Cholesterol After 40 on Keto: What the 2026 Guideline Says About LDL, ApoB, and Lp(a)

The sentence that stopped me was a short one, sitting in the middle of a table: measure Lp(a) at least once in every adult.

I eat keto. I do one meal on Fridays. I lift three to five days a week and I am fifty-four. That is a profile that gets a lot of attention from cholesterol arguments online, and for a long time I handled those arguments the way most people on this way of eating do: I read the half that agreed with me.

Then the American College of Cardiology and the American Heart Association, along with nine partner organizations, published a new dyslipidemia guideline in 2026. It replaces the 2018 cholesterol guideline that most clinicians have worked from for the better part of a decade. I did something I should have done years ago. I stopped reading what people said about it and read what it said.

It is a long document, and it does mention people like me. Twice, briefly, and not in the way either camp online tends to claim. I will get to both passages. More broadly, it changes what a lipid panel means for any man or woman past forty, and it changes it in ways that land squarely on people who eat the way I eat.

What follows is what the guideline recommends, what the best evidence says about keto and LDL, and the part the loudest voices on both sides skip: what that evidence cannot tell you. I read each source itself, not a summary, and left out any number I could not trace.

What Actually Changed in 2026

Start with the shape of the thing. The 2026 guideline was published in March in Circulation, with a joint publication in the Journal of the American College of Cardiology, chaired by Roger Blumenthal with Pamela Morris as vice chair. It retires the 2018 blood cholesterol guideline.

First, risk is estimated differently. For adults aged 30 to 79 with an LDL cholesterol between 70 and 189 mg/dL and no known heart disease, the guideline uses the PREVENT equations to estimate 10-year risk of atherosclerotic cardiovascular disease: under 3% is low, 3% to under 5% is borderline, 5% to under 10% is intermediate, and 10% or higher is high. It also estimates 30-year risk, which is the more useful horizon at forty-five than at sixty-five.

Second, numeric treatment goals are back. The 2026 guideline attaches LDL and non-HDL targets to the risk groups. For a borderline-risk adult who starts a statin, the goal is an LDL cholesterol under 100 mg/dL and a non-HDL cholesterol under 130. For high-risk adults it is under 70 and under 100. For people with established heart disease it is under 70, or under 55 at the very-high-risk end, with non-HDL under 100 and under 85.

The way I read the table, those goals describe what you and a clinician aim for once you have decided to treat. They are not a pass-or-fail grade on a blood draw.

Third, the guideline widens what counts as information. Coronary artery calcium scoring is given a larger role for men from forty and women from forty-five. Apolipoprotein B is recommended in specific situations. And lipoprotein(a), or Lp(a), is to be measured once in every adult.

One detail from the primary-prevention table is worth pausing on, because it applies to a group of people who eat a lot of fat. Even in an adult whose 10-year risk is low, an LDL cholesterol of 160 to 189 mg/dL, or a 30-year risk of 10% or more, is enough for the guideline to say that a moderate-intensity statin is reasonable. The guideline's stated reason is to reduce cumulative exposure to atherogenic lipoproteins. A low 10-year score does not make a high LDL a non-event.

Two Tests Most People Have Never Had

These are the two tests the 2026 guideline moves toward the middle of the conversation, and most people I know have had neither.

Lp(a). The guideline gives this its strongest recommendation, Class 1: measure it at least once in all adults to refine risk assessment. It does not require fasting. It also ties a result to risk in a way you can read. An Lp(a) of 125 nmol/L or higher, which is about 50 mg/dL, is treated as a risk-enhancing factor, associated with roughly a 1.4-fold higher risk of atherosclerotic cardiovascular disease. At 250 nmol/L, around the 95th percentile, the risk is about double. At 430 nmol/L, around the 99th, it is roughly four times, similar to heterozygous familial hypercholesterolemia. Those multiples come from UK Biobank data and are a general guide.

What the guideline recommends if yours is high is not a diet. It notes that Lp(a) concentrations stay similar over time and are minimally modified by lifestyle factors, which is why a single measurement is generally enough. For an Lp(a) of 125 nmol/L or higher it recommends early, optimal control of the modifiable risk factors, and its take-home messages add that a high Lp(a) should be an indication for more intensified LDL lowering. That matters for this audience, because one of the quiet assumptions in low-carb circles is that a good enough diet fixes the lipid panel. Whatever a diet does to LDL and triglycerides, the guideline's own text says it is not going to move Lp(a), and a high Lp(a) makes the LDL conversation more pressing, not less.

ApoB. LDL cholesterol measures the cholesterol carried inside certain particles. ApoB is a count of the particles themselves: the guideline describes one apoB molecule per LDL, VLDL and Lp(a) particle. Its position is more cautious than the internet's. In adults already on lipid-lowering therapy, especially people with heart disease, type 2 diabetes, cardiovascular-kidney-metabolic syndrome or high triglycerides, measuring apoB is reasonable to guide further treatment once LDL goals are met (Class 2a). The take-home messages say it is most useful when triglycerides are 150 mg/dL or higher, when there is diabetes, or when the achieved LDL is already under 70, because those are the settings where the standard panel can underestimate risk. I found no numeric apoB goal.

Why LDL Still Gets a Vote

Everything above assumes LDL matters. On keto forums that is sometimes treated as a belief held by people who have not read the right paper. Here is what the evidence looks like.

In 2017 a consensus panel of the European Atherosclerosis Society published a statement in the European Heart Journal asking whether LDL causes atherosclerotic cardiovascular disease. They drew on genetics, prospective cohorts, Mendelian randomization studies and randomized trials of LDL-lowering treatments. Meta-analyses covering more than 200 studies, more than 2 million participants and over 150,000 cardiovascular events showed a consistent, log-linear relationship between LDL exposure and risk. Rare mutations that raise LDL raise risk. Rare variants that lower it lower risk. And the effect appears to grow with the duration of exposure. The 2026 guideline says the same in its own words: duration of exposure to high LDL is an important driver of risk.

The panel concluded that LDL is a cause, not just a marker, and that a mechanism lowering LDL should lower risk in proportion to the absolute reduction and how long it is sustained, provided particle number falls with it and there are no harmful off-target effects. Hold on to that “provided.” It is where the entire keto argument lives. Time at a given level is part of the dose.

What Keto Does to LDL on Average

Now to the diet. The cleanest summary I could open was a 2020 meta-analysis in PLOS ONE by Dong and colleagues, pooling twelve randomized controlled trials with 1,640 participants. Participants were aged 31 to 65. The low-carbohydrate arms ate under 40% of calories from carbohydrate, the comparison arms 45% to 55%, and every trial followed people for at least three months.

Pooled across those trials, LDL cholesterol was higher on the low-carbohydrate diets by 0.11 mmol/L, with a 95% confidence interval of 0.02 to 0.19. That converts to about 4 mg/dL on average, with an interval of roughly 1 to 7 mg/dL. HDL cholesterol was higher by 0.10 mmol/L. Triglycerides were lower by 0.15 mmol/L. Systolic blood pressure was lower by about 1.4 mmHg.

Heterogeneity on LDL was high (I² of 71%), so individual trials disagreed about the size of the effect. A “low-carb” arm meant anything under 40% carbohydrate, a much wider net than strict ketosis. Most trials ran under two years, and the authors say plainly that effects on heart attacks, strokes and death are unknown.

An average is a fact about a group. Your LDL is a fact about you.

The Hyper-Responder Question

Here is where keto earns its own section. A subset of people, often lean and metabolically healthy, see their LDL cholesterol climb to very high levels on a ketogenic diet while their HDL rises and their triglycerides fall. The pattern has a name, the lean mass hyper-responder, and a group of researchers has been testing whether it carries the risk its LDL number would suggest.

Two studies are worth knowing.

The KETO trial (Budoff and colleagues, JACC: Advances, 2024) is an observational comparison, despite the name. It compared 80 people who had developed very high LDL on a ketogenic diet, defined as LDL of 190 mg/dL or more with HDL of 60 or more and triglycerides of 80 or less, without familial hypercholesterolemia, with 80 controls matched on age, sex, race, hypertension, hyperlipidemia and smoking from a separate population cohort. The study was funded by the Citizen Science Foundation. The keto group had been eating this way for a mean of 4.7 years. Mean LDL was 272 mg/dL in the keto group, with a maximum of 591, against 123 in the controls.

Coronary imaging found no significant difference in calcium scores between the two groups (P = 0.520), and no significant correlation between LDL cholesterol and plaque within the keto group. The authors' conclusion was that plaque in the keto group was not greater than in a matched cohort whose average LDL was 149 mg/dL lower.

The Keto-CTA study followed 100 people with the same lipid pattern for one year, scanning their coronary arteries at the start and end. As of this writing it is a preprint, posted to medRxiv in January 2026, which means it has not been peer reviewed. The mean LDL cholesterol was 242 mg/dL. The median increase in non-calcified plaque volume over the year was 5.6 mm³, a 37% relative increase. About 15% of participants showed regression despite sustained high LDL. And, the authors report, baseline plaque predicted progression while apoB and cumulative LDL exposure did not.

If you are a keto enthusiast looking for permission, you will read the first study and stop. If you are a skeptic looking for a gotcha, you will read the 37% and stop. I would ask you to read both and then read the next section.

What This Evidence Cannot Tell You

It cannot tell you that very high LDL is harmless on keto. The KETO trial is a cross-sectional comparison at one point in time. It found no difference in plaque so far. The mean exposure was 4.7 years. The consensus statement above says that the effect of LDL grows with how long you are exposed. Nobody has followed these people for the decades that kind of exposure implies.

It cannot tell you the comparison was clean. Eighty people per group is small, and a small study finding no significant difference is not the same as showing there is none. The groups differed in ways the authors acknowledge: BMI averaged 22.5 in the keto group and 25.8 in the controls, and HDL and triglycerides differed by design. Twenty-six of the 80 controls, about a third, were taking medication for hyperlipidemia, while none of the keto group were. LDL was a single measurement per person. And no particle-size or particle-count profile was reported.

It cannot tell you that very high LDL is dangerous on keto either. Neither study measured heart attacks, strokes or deaths. Imaging is a stand-in for those. The participants are also, almost certainly, a self-selected group: people whose LDL was extreme and who chose to be studied for it.

The Keto-CTA result cuts in two directions. Plaque grew measurably in a year in a group with no comparison arm, yet a sizable minority regressed, and neither LDL nor apoB predicted who progressed. The authors read that as LDL not driving plaque here. A study of 100 people all selected for high LDL, with no low-LDL comparison, is poorly placed to rule out a role for it, and as a preprint it deserves a provisional reading.

It cannot tell you about you. These are lean people with a specific lipid triad, and a wide mix of low-carb eaters averaged together. If your LDL on keto is 140 and you are carrying extra weight, neither group describes you well.

And it cannot tell you what Lp(a) is doing. None of the studies above set out to answer that. This is why I keep returning to the Lp(a) recommendation. It is the one number a good diet is least likely to be hiding.

Imaging is a stand-in for heart attacks. A good stand-in is still a stand-in.

What the Guideline Actually Says About Keto

Two passages speak to this directly, and I will give them as plainly as I can.

In the section on weight, the guideline says that dietary patterns lower in carbohydrate have been associated with greater weight and triglyceride reduction. It says ketogenic diets are associated with significant short-term weight and triglyceride reduction, but that LDL cholesterol can rise substantially in some individuals, particularly when there is high consumption of animal products. It also calls the approach restrictive and hard to sustain. Predominantly plant-based patterns, meaning Mediterranean, DASH and vegetarian or vegan, are the ones it says have the strongest evidence for lowering LDL and for improving cardiovascular outcomes.

In the section on severe hypercholesterolemia, an LDL of 190 mg/dL or higher, it describes ketogenic eating of about 20 to 50 grams of carbohydrate a day, notes that the carbohydrate is mostly replaced by fat, often saturated, and says some people may develop a severe hypercholesterolemia phenotype as a result. Everyone with that phenotype should get a careful dietary history to exclude a high-saturated-fat diet. For those whose cause is keto, it says to counsel them on ways to mitigate the LDL elevation while still achieving the primary goal of the dietary change.

The guideline does not tell anyone to abandon keto, and it does not wave the LDL away. It treats the diet as a plausible cause to examine and the LDL as something to reduce.

Elsewhere, the guideline says that healthy eating patterns that limit saturated fat while increasing unsaturated fat produce more consistent LDL lowering than restricting dietary cholesterol. It lists the sources of unsaturated fat: fish, nuts, seeds, olive oil, avocados.

So the collision is saturated fat, and the guideline names it. A keto plate built on butter, bacon and cheese and a keto plate built on olive oil, avocado, nuts and fish can both be ketogenic. Only one of them matches the line the guideline draws. That is an argument for noticing which fats do the work, not for or against keto.

I wrote about the fiber side of this in the piece on psyllium husk and again in the one on how much fiber you need after forty. It is the part of the guideline's advice that fits keto most easily and gets the least attention.

The Part That Isn't Food

In its section on physical activity, the guideline points to moderate-to-vigorous aerobic activity supplemented with resistance exercise on at least two days a week, and says this lines up directly with the federal recommendations.

I lift. That box is checked. The other one makes me do some honest accounting. I do low-intensity cardio most days, and low-intensity does not automatically count as moderate-to-vigorous. If I want to claim that part of the guideline as met, I have to measure it, not assume it.

That is the lesson I keep relearning in my other job. In platform engineering, the failure modes you never page on are the ones that take you down. We do not accept a service with no alert on its most important failure, yet most of us carry a body for decades with no alert on its biggest long-term risk. A lipid panel is a dashboard. Lp(a) and apoB are two panels most people never added.