CRP) barely budged. And their heart attacks still fell — about 15% fewer major events, consistent across every level of inflammation. Fewer vans, same road. Fewer events. A second enormous trial, ODYSSEY OUTCOMES, repeated the feat with a similar drug (alirocumab) in 18,924 people who had just had a heart attack: again, big LDL cut, big event reduction. ¹²

Both experiments are randomized, both are huge, and they point in opposite directions from the same starting point: inflammation is causally harmful on its own. LDL is causally harmful on its own. Neither cancels the other.

Which means the version of my own argument I most wanted to be true — fix the fire and the traffic stops mattering — does not survive contact with that evidence. I am telling you this because it is true, and because a physician who only tells you the parts of the science that flatter his thesis is a salesman, not a doctor.

And there is one more thing I have to put in, because leaving it out would be dishonest.

Some people are born with genes that keep their LDL very high for life. The severe form, homozygous familial hypercholesterolemia (HoFH), shows up in childhood — LDL often over 500. Those children develop serious artery disease young. Some have heart attacks as teenagers — there is a 2026 case report of a myocardial infarction in a 13-year-old with this condition. ²³ If untreated, these patients typically have a cardiovascular event in childhood or adolescence and die before thirty. ²³

A twelve-year-old has not spent three decades eating badly. There has been no time to wreck the road. The vans did it alone. You can argue — and I have argued — that these children still eat a modern diet and still carry some inflammation. That is fair for the adults with the milder form. It is much harder to argue for a child whose arteries are filling in grade school.

And then there is the genetics work. Researchers can compare people who happened to be born with genes that keep LDL slightly lower for life against people who were not — the closest thing to a lifelong randomized trial we will ever get, because your genes were dealt before your diet was. Lower lifelong LDL, less heart disease. Consistently. Dose for dose. ²⁴

I do not get to wave that away. So I won’t.

What all of this does to my thesis: it doesn’t kill it — the road is real, the fire is real, and the fire is causally dangerous on its own. But it forces my thesis into its honest final form: the number of vans matters, the state of the road matters, they add, and the “keto panel” that shows great triglycerides and HDL but sky-high LDL is giving you good news and bad news at the same time. Anybody who tells you only one side of that is selling you something.

The road, up close — and the fire your test is missing

Let’s spend a minute on the half of the argument your doctor’s standard lipid panel is blind to.

The modern scientific version of “cracks in the road” has a name: the response-to-injury hypothesis — and its defining statement is a paper titled simply ”Atherosclerosis — An Inflammatory Disease,” by Russell Ross, published in the New England Journal of Medicine in 1999. He argued — with evidence that has since been cited over 36,000 times — that atherosclerosis is not primarily a plumbing problem. It is a wound-healing problem that goes wrong. Injure the lining (the endothelium/intima), and the body mounts an inflammatory repair response: immune cells arrive, the lining becomes leaky, cholesterol seeps into the wall, and the “repair” becomes the plaque. ¹⁶

This is my thesis, in mainstream peer-reviewed form, from before I started practicing medicine. The fire is not a side effect of the disease. The fire is the disease.

What starts the fire? In a modern Western body: a lot of things, and they stack. Excess sugar and refined carbs producing repeated insulin surges. Fat storage cells (adipocytes) that are overstuffed and start sending out alarm signals — inflamed fat tissue is an active, fire-spreading organ, not a passive closet. Poor sleep. Chronic stress. Smoking. High blood pressure mechanically battering the lining. Autoimmune simmer. And the mitochondria — the furnaces — running overloaded and spilling sparks.

How do you see the fire on a blood test? Three old friends:

• hs-CRP — the best single gauge of systemic fire we have for heart risk
• ESR — the old man of the trio, still useful
• Ferritin — the double agent (iron stores and fire alarm at once)

And here is where I defend a clinical observation I have made for years: when I see patients with high CRP, high ESR, high ferritin — no matter what medicines they are on — I do not see the outcomes getting better. The lipid-lowering drugs are doing their lipid job, and the patients are still burning. Patients in that state can have a “perfect” LDL on paper and still be walking time bombs, because their road is full of cracks and nobody is looking at the road — only at the traffic report.

What the research says now: this observation has moved from “Dr. So-and-So’s clinical hunch” to one of the most-studied phenomena in cardiology. It has a name: residual inflammatory risk — the risk that remains after you’ve fixed the LDL, because the fire is still burning.

• Across the big secondary-prevention trials, 43 to 47 percent of patients still had an hs-CRP of 2 or higher — nearly one in two, with an actively burning road, on treatment. ¹⁷ In Americans who had already had a heart attack, the figure was 61 percent, and standard risk factors explained only about a fifth of the variation. Your usual risk checklist is not measuring this.
• And here is the cleanest evidence I can give you that the fire matters on its own. In FOURIER, the trial where LDL was driven down to a median of 30, look at the patients who got it below 20 — as low as medicine can currently push it. Over three years, their rate of heart attack, stroke or cardiovascular death was 9.0 percent if their CRP was under 1, 10.8 percent if it sat between 1 and 3, and 13.1 percent if it was above 3. ¹⁷ Same crushed cholesterol. Different fires. Different outcomes.
• And when patients with persistently high CRP were followed after stenting, their rate of heart attack, stroke or death ran at about 15 percent a year, against 6 percent in those whose CRP stayed low — more than double the risk, and more than triple the death rate, after adjusting for everything the researchers could think of. ¹⁷
• Those patients — LDL fixed, fire burning — continue to have high rates of recurrent events, and in study after study their hs-CRP on treatment predicts future heart attacks, strokes, and death better than their LDL does — a finding confirmed in a 2023 re-analysis of FOURIER. ¹⁷
• And when you do treat the fire — canakinumab in CANTOS, colchicine in two trials, COLCOT and LoDoCo2 — you get event reduction with LDL unchanged. In COLCOT, 4,745 people just after a heart attack took a single 0.5 milligram pill a day, and their heart attacks, strokes, and cardiac deaths fell from 7.1% to 5.5%. In LoDoCo2, 5,522 people with established heart-artery disease saw their events fall from 9.6% to 6.8% — a 31% relative reduction. ¹³¹⁴ Honesty demands one more sentence: in LoDoCo2 there was a signal toward more non-cardiovascular deaths with colchicine that the investigators themselves flagged as unclear. The fire is dangerous. The fire extinguishers we have are imperfect. That is the current state of the art.

And the wild card that ties it all together: the JUPITER trial. It took 17,802 healthy people — LDL under 130 (which guidelines said was fine), no heart disease — who happened to have hs-CRP of 2 or higher (fire present, traffic normal). Half got rosuvastatin 20 mg, half got a dummy. The trial was stopped early because the statin group was winning by so much: primary events 0.77 vs 1.36 per 100 person-years — a 44 percent relative risk reduction. ¹⁵

Read that carefully, because it cuts both ways:

• It proves the fire predicts. People with normal LDL and a burning road were at real risk — the “normal cholesterol” reassurance would have missed them. That is exactly the “road report” argument my side has been making for years.
• But the thing that saved them was a statin — and statins lower LDL and CRP at once. In JUPITER the benefit tracked how far the LDL came down, and the pattern is hard to argue with. People whose LDL did not fall got almost no benefit. People whose LDL fell by less than half got a moderate benefit. People whose LDL fell by more than half got the largest benefit of all. The drug’s message was: bring the van count down and you win. JUPITER did not test whether a diet-and-lifestyle “super anti-inflammatory state” would have done the same job. Nobody has ever run that trial. If you meet a doctor or a podcaster who claims it has, they are wrong, and you can tell them I said so.

And now the caveat I owe you about my own side of this argument, because it is a serious one.

CRP is a fire alarm. It tells you a fire is burning. But some people are born with genes that keep their CRP high for life, and when researchers look at those people, they do not find more heart disease. That is the same kind of natural experiment that convicted LDL — and for CRP it comes back with nothing.

So the honest reading is this. CRP is a thermometer, not the fire. Measuring it tells you something real about your risk. Lowering the number itself would probably do nothing. What CANTOS showed is that quieting a specific alarm deep in the immune system — interleukin-1 beta, the one sitting beneath that smoke-alarm switch I described earlier — reduces heart attacks. The signal is real. The thermometer is not the thing to treat.

Anyone who tells you to chase your CRP number the way you chase your LDL number has skipped that distinction. I have made that mistake in conversation, and I am correcting it here.

One more honest note about JUPITER: the trial was stopped early, after a median of under two years, because the statin group was so far ahead. Trials stopped early for benefit tend to overstate the size of that benefit. The direction is not in doubt. The exact magnitude is a little softer than the number suggests.

So: the fire matters, the fire is measurable, the fire predicts events independently of LDL, and quieting the fire with drugs reduces events even when LDL sits still. And yet your standard doctor’s appointment will likely never look at your CRP — only your LDL. That asymmetry is the single most important fact in this whole debate, and we are going to build on it in the next part, because it showed up in my clinic every single day.

What I see when the fire is burning — and whether the science backs me up

Someone is going to read this article and say the clinical observations are anecdote dressed as evidence. They’re right that they’re anecdote — these are my cases, not randomized trials. That’s exactly why I ran the literature on each one, and below each observation I’ve written what the research actually says: where it backs me up, where it doesn’t, and where it’s genuinely undecided. Here they are, checked.

Observation 1: “Their ‘normal’ is their inflamed friends.”

Patients tell me they sleep fine, feel fine, think fine — and their labs and their lives say otherwise. Their baseline isn’t health; it’s the health of everyone around them. I knew I was in serious trouble before I changed course — BMI about 31–32, looked better than many people my age and younger, and I was a wreck. Most patients never get the wake-up call, because they’re too busy paying rent to notice they’ve been running on 40% power for a decade.

The science check: There’s no randomized trial of “personal baseline delusion,” so I’ll be honest — this is exactly what it is, a clinical observation. What I can say is that it sits on solid ground: the metabolic syndrome (the cluster of insulin resistance, high triglycerides, low HDL, high blood pressure, and abdominal fat — the inflamed-metabolism package) is now present in over a third of American adults, meaning the “normal” a patient compares himself to is frequently itself a disease state. And the observation has a known mechanism behind it: your brain’s internal norms are built from your environment, and metabolic decline is slow — a decade of slow degradation doesn’t come with a “you are now worse than you were” notification. It’s seductive and it’s dangerous, and it’s why I ask about deep sleep and dreams rather than just “do you sleep” — that is the subject of Observation 2.

Observation 2: “If they fall asleep, they call it sleep.”

They don’t remember what it’s like to sleep deep and dream. Falling asleep isn’t the same as restoring — the body runs a nightly repair shift in deep (slow-wave) sleep and a nightly memory-consolidation shift in dream (REM) sleep, and an inflamed, disrupted night skips both.

The science check: the research is unambiguous in direction. A landmark meta-analysis of 72 studies found that sleep disturbance is associated with higher CRP and IL-6 — the fire markers — in adults. ²⁵ Experimental partial sleep deprivation — just a few nights of short sleep — reliably raises IL-6 and CRP in healthy volunteers; a 2025 meta-analysis of human experimental studies concludes that partial sleep deprivation for at least three nights triggers measurable increases in both. Backs me up: poor sleep fans the fire, and an inflamed body sleeps poorly — a vicious circle the patient experiences as “I just don’t sleep the way I used to.”

Observation 3: The phone number.

For a while my short-term memory was bad enough that I couldn’t hold a seven-digit phone number in my head. My workaround: memorize three digits, write them down, look back, get the next four. I got through life. That’s the thing — the workaround works. So patients don’t think of it as a problem. They don’t notice they’ve degraded, because they’re too busy doing the workaround. (They juggle three phone numbers at once and call the workaround a success.)

The science check: this observation — metabolic illness tracking with early cognitive decline, especially working memory — has real literature behind it. The 2004 JAMA study by Yaffe and colleagues was the first to document that the metabolic syndrome is associated with poorer cognitive outcomes (in older women followed over time). ²⁶ Follow-up work has repeatedly found the metabolic syndrome tracks with accelerated cognitive decline, and more metabolic syndrome components means worse cognition even in midlife. ²⁶ Honest caveats: these are associations, not proof of mechanism, and “my memory got worse” is not a diagnosis. But the direction is exactly what I see: the metabolically sick brain ages faster, and the person copes with workarounds instead of ever noticing.

Observation 4: They can’t stay in the gym.

I’ve watched patients start a workout program with real motivation and quit within weeks — not from laziness, but because inflamed muscles and joints cannot sustain it. They push, they’re laid up for two weeks, they quit. We call the resulting decline “normal aging.” It isn’t. It’s inflammation’s tax on the engine.

The science check: here the research is genuinely on my side. In the InCHIANTI study — a famous Italian population study of aging — higher blood levels of IL-6 and CRP predicted significantly greater loss of muscle strength (and, for IL-6, muscle mass) over three years. ²⁷ A 2024 meta-analysis confirms: higher IL-6 is associated with sarcopenia — the loss of muscle mass — in older adults. ²⁷ The honest footnote: the arrow could point both ways — sick mitochondria and sedentary life feed inflammation, and inflammation feeds muscle loss. It’s a loop, not a one-way street. But the clinical reality — inflamed people can’t train, so they don’t, so they get more inflamed — is exactly what the loop predicts.

Observation 5: The bicep that popped.

I had a patient, a retired sanitation worker in his early sixties, who was in the gym doing curls when his left bicep popped — a distal biceps tendon rupture. Two or three years earlier he’d been lifting 85-pound cans three or four hours a day. I’ve seen more than one patient like this. Their bodies are inflamed, their tendons and muscles are degraded, and so the structures fail at loads they once handled easily. He had horrible labs. He was a walking time bomb with a fresh tendon rupture.

The science check: the honest read here is mixed — with real literature pointing the right way, and one important correction to my own assumption. I originally assumed the statins most of these patients take were a major suspect. The true picture:

• There is a body of case reports, pharmacovigilance data, and a review concluding statins may be associated with tendinopathy and tendon rupture, especially of the Achilles, quadriceps, and distal biceps — precisely the tendon in question. ³¹
• But a large propensity-matched cohort study found no increased risk of Achilles or biceps rupture in new statin users — the strongest design on this question came back negative.
• Independently of statins, there is an established body of work linking hyperlipidemia, diabetes, and metabolic syndrome with tendon degeneration — “tendinosis” — a story that fits my patient well: years of hard physical labor on a metabolically failing, inflamed body, then rupture at a routine load.

Verdict: the observation is biomechanically plausible and consistent with the connective-tissue literature, but I cannot honestly claim any single cause for his rupture, and I don’t know what medications he was on. I’m keeping the story and downgrading the certainty. That’s what a check-up is for.

Observation 6: “Poisoning their mitochondria.”

People eating massive carbs, inflamed, metabolically ill — I say their mitochondria are being poisoned. The furnaces are being force-fed.

The science check: this one is real mechanism, not metaphor — with one caveat about direction. There is a large literature on lipotoxicity and glucotoxicity: when cells are flooded with more fuel than the mitochondria can burn, the overflow damages the machinery — spilled sparks (reactive oxygen species), stressed-out quality-control compartments (ER stress), and alarm cascades (JNK) — which desensitizes cells to insulin and amplifies inflammation. Excessive refined carbohydrate, and the insulin surges it causes, are among the most reliable ways to run this overload in a human. ³³ The caveat: for many associations it’s still debated whether mitochondrial damage is the cause or the consequence of metabolic disease — probably both, in a loop. But “eat less refined carb, give the furnaces a break” is one of the best-supported interventions in metabolic medicine, so the practical conclusion survives the caveat.

The one clinical claim I cannot back up — and I’m not going to pretend otherwise. I believe — strongly — that a “super anti-inflammatory state” (low-carb metabolic health, no excess energy, sleep, movement, minimal toxins) protects the road so well that high LDL loses most of its power to hurt. I have seen it change people’s lives, including mine. And when I went looking for the proof that such a state erases LDL risk, I found three things. First: proof that inflammation causes events and quieting it with drugs reduces them — the CANTOS, COLCOT, and LoDoCo2 trials in Parts Seven and Eight. Second: proof that an anti-inflammatory lifestyle improves the road report — triglycerides, HDL, CRP —, which I set out below. Third, and this is the gap at the center of my belief system: no randomized trial anywhere has tested a lifestyle anti-inflammatory state against heart attacks directly, and no study shows that a great road report cancels high LDL. That’s the gap at the center of my belief system, and this article is me standing in it, in public, at 4–6 mmol/L of ketones, day 78 of 180. And I am telling you this before I have my own numbers, not after. If the panel comes back ugly, you will read it here. That is the deal I made with myself, and the reason I made it in public was so that I could not quietly renegotiate it later.

What actually moves which number — the diet manual in plain language

Let’s be practical now. You want the map of which lever moves which number. Here it is, with the studies.

Carbohydrates — especially sugar and refined starch. Raised triglycerides? Carbs are your first suspect. The systematic review by Te Morenga and colleagues (2014) found dietary sugars raise blood pressure and serum triglycerides, independently of weight. ²⁸ A network meta-analysis including 30+ trials found that substituting sugar for starch raises LDL — meaning the reverse, swapping sugar out for starch or whole foods, lowers LDL. ²⁸ So the popular claim “my high LDL came from carbs” is partly true: excess sugar pushes triglycerides up and LDL up, and HDL down. But the truth on a typical high-carb American diet is that the LDL number is dominated less by carbs than by the saturated fat riding along with the processed food — and, in lean people, cutting carbs can raise LDL — the lean rocketer again. The honest instruction to a patient with high LDL: cut the sugar and the refined carbs first — it reliably fixes triglycerides and HDL and modestly helps LDL — and then look at fat quality.

Fat quality. Saturated fat (butter, lard, fatty red meat, palm oil) raises LDL — this is one of the most reproduced findings in nutrition. Unsaturated fat (olive oil, nuts, avocados, fatty fish) lowers LDL and triglycerides. A 2021 systematic review of food effects on LDL ranks this landscape clearly. ³⁵ If your LDL is the problem and your triglycerides are already low, the fat mix is where to look before you panic about eggs (dietary cholesterol barely moves blood cholesterol in most people).

Exercise. The 2007 meta-analysis of 25 randomized trials found aerobic exercise raises HDL by about 2.6 mg/dL on average — modest, but real — with each extra 10 minutes per session adding roughly another 1.4 mg/dL, and a minimum volume (about 120 min/week) below which nothing happens. Exercise also lowers triglycerides meaningfully; LDL, honestly, barely budges with exercise alone. ²⁹ Translation: exercise is a road drug, not a van drug. It’s essential, and it won’t fix a bad LDL by itself.

Protein. The claim that higher protein intake helps HDL is supported: a meta-analysis of 88 trials found higher-protein diets raised HDL, and the observational data agree. ³⁰ The effect is modest, and it’s confounded by the fact that protein usually displaces carbs. Eat the protein, don’t worship it.

The “keto panel,” summarized honestly. Cutting carbohydrates (very low-carb / ketogenic):

• Triglycerides: reliably down — usually dramatically
• HDL: reliably up
• LDL: variable — unchanged or slightly up in most trials; up a lot in lean people (the lean-rocketer minority)

That is the whole controversy in one sentence, and both sides of this fight are looking at the same three facts. The 2026 systematic review of ketogenic diets and lipids states it plainly: keto “significantly improves triglycerides and HDL-C but also leads to modest increases in LDL-C.” ³⁴

So, concretely, if your LDL is high:

1. Cut added sugar and refined carbs. Watch triglycerides and HDL fall. This is the single highest-value diet change for most people.
2. Swap saturated fat for unsaturated fat where you can. Watch LDL fall.
3. Move, at least 2.5 hours a week. Watch HDL and triglycerides improve.
4. Don’t smoke, protect your sleep (Observation 2), and get your CRP, ApoB, and — once — your Lp(a) measured, because those tell you about the road and the fleet composition the standard panel ignores.
5. If you’re lean and your LDL rocketed on keto — this is the important part — do not assume you’re fine because your triglycerides are 60. That’s the exact assumption the retracted study was built on. And do not assume you’re doomed either — the replacement analysis shows far less plaque progression than the withdrawn numbers. You’re in the “we don’t know” zone, and in that zone, the highway patrol answer is the right one: get an actual look at the road somewhere down the line (a calcium score or CT angiogram, discussed with your doctor), measure ApoB and Lp(a), and manage the fire. Not panic. Not denial. Information.

Statins — the honest ledger

Your doctor probably mentioned statins, so we’re going to do this the way I do it with patients: the real effects, the real side effects, what the research actually shows, and why patients are scared.

The benefit. The single best summary of the statin evidence is the Cholesterol Treatment Trialists’ Collaboration — a pooling of the individual data of 174,000 patients in 27 randomized trials. Its conclusion: for every 1 mmol/L reduction in LDL (about 38 mg/dL), the risk of a major vascular event falls by roughly a fifth. ²² This is real, measured, and one of the more solid findings in all of medicine.

But “roughly a fifth” is a relative reduction. What it means in your life depends entirely on how big your risk was to begin with. In JUPITER, the statin group had 1.36 events per 100 people per year on placebo vs 0.77 on the drug: you’d need to treat about 95 people for 2 years — or 31 for 4 years — to prevent one event. The man who has already had a heart attack is playing a different game: his baseline risk is multiple times higher, and the same twenty-percent produces enormous absolute benefit. That is why I tell patients: statins are a great insurance policy for a burning house and a so-so one for a dry house. And that aligns exactly with your observation — if you take the same badly inflamed patients eating the same terrible diet and put some of them on a lipid drug, the drugged group will be statistically better off. And it won’t be thrilling, because for the inflamed walkers the fire is still burning. Statistically better, not transformed. That’s residual inflammatory risk, and it’s now one of the most documented phenomena in cardiology.

The muscle-pain story — the part the internet gets wrong. You have read that statins cause muscle pain. Let me show you the best evidence, because it is not what the internet says. Researchers ran a clever kind of trial: each patient took — in random order, without knowing which was which — a statin some months, an identical dummy pill other months, and nothing other months, and rated their symptoms daily.

• In the SAMSON trial, sixty people who had already given up on statins because of side effects took twelve bottles over a year. Four held a statin, four held a dummy pill, four were empty. They scored their symptoms every day on a phone. Symptom scores came out at 16.3 on the statin and 15.4 on the dummy — no meaningful difference between them — while months with no tablet at all scored 8.0. About 90 percent of the entire symptom burden was reproduced by the dummy pill. ¹⁸
• And here is the finding that should change how you think about your own experience. Stopping the tablets brought fast relief — scores more than halved within three days on 55 percent of occasions. That happened just as often when the tablet was a dummy as when it was a statin. Neither how hard the symptoms hit on starting, nor how quickly they lifted on stopping, told you which pill it had been. ¹⁸
• Think about what that means. The test that you and I would both instinctively use — stop it, feel better, start it, feel worse, therefore the drug did it — produces exactly the same result when the pill contains nothing. People stopped a statin month early 22 percent of the time and a dummy month early 17 percent of the time, and that gap was not statistically meaningful. The informal experiment that feels like proof is not proof. It is the one thing in this whole article I would most want a patient to understand.
• In the StatinWISE trial — 200 patients who had already stopped statins because of muscle symptoms — symptom scores were identical on drug and placebo; mean difference −0.11 on a scale of 10. ¹⁹

Read that carefully, because it is easy to hear it as an insult and it is not one. The pain is real. Those people genuinely hurt. What the trials show is that most of the pain was not coming from the drug. That is not weakness. It is how powerfully expectation shapes physical sensation in all of us. And here is the part that tells you these patients were honest: after seeing their own data — half of SAMSON’s participants, and two-thirds of StatinWISE’s completers — went back on a statin. Real statin intolerance exists. It is simply much rarer than you’ve been led to believe. One honest footnote. The rare severe forms are real and serious. Muscle breakdown — rhabdomyolysis — runs at roughly one in ten thousand per year. In JUPITER, serious muscle problems occurred in one patient in a thousand, and the rate was exactly the same in the group taking the dummy pill. ¹⁵ Rare is not never, and a real reaction deserves a real conversation with your doctor. But the number is far smaller than the internet suggests.

The side effect that is real, and that almost nobody talks about: type 2 diabetes.

Statins do raise the chance of being diagnosed with diabetes. The best evidence we have is a pooling of the individual records of nearly 124,000 people from nineteen blinded trials. Standard doses raised new diabetes diagnoses by about 10 percent. High doses raised them by 36 percent, and the effect climbs with the dose. ²¹

That is real. It is not a rumour and it is not a nocebo effect. And when I first read those numbers I thought I had found my strongest argument, because here is a medicine that nudges your metabolism in the wrong direction in order to improve a number on a blood test. For a man whose whole case is that metabolic health sits upstream of everything, that felt like a gift.

Then I read the rest of the paper, and I have to give most of it back.

First, the actual change is tiny. Average blood sugar rose by four hundredths of one millimole. Average HbA1c, the three-month sugar average, rose by six hundredths of one percent on standard doses and eight hundredths on high doses. Those are not numbers you would feel. They are barely numbers your laboratory would notice.

Second, the 36 percent is partly an artefact of looking harder. In the high-dose trials, seven out of ten participants had their HbA1c checked during follow-up. In the standard-dose trials, three out of a hundred did. The authors say plainly that what drove the difference between those two figures was the amount of testing, not the drug. Measure more people, find more diabetes. Some of that 36 percent is a magnifying glass, not a disease.

Third, it is happening almost entirely to people who were already at the line. Around 62 percent of the extra diagnoses came from the quarter of participants whose sugar was already closest to the diagnostic threshold. The statin did not send them on a journey. It nudged them over a line they were already standing on.

Fourth, and this is the one I cannot get around. Those same trials measured heart attacks and strokes. Whatever harm the small sugar rise causes is already sitting inside that result — it happened to those same people during those same years, and the benefit was measured after it. So the diabetes effect is not a hidden cost waiting to be added on top. It has already been paid, and the trials still came out in favour of the drug. The authors state it directly: the benefits greatly outweigh the excess diabetes risk.

And JUPITER shows you exactly who this happens to, in one comparison. Among participants who already had at least one risk factor for diabetes, the statin raised new diabetes by 28 percent — and cut their heart attacks and strokes by 39 percent. Among participants with no diabetes risk factors, the statin produced no increase in diabetes at all, and cut their events by 52 percent.

Read that twice. In the people who were metabolically fine, the diabetes effect did not appear. In the people who were already vulnerable, it did — and even in them, the heart protection was three times larger than the diabetes cost.

So where does that leave my argument? Smaller than I wanted, and still not nothing. A drug that pushes glucose the wrong way, however slightly, in a person whose problem is metabolic, is worth knowing about and worth discussing with your doctor — especially if your sugar is already near the line, because that is exactly who this happens to. But I am not going to tell you that statins will give you diabetes. The evidence says they will move your sugar a little, that this will tip a few people who were nearly there anyway, and that the trade still favours the drug in people who need it.

That is the third time in writing this article that I have gone looking for support and come back with less than I set out with. I am leaving all three in.

The dementia question — because you asked, and because the internet loves this one. Is dementia one of the statin side effects? Here is the honest, current answer: the fear is understandable; the evidence does not support it. Randomized trials — the strongest evidence — have found no excess of cognitive decline or dementia on statins; a meta-analysis pooling 25 randomized trials specifically looking at cognition found no adverse cognitive effect. ³² The largest analysis ever conducted — 55 studies, over 7 million patients, published in 2025 — found statin use associated with a lower risk of dementia — observational, so it can’t prove cause and effect, but the direction is categorically not “statins cause dementia”. ³² Where does the fear come from? Case reports of memory fuzziness, an FDA label note based on those reports, and media amplification. The honest summary I give patients: ”If retaining your memory is the goal, the cardiovascular protection statins provide in people who need them may actually protect the brain’s blood supply. There is no good evidence they cause dementia. There is some evidence they might protect against it. The memory-fuzz anecdote is a real thing in a tiny number of people and is usually reversible on stopping the drug — worth mentioning to your doctor, not worth refusing the drug on.”

Why patients fear statins, then — the full answer:

• The echo of real side effects in the news — diabetes, rare myopathy — amplified.
• The nocebo effect — expectation produces symptoms in about 9 of 10 people who experience them.
• The FDA memory warning and case reports.
• The collapse of trust in pharma.
• And, honestly, the fact that a drug that makes you “feel” nothing while preventing a distant event is psychologically weak — dread of the possible side effect beats gratitude for the event that didn’t happen, every time.

None of that is irrational. All of it is human. What I try to give patients instead is the number: your risk, my risk assessment, the absolute benefit, the real side-effect odds, and the plan for the fire. Most patients, given the actual numbers instead of the panic, make the right choice for them.

Familial hypercholesterolemia — the strongest counterargument, dealt with honestly

Let’s steel-man the other side, because this is where they beat me.

My argument says: the problem is the inflamed road, not the vans — high LDL is dangerous mainly when it meets inflammation. Their answer: familial hypercholesterolemia. People born with genes that keep LDL high for life — no bad diet, no years of abuse. Severe cases get heart attacks as children. A 12-year-old has not spent three decades eating badly; there is no time to have wrecked the road; the vans alone did it.

My first instinct was to wave this away: “those patients still eat a modern diet, they still carry inflammation.” That’s fair for the adults. It is much harder for a child. And the Mendelian randomization studies — nature’s lifelong experiments, where genes decide your LDL before your diet does — show dose-for-dose, lower lifelong LDL means less heart disease. ²⁴

Now here’s my actual rejoinder, and I think it’s the honest one. You demanded of me: ”show me an FH patient on strict keto, very low inflammation, exercising, who still developed blockages — then you’d have something to argue against.” OK. You’re right that nobody has that study — it doesn’t exist, and I’ll concede the point costs me. But notice what the same challenge does to the other side: show me* an FH patient cured of blockages by fixing only LDL while the fire burned on — that study doesn’t exist either, and actually we know the answer in advance: it’s false. Half of statin-treated patients still have a burning road (hs-CRP ≥ 2) and keep having events. LDL-fixing alone has never been shown to fully erase risk — it cuts it. So the mature position is: the vans clearly matter (FH says so). The road clearly matters (CANTOS/colchicine say so). Neither is sufficient, neither is optional. Familial hypercholesterolemia is the best evidence LDL alone causes disease; CANTOS is the best evidence inflammation alone causes disease. Both sides of this fight get to keep their founding document. My side just had to give one back — the retracted one.

Where I actually land, ranked by how sure I am

You asked for honesty, not certainty. Here is my certainty ledger.

Established. LDL — the van count — is causally involved in artery disease. The genetics and the drug trials agree. This is one of the better-supported findings in medicine. My side loses this one, and I’ve said so in writing.

Established. Inflammation is separately causally involved. Quieting it reduces heart attacks even when cholesterol does not move at all — the CANTOS, COLCOT, and LoDoCo2 trials. ¹⁰¹³¹⁴

Established. Cutting carbohydrates reliably lowers triglycerides and raises HDL in nearly everyone.

Established. In lean people, cutting carbohydrates often raises LDL substantially; in heavier people it usually does not (the 41-trial meta).

Plausible (strong candidate, not proven). Triglycerides largely determine whether your vans are large and fluffy or small and dense — small, dense vans slip into cracks more easily; the triglyceride/HDL ratio tracks insulin resistance. Cutting carbs makes the vans bigger and fewer of them small. So it is possible to improve the quality of your cholesterol while the headline number gets worse. Plausible — and precisely the mechanism the whole “keto panel is fine” case rests on.

Unknown. Whether an excellent road report (great triglycerides, high HDL, low CRP) actually cancels a high van count. That is exactly the question the retracted study was built to answer. We are back to not knowing. And whether a lifestyle “super anti-inflammatory state” protects the road the way the anti-inflammatory drugs proved they can — plausible mechanism, zero randomized testing, nobody has done it.

What I will not say. I will not tell you that high LDL on a ketogenic diet is safe. I believed something close to that a year ago. The paper that supported it no longer exists. I will also not tell you that high LDL guarantees a heart attack — the replacement data shows that would be wrong too. What I will tell you is: the road and the traffic both matter; your doctor is measuring only half the road; and the fire — measurable, unfixable by statins, fixable by metabolic health — is the part everybody is ignoring.

When your doctor is right, and you should listen

I said at the start I would not give you permission to ignore your doctor, and I meant it.

Go and see your own doctor if any of this is true for you:

• You already have plaque on a scan, or a high coronary calcium score
• You have already had a heart attack or a stroke
• A parent or sibling had one early (before 55 men / 65 women)
• Your Lp(a) is high — the sticky-van gene
• You are diabetic — the single most powerful road-damager there is
• Your LDL stays very high (over ~190) regardless of anything you do
• You are on a prescribed cardiovascular medicine and thinking about stopping it

In every one of those situations, the doctor telling you to be careful is not being unimaginative. He is being right. And do not stop a prescribed medicine because of an article — not this one, not anybody’s. Take the article to the appointment instead. Ask better questions. That is what it is for. The thing I want for you is not defiance. It is understanding, so you can make a real decision with someone who has examined you — which I have not.

What I actually do

Not a prescription. Just what a physician who is his own patient does.

• Measure ApoB, not only LDL. Counting vans beats weighing cargo.
• Measure Lp(a) once. It’s largely genetic, it doesn’t change much, and it changes how seriously you take everything else.
• Track triglycerides and HDL together — the road report and the insulin-resistance ratio.
• Watch the fire: hs-CRP, and a ferritin and ESR when I want the full picture — because half this argument lives there and standard panels ignore it.
• If there’s any question about the road itself, image it — a calcium score or CT angiogram and look at the actual highway instead of arguing about the traffic report. This is the only way to settle the “is my keto LDL fine?” question for a particular person, and it is astonishingly underused.
• And publish my numbers even when I do not like them. Which brings me back to the thing this article has been circling. I am on day 78. I have published a blood ketone number every single one of those days, including the ones I was ashamed of. And I have not yet drawn the lipid panel that this entire article is about.

I am not going to dress that up. It is the gap. But I would rather show you the gap than close it quietly and hand you a finished story.

So the commitment is this, and it is specific: the panel gets drawn, and when it does, every number goes up — here, and in the newsletter, in full, in the same week I get them. LDL-C, ApoB, Lp(a), triglycerides, HDL, hs-CRP, ferritin. If they are good, that proves nothing on its own and I will say so. If they are bad, you will read that too, and it will be the most interesting thing I have published all year.

That is a follow-up piece with a date on it, not a promise that fades.

Where all this comes from

1. Soto-Mota A, et al. Increased LDL-cholesterol on a low-carbohydrate diet in adults with normal but not high body weight: a meta-analysis (41 RCTs). Am J Clin Nutr 2024. https://pubmed.ncbi.nlm.nih.gov/38237807/
2. Norwitz NG, et al. Elevated LDL cholesterol with a carbohydrate-restricted diet: evidence for a “lean mass hyper-responder” phenotype. Curr Dev Nutr 2021. https://pmc.ncbi.nlm.nih.gov/articles/PMC8796252/
3. McLaughlin T, et al. Is there a simple way to identify insulin-resistant individuals at increased risk of cardiovascular disease? (TG/HDL cutpoint 3.5). Am J Cardiol 2005. https://pubmed.ncbi.nlm.nih.gov/16054467/
4. Norwitz NG, Cromwell WC. Oreo Cookie Treatment Lowers LDL Cholesterol More Than High-Intensity Statin Therapy in a Lean Mass Hyper-Responder on a Ketogenic Diet. Metabolites 2024;14(1):73. https://www.mdpi.com/2218-1989/14/1/73
5. Soto-Mota A, Norwitz NG, Manubolu VS, Kinninger A, Wood TR, Earls J, Feldman D, Budoff M. Longitudinal Data from the KETO-CTA Study: Plaque Predicts Plaque, ApoB Does Not. JACC: Advances 2025;4(7):101686, July 2025 issue. — RETRACTED. Retraction notice doi:10.1016/j.jacadv.2026.102824. https://www.sciencedirect.com/science/article/pii/S2772963X26002450

The notice gives NO date and NO detail beyond “methodology.” It does not mention Cleerly, unblinded readings, or any conflict of interest. Every one of those claims must be sourced from refs 6 and 7, and attributed to them in the text.

1. Norwitz N. “We Want to Retract Our Own Paper” (the authors’ full account, incl. the unblinded Cleerly analysis and the blinded re-runs). StayCurious Metabolism, 2026. https://staycuriousmetabolism.substack.com/p/we-want-to-retract-our-own-paper
2. Retraction Watch. “Widely criticized keto diet study retracted,” May 22, 2026. https://retractionwatch.com/2026/05/22/widely-criticized-keto-diet-study-retracted/
3. López-Moreno M, López-Gil JF. Letter to the Editor: The KETO CTA Study (selective reporting; absence of a comparator group). JACC: Advances 2025;4(7):101861. https://pmc.ncbi.nlm.nih.gov/articles/PMC12163138/
4. Budoff MJ, et al. The Impact of Sustained LDL-C Elevation on Plaque Changes: Primary coronary plaque progression results from the Keto CTA Study. medRxiv preprint, Jan 16, 2026 — NOT peer-reviewed. https://www.medrxiv.org/content/10.64898/2026.01.15.26343955v1
5. Ridker PM, et al. Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease (CANTOS). N Engl J Med 2017;377:1119–31. https://pubmed.ncbi.nlm.nih.gov/28845751/
6. Sabatine MS, et al. Evolocumab and Clinical Outcomes in Patients with Cardiovascular Disease (FOURIER). N Engl J Med 2017;376:1713–22. https://pubmed.ncbi.nlm.nih.gov/28304224/
7. Schwartz GG, et al. Alirocumab and Cardiovascular Outcomes after Acute Coronary Syndrome (ODYSSEY OUTCOMES). N Engl J Med 2018;379:2097–107. https://pubmed.ncbi.nlm.nih.gov/30403574/
8. Tardif JC, et al. Efficacy and Safety of Low-Dose Colchicine after Myocardial Infarction (COLCOT). N Engl J Med 2019;381:2497–505. https://www.nejm.org/doi/full/10.1056/NEJMoa1912388
9. Nidorf SM, et al. Colchicine in Patients with Chronic Coronary Disease (LoDoCo2). N Engl J Med 2020;383:1838–47. https://pubmed.ncbi.nlm.nih.gov/32865380/
10. Ridker PM, et al. Rosuvastatin to Prevent Vascular Events in Men and Women with Elevated C-Reactive Protein (JUPITER). N Engl J Med 2008;359:2195–207. https://www.nejm.org/doi/full/10.1056/NEJMoa0807646
11. Ross R. Atherosclerosis — An Inflammatory Disease. N Engl J Med 1999;340:115–26. https://pubmed.ncbi.nlm.nih.gov/9887164/
12. Pradhan AD, et al. Residual Inflammatory Risk on Treatment with PCSK9 Inhibition and Statin Therapy. JACC 2019;73:2754–64. https://www.jacc.org/doi/10.1016/j.jacc.2019.02.056 ; and Ridker PM, et al. Inflammation and Cholesterol as Predictors of CV Events and Mortality (hsCRP > LDL on-treatment). Circulation 2024. https://www.ahajournals.org/doi/10.1161/CIRCULATIONAHA.123.066213
13. Wood FA, et al. N-of-1 Trial of a Statin, Placebo, or No Treatment to Assess Side Effects (SAMSON). N Engl J Med 2020;383:2182–4; full data: Howard JP, et al. JACC 2021;78(16):965–72. https://pmc.ncbi.nlm.nih.gov/articles/PMC8453640/
14. Herrett E, et al. Statin treatment and muscle symptoms: series of randomised, placebo controlled n-of-1 trials (StatinWISE). BMJ 2021;372:n135. https://pmc.ncbi.nlm.nih.gov/articles/PMC7903384/
15. Sattar N, et al. Statins and risk of incident diabetes: a collaborative meta-analysis of randomised statin trials. Lancet 2010;375:735–42. https://pubmed.ncbi.nlm.nih.gov/20167359/
16. CTT Collaboration (Reith C, et al.). Effects of statin therapy on diagnoses of new-onset diabetes and worsening glycaemia. Lancet Diab Endocrinol 2024. https://www.thelancet.com/journals/landia/article/PIIS2213-8587(24)00040-8/fulltext
17. CTT Collaboration (Fulcher J, et al.). Efficacy and safety of LDL-lowering therapy: meta-analysis of individual data from 174,000 participants in 27 randomised trials. Lancet 2015;385:1397–405.
18. Wiegman A, et al. Familial hypercholesterolaemia in children and adolescents (EAS consensus). Eur Heart J 2015;36:2425–37. https://pmc.ncbi.nlm.nih.gov/articles/PMC4576143/ ; World Heart Federation: https://world-heart-federation.org/heart-health/cholesterol/familial-hypercholesterolemia/ ; 13-year-old MI case: https://www.jacc.org/doi/10.1016/j.jaccas.2026.108242
19. Ference BA, et al. Effect of long-term exposure to lower LDL-C beginning early in life on the risk of CHD. JACC 2012;60:2631–9. https://pubmed.ncbi.nlm.nih.gov/23083789/ ; Holmes MV, et al. Eur Heart J 2015;36:539–50. https://academic.oup.com/eurheartj/article/36/9/539/506920
20. Irwin MR, et al. Sleep disturbance, sleep duration, and inflammation: a systematic review and meta-analysis. Biol Psychiatry 2016;80:40–52. https://pmc.ncbi.nlm.nih.gov/articles/PMC4666828/
21. Yaffe K, et al. The metabolic syndrome, inflammation, and risk of cognitive decline. JAMA 2004;292:2237–42. https://www.sciencedaily.com/releases/2004/11/041116220336.htm ; follow-up: Yaffe K, et al. Arch Neurol 2009. https://jamanetwork.com/journals/jamaneurology/fullarticle/1033005
22. Schaap LA, et al. Inflammatory markers and loss of muscle mass (sarcopenia) and strength (InCHIANTI/Health ABC). Am J Med 2006;119:526.e9–17. https://pubmed.ncbi.nlm.nih.gov/16750969/ ; Ding J, et al. IL-6 and sarcopenia meta-analysis. Ann Med 2024. https://www.tandfonline.com/doi/full/10.1080/07853890.2024.2384664
23. Te Morenga LA, et al. Dietary sugars and cardiometabolic risk (systematic review). Am J Clin Nutr 2014;100:65–79. https://pubmed.ncbi.nlm.nih.gov/24808490/ ; Schwingshackl L, et al. network meta-analysis. Am J Clin Nutr 2020. https://pubmed.ncbi.nlm.nih.gov/31711109/
24. Kodama S, et al. Effect of aerobic exercise training on serum levels of HDL-C. Arch Intern Med 2007;167:999–1008. https://pubmed.ncbi.nlm.nih.gov/17533202/
25. Santesso N, et al. Effects of higher- versus lower-protein diets on health outcomes (meta-analysis). Eur J Clin Nutr 2012. https://www.nature.com/articles/ejcn201237 ; Pasiakos SM, et al. Am J Clin Nutr 2015. https://pubmed.ncbi.nlm.nih.gov/25733478/
26. Statin and tendon: Deren ME, et al. Am J Sports Med 2016. https://pubmed.ncbi.nlm.nih.gov/27490216/ ; Eliasson P, et al. Sci Rep 2019. https://www.nature.com/articles/s41598-019-53238-7 ; Moniri NH, et al. Mayo Clin Proc 2018. https://www.mayoclinicproceedings.org/article/S0025-6196(18)30589-5/fulltext ; propensity-matched cohort (no increased rupture risk): https://bcdsp.org/the-risk-of-achilles-or-biceps-tendon-rupture-in-new-statin-users-a-propensity-score-matched-sequential-cohort-study/
27. Zhou Z, et al. Effect of statin therapy on cognitive decline and incident dementia: meta-analysis of randomized trials. JACC 2021. https://www.jacc.org/doi/10.1016/j.jacc.2021.04.075 ; Westphal Filho FL, et al. Largest statin-dementia meta-analysis (55 studies). 2025. https://pubmed.ncbi.nlm.nih.gov/39822593/
28. Lipotoxicity/glucotoxicity mechanisms. https://pmc.ncbi.nlm.nih.gov/articles/PMC4931958/ ; mitochondrial dysfunction and insulin resistance. https://pmc.ncbi.nlm.nih.gov/articles/PMC6510277/
29. Chang C, et al. The impact of the ketogenic diet on the lipid profile in adults: a systematic review. 2026. https://www.sciencedirect.com/science/article/abs/pii/S1530891X26000261
30. Schoeneck M, Kahan D. The effect of foods and nutrients on LDL cholesterol: a systematic review. Nutr Metab Cardiovasc Dis 2021. https://www.nmcd-journal.com/article/S0939-4753(21)00002-8/fulltext