Sunday Memos

Subscribe

The keto CTA study (KETO-CTA): what it found about LDL cholesterol, ApoB and coronary plaque

The KETO-CTA study scanned 100 people on a keto diet with very high LDL for plaque. Its 2025 paper was retracted, and this is what the study can and can't show.

22 minutes
Hand-drawn gold ink sketch

In 2015, a self-described engineer named Dave Feldman began a low-carb, high-fat diet, and his “lipid numbers spiked substantially”[10]. So he did what an engineer does with a number that won’t behave, and went looking for a pattern[10]. Some years later, his Citizen Science Foundation crowd-sourced the money for a study of people on a keto diet with very high LDL cholesterol, known as the KETO-CTA study[1, 11]. The question behind it is one plenty of low-carb eaters have put to their own lab results: if cutting carbs sends your LDL way up, is plaque quietly building in your arteries? (There are more questions like it on the Body page.)

The study followed 100 people, with two heart scans each, a year apart[1]. And if you’ve heard of it, you probably heard about the ending first: its 2025 paper was retracted in 2026 (notice published May 2026) “at the request of the authors and the Editors,” over concerns about its methods[3].

Now, “keto study retracted” sounds like a verdict on keto. And that makes sense, right? A paper says one thing, the paper gets pulled, so the opposite must be true. Well, it isn’t one. The notice says nothing about whether keto is safe or harmful[3]. And a one-group, one-year study couldn’t tell you whether diet-raised LDL causes plaque[12, 21]. And that’s what I want to argue here: both camps have been arguing over a study that, by its design (one group, one year, no comparison group), couldn’t settle their disagreement, with the retraction or without it.

So what was all the fuss about? Well, the short version is that the same 200 heart scans got run through three different computer programs, and two of them came back with nearly the same number while the third came back with a much bigger one. The long version starts with how the study was built, and I think it’s the better story.

Before that, the one thing I’d want a friend to hear. If you eat low-carb and your LDL cholesterol has climbed, this study can’t tell you what that means; that’s a conversation for your doctor, with your numbers on the table.

Help wanted, 100 people on a ketogenic diet with very high LDL-C: The keto CTA study design

The plan itself was simple enough: one group of 100 people on a ketogenic diet with very high LDL cholesterol, and each of them got a heart scan at the start and another one a year later[1]. So let’s start with the people. Feldman gave a name in 2017 to the people whose LDL rises most on low-carbohydrate diets, “lean mass hyper-responders”, and Nick Norwitz and colleagues later studied the pattern[13]. That’s a lot of syllables, so the papers shorten it to LMHRs. The LMHR phenotype (a phenotype is just a recognisable pattern) is very high LDL, high HDL and low triglycerides, a type of fat in the blood[13, 14] (more on that name in the lean mass hyper-responders post).

So where do you find 100 of them? Well, the study was run from the Lundquist Institute for Biomedical Innovation at Harbor-UCLA, with Matthew Budoff as principal investigator, and it recruited its 100 people through social media[1, 12]. The published plan says recruitment was initially through social media, mostly Twitter and Facebook, where one page devoted to LMHRs had over 10,000 members[18].

To get in, you had to tick every box on a checklist (the papers call it the inclusion criteria): at least two years on a ketogenic diet, an LDL cholesterol (the LDL-C on a lab report) of 190 mg/dL or more, a high-density lipoprotein cholesterol (HDL) of 60 mg/dL or more, triglycerides of 80 mg/dL or less, and no statins[1]. So, a pretty exclusive guest list. Each person had a CT scan of the coronary arteries (coronary computed tomography angiography, or CCTA, in the papers), a heart scan that shows plaque in the artery walls, once at the start and once a year later[1]. That scan is the “CTA” in the name. And 100 people with two scans each makes 200 scans, which is a number you’ll see again.

Now, you’ll often see it called a trial. But everyone was ALREADY on keto when they joined, and there was no comparison group[1, 12]. So nobody was told what to eat, and there was nobody eating differently to compare them with. On the registry it’s filed as an observational study, a prospective cohort study, which is a fancy way of saying one cohort: a single group, scanned twice. Hold on to that detail, because the whole argument later turns on it.

Then there’s the study plan, the paper where the researchers set down how the study was meant to be run (it’s dry reading, and it turns out to matter a lot). The plan was published in June 2024, and it set out three things. A program called QAngio would measure the plaque. The scans would be “blindly read.” And the main outcome, the number the whole study was built to report, would be the change in soft plaque, the kind that hasn’t hardened with calcium[12, 18, 19].

Remember all three, because every one of them comes back.

Why scan anyone’s arteries at all: What low-density lipoprotein cholesterol has to do with atherosclerosis and heart disease

Before the results, there’s a fair question: why would anyone scan the arteries of people with high LDL in the first place? The short answer is a word you’ll see a lot from here on, atherosclerosis. But first, the name. LDL is short for low-density lipoprotein, and the number on a lab report is, in full, low-density lipoprotein cholesterol (which nobody says out loud), and the papers write it as LDL-C[18]. In people like these, LDL-C and ApoB both run high, and the papers report both[1].

The worry with high LDL is atherosclerosis, the slow build-up of atherosclerotic plaque in the walls of the arteries. And when that happens in the arteries that feed the heart, it’s called coronary artery disease[20].

The big statement on this is a 2017 consensus from the European Atherosclerosis Society[24]. Its authors pulled together more than 200 studies covering over 2 million people, and they reported “a remarkably consistent dose-dependent log-linear association” between how much LDL-C the arteries are exposed to and the risk of atherosclerotic cardiovascular disease, an effect that “appears to increase with increasing duration of exposure to LDL-C”[24]. Which is a mouthful, so here it is in plain words: across those studies, more LDL-C for longer went along with higher cardiovascular disease risk. So whenever people argue about cholesterol and the risk of heart disease, this is one of the documents they end up pointing at.

And both camps in this story point at it. Norwitz, one of the authors, quotes its “dose-dependent” language in his reply post[23], and the paper itself says ApoB “does not drive atherosclerosis in a dose-dependent manner”[1]. The critics’ letter, from the other corner, says one short study “does not negate decades of evidence linking elevated LDL-C with cardiovascular risk”[21]. So the KETO-CTA study was one unusual group looked at against all of that, and that’s a big part of why people cared so much about how it came out.

Sounds like a test result, isn’t one: What “metabolically healthy” means in the keto CTA study papers

There’s a phrase the papers keep using for the people in them. The 2025 paper calls them “lean metabolically healthy people”, and the 2024 paper calls them “metabolically healthy individuals”[1, 4]. Which sounds great, a bit like a gold star on a report card. So what does it actually mean here?

Well, the papers never stop to define it. What the inclusion criteria actually asked for was a blood-fat pattern: high HDL and low triglycerides sitting right next to the very high LDL[1]. And on average these were lean individuals, with a mean body mass index of 22.5[1]. So “metabolically healthy” here describes who was studied. It isn’t something the scans found about their arteries.

And the 2025 paper, the one that was later retracted, actually said so itself: “However, this does not mean that the LMHR population is without risk”[1]. And in their published reply to the critics, the authors wrote that they did not want the results read “as if we believe all lean mass hyper-responders (LMHRs) to be of uniform risk”, and that “all patients should still seek personalized medical advice”[22].

Well, that’s not what I expected: The 2024 snapshot of coronary plaque, which was not retracted

The first results arrived in August 2024. Budoff and colleagues took 80 of the participants and matched them one to one with 80 people from the Miami Heart study, on things like age, gender and smoking[4]. Then they compared the first scans alone[4]. So on one side you have keto eaters whose LDL is very high, and on the other, matched people whose LDL is much lower. Who has more plaque?

I know what I’d have guessed. But the keto group’s LDL averaged 149 mg/dL higher, and their plaque was “not greater”[4]. Or, in the paper’s own summary, “no significant difference in coronary plaque burden”[4].

I found that surprising, and it’s easy to carry it too far. This was a snapshot, and one moment in time can’t show cause, or which way plaque is heading[4]. The keto group was also leaner, and only 35 of the 80 (so fewer than half) met the full lean mass hyper-responder cut-offs[4].

One thing to file away before we go on: this paper has never been retracted or flagged[3, 4]. It is a different paper from the one that came next.

Big title, one missing number: Longitudinal data from the KETO-CTA study, and whether ApoB could predict plaque

The one-year paper reported that neither LDL nor ApoB predicted whose plaque grew (plaque progression, in the researchers’ language); the plaque already there at the start (baseline plaque) did[1]. Or, as the paper put it (deep breath), “neither total exposure nor changes in baseline levels of ApoB and LDL-C were associated with changes in plaque”[1]. This is the paper that was later retracted.

It went online on April 7, 2025[1]. (ApoB, by the way, is apolipoprotein B, a protein that rises along with LDL in people like these[1].) And the original title didn’t mumble: “Plaque Begets Plaque, ApoB Does Not”[2]. The Lundquist Institute’s press release went further, saying keto-induced high cholesterol “does not predict heart disease”, and that the study “challenges the long-held belief that high cholesterol correlates and even directly causes coronary artery disease”[20]. That is a very big sentence to rest on 100 people and one year. The paper itself was more careful than its press release (press releases are rarely the shy ones), as you saw a minute ago.

Something was missing, though. Remember the plan’s third item: the change in soft plaque was the main outcome. In the paper it appeared in a figure but had no number in the text[6, 21]. So the main thing the study was set up to measure, the change in plaque volume, was the one number the first version of the paper didn’t write out in its text.

And people noticed. On May 7, 2025, two researchers, López-Moreno and López-Gil, wrote to the journal that this “raises concerns of selective reporting”[21]. The authors apologized (“This was a sincere oversight, not intentional selective reporting”) and gave the number: a median rise of 18.9 cubic millimeters of soft plaque (non-calcified plaque volume, in the papers) in a year[22]. It became the critics’ favorite number.

Which leaves two facts sitting awkwardly side by side. By the paper’s own measurement, soft plaque rose in a group with very high LDL. And the paper’s headline was that LDL didn’t predict whose plaque grew. Can both be true? Well, yes, they can, and the reason is the detail I asked you to hold on to.

How both can be true: Everyone’s LDL cholesterol was high, so the study couldn’t show whether high LDL matters

Brad Stanfield, a family doctor in New Zealand, pointed out that “there was no ‘low LDL’ group to contrast against”[6]. When everyone’s LDL is high, LDL can fail to sort fast plaque-growers from slow ones and still be the reason plaque grows at all[6].

Picture testing whether rain makes people wet by standing a hundred people in a downpour. Some end up wetter than others, thanks to coats and umbrellas, so how hard it rained on each person won’t explain much. But rain is still what’s soaking them. Here the rain is the LDL elevation everyone in the study had, and the coats are whatever else differs between people.

Now, one answer to that is that “high” still covered a lot of ground inside the group[1, 4]. The consensus on LDL describes a “dose-dependent” link[24], so you’d expect people at 400 to gain plaque faster than people at 200. In the authors’ data they didn’t[1]. The critics’ reply: a year is short, and the finding is only “narrowly true” within the group[6, 21].

What surprised me is where the two sides meet. The authors’ own reply says their results “are compatible with a causal role of apolipoprotein B in atherosclerosis”[22]. Norwitz’s picture is the bottom block in a Jenga tower: something can be “causal, and not predictive”[23]. That’s a long way from the press release.

Up to here, everyone was arguing about what the 18.9 meant. The next question was whether 18.9 was the right number in the first place.

One set of scans, three programs: Same scans, different software

The same 200 scans ended up going through three programs for assessing plaque volume, Cleerly, QAngio and HeartFlow, and the three of them did not give the same number[1, 8, 9]. To see how that happened, go back to the plan’s first item. It named QAngio. The 2025 paper had measured plaque with Cleerly, an “artificial intelligence-guided” program[1, 18]. And this is where the engineer comes back into the story. Feldman says he got the raw Cleerly data on April 18, 2025, “eleven days after publication,” and “found several concerning issues”[8].

Wait, eleven days AFTER? His name is on the paper. Well, three of the authors (Feldman, Norwitz and Adrian Soto-Mota) told Retraction Watch they didn’t have access to the data before the study was published[16]. So the first time Feldman could go through the raw numbers himself, the paper had already been out for a week and a half.

Then the plan’s second item, the blind reading. The authors say the Cleerly analysis wasn’t blinded: whoever measured wasn’t kept from knowing which scan was “before” and which “after,” the safeguard that stops expectations nudging the numbers[5, 16]. And one co-author was Cleerly’s chief medical officer; the journal says he disclosed his Cleerly affiliation when the paper was submitted and his equity in the company on acceptance, but the published paper listed only the affiliation[1, 16].

So the same 200 scans were read again. And honestly, I’d have assumed the plaque assessment was the boring part: a scan is a picture, you measure the plaque in the picture, and you get a number. But the measuring is done by software, and it turns out the software doesn’t have to agree. Cleerly had found a median rise of 18.9 cubic millimeters[1]. QAngio found 5.6, in a preprint six members of the research team posted on January 16, 2026[9]. HeartFlow, in a reading the authors describe as blinded, found 5.5[8]. (So the first reading came out more than three times as big as either re-read.)

Same people, same pictures, three programs, three answers.

Same people, same pictures, three programs, three answers.

Before leaning on the two lower ones, know that both are the authors’ own and not yet peer reviewed[8, 9]. If they hold up, the critics lose their favorite number. The authors lose something too: their ApoB headline now rests mainly on that preprint, which isn’t peer reviewed[3, 9]. (Its exact words: “neither ApoB levels nor cumulative LDL-C exposure predicted plaque progression in this population”[9].) And Stanfield says the reanalysis “raises new concerns rather than resolving the old ones”[16]. The Retraction Watch story that quotes him doesn’t spell out what those new concerns are, so I can’t tell you either.

Why was the keto CTA study’s 2025 paper retracted?

The journal’s notice gives one reason: concerns about the paper’s methodology and the reliability of its data[3]. Okay, but WHICH methods? What errors? Raised by who? Well, the notice doesn’t tell us. It names no specific error. The more detailed explanation, the one about Cleerly, comes from the authors, not the notice[3, 5, 8].

Figure 1From study plan to retraction
  1. Jun 2024Study plan published: QAngio, blinded reads [18]
  2. Aug 2024Snapshot paper: no more plaque than matched people with lower LDL (still stands) [4]
  3. Apr 7, 2025One-year paper online: “Plaque Begets Plaque, ApoB Does Not” [1]
  4. Apr 18, 2025Feldman gets the raw Cleerly data (by Feldman’s own account) [8]
  5. May 7, 2025Critics’ letter and authors’ reply; median rise given as 18.9 mm³ [21, 22]
  6. Jan 16, 2026Authors’ re-read posted as a preprint: 5.6 mm³ (not peer reviewed) [9]
  7. Jan 23, 2026Journal attaches an expression of concern [2]
  8. Mar 5, 2026Norwitz: “We Want to Retract Our Own Paper” [5]
  9. 2026The 2025 paper is retracted; the notice runs in the May 2026 issue [3]
Only the 2025 paper was retracted. The 2024 snapshot still stands. The April 18 entry is Feldman’s own account, not a published document.

Here’s how it unfolded. On January 23, 2026, the journal attached an expression of concern, which the authors say they pushed for[2, 16]. On March 5, Norwitz published a post with a title you don’t see every day: “We Want to Retract Our Own Paper”[5]. His line, which the authors still hold: the size of the plaque rise “has been unequivocally and demonstrably refuted,” but the finding that LDL and ApoB don’t predict plaque growth in this group “remains robust”[5].

The paper was retracted later in 2026, and the notice ran in the journal’s May 2026 issue[3]. Here is its whole explanation:

This article has been retracted at the request of the authors and the Editors. Following publication, concerns were raised regarding the methodology in this article, which effect [sic] the reliability of the data. The authors and the Editors agree that the identified errors are too great to be corrected with a corrigendum.[3]

That’s all of it. It doesn’t name Cleerly, blinding, a conflict of interest or a single number, doesn’t say fraud or misconduct, and says nothing about diet, LDL or hearts beyond repeating the paper’s title[3]. Stanfield called the retraction “warranted” and the re-framing of the same data, “in my view, more advocacy than science”[16].

So the whole official explanation is three sentences long, and people on both sides have been reading entire stories into it.

So the whole official explanation is three sentences long, and people on both sides have been reading entire stories into it.

What does it mean if you eat low-carb?

On its own, it tells you very little either way. The KETO-CTA study can’t say whether high LDL on a ketogenic diet is harmless or harmful.

Take the retracted paper away and look at what’s left. There’s the 2024 snapshot, which can’t show cause[4]. There’s the authors’ published reply to the critics, which reports the same Cleerly analysis[22]. And there’s an unreviewed re-read with the same one-group, one-year design[9, 12]. Read generously, KETO-CTA says that in 100 lean people on keto who all had high LDL, followed for a year, ApoB didn’t sort fast plaque-growers from slow ones, in the authors’ re-read[9].

What it can’t say is whether that high LDL is harmless. That needs people with lower LDL to compare against, and more than a year[6, 21].

Meanwhile, the older evidence is still there. A 2017 European Atherosclerosis Society consensus, drawing on more than 200 studies and over 2 million people, concluded that the evidence “unequivocally establishes that LDL causes ASCVD” (atherosclerotic cardiovascular disease)[24]. The weight of other evidence still says high LDL matters, and this study doesn’t change that[24].

It says even less about most people on keto. Only a minority of people who cut carbs see an elevation in LDL like this, and these participants were picked for it[1]. (The 2023 “keto doubles heart attack risk” headline was a different study[27]: the post on that study.)

Is it the KETO-CTA trial or the KETO-CTA study?

Both, honestly, because the papers themselves couldn’t pick one. The 2024 paper’s title ends with “The KETO Trial”[4]. The 2025 paper went online as “Plaque Begets Plaque, ApoB Does Not: Longitudinal Data From the KETO-CTA Trial”, and the version on PubMed now is titled “Longitudinal Data From the KETO-CTA Study: Plaque Predicts Plaque, ApoB Does Not”[1, 2]. So if you typed “KETO-CTA trial” into a search bar and landed here, you’re in the right place.

And the registry entry adds a third name, just to keep things interesting: “Diet-induced Elevations in LDL-C and Progression of Atherosclerosis (Keto-CTA)”, filed as observational[12]. When most of us hear “trial”, we picture people getting assigned to one thing or another and then compared. Nobody was assigned to anything here (everyone showed up already on keto), which is why I’ve called it a study the whole way through. If you’d like more food and health questions read the same way, they’re on the Body page.

If you’d like to read more on this:

The bit you can send to the group chat: Quick answers about the keto CTA study (KETO-CTA)

What was the KETO-CTA study designed to investigate?

The KETO-CTA study was designed to describe how plaque buildup in the coronary arteries changed over 12 months in people following a ketogenic diet whose LDL cholesterol had climbed very high[18]. So, a prospective study of 100 individuals with the LMHR phenotype or close to it, and nobody to compare them with[1, 12]. The main outcome was the change in noncalcified plaque volume between the two heart scans, which the plan said would be “blindly read” by cardiac computed tomography readers[1, 18] (and that part, as you saw, got complicated).

What were the main findings of the KETO-CTA study?

The KETO-CTA study’s main findings come in two papers, and only the second was retracted[3]. The 2024 paper reported that the amount of plaque in 80 participants was “not greater” than in 80 matched people whose LDL averaged 149 mg/dL lower[4]. The 2025 paper reported that neither the change in ApoB, nor baseline levels of ApoB, nor total LDL-C exposure was associated with the change in noncalcified plaque volume or in the total plaque score[1]. But baseline plaque was associated with the change in noncalcified plaque volume[1].

What did the study find about ketogenic diets, high LDL cholesterol and coronary plaque?

Within this one group, the retracted 2025 paper reported that LDL cholesterol and apolipoprotein B did not predict whose coronary plaque grew over a year[1, 22]. Plaque did grow, though, by a median of 18.9 cubic millimeters on the paper’s own reading[1, 22], and the predictor of plaque progression was the plaque already there[1]. The authors’ later re-read says 5.6, and that one isn’t peer reviewed[9]. And neither number can show whether high LDL causes plaque, because there was no low-LDL group to compare with[6].

What are the study’s limitations, and how strong are its conclusions?

The study’s conclusions are limited by its design: one group, one year and no comparison group[1, 21]. The critics’ published letter said the short follow-up “limits the ability to assess long-term cardiovascular impact”, and that the result “does not negate decades of evidence linking elevated LDL-C with cardiovascular risk”[21]. And the authors’ own reply agreed that their exploratory results, especially “those suggesting a null association between plaque progression and ApoB”, “should be interpreted with caution”[22]. On top of all that, the 2025 paper has since been retracted[3].

Was the whole KETO-CTA study retracted?

No, and this is the part I’d most want people to get right. Only the 2025 one-year paper was retracted in 2026 (notice published May 2026), “at the request of the authors and the Editors”[3]. The 2024 paper, the one that compared first scans with a matched group, has not been retracted[3, 4]. As for why, the notice points to concerns about the methodology and the reliability of the data, names no specific error, and says nothing at all about diet[3].

The numbers, if you want them

  • The 2024 snapshot: mean LDL 272 mg/dL against 123; median coronary artery calcium score 0 against 1, no significant difference; body mass index (BMI) 22.5 against 25.8[4].
  • Why 80 and not 100 in the 2024 snapshot: the pairs were matched for age, gender, race, hyperlipidemia, hypertension and smoking status, and the other 20 participants fell outside the Miami Heart study’s age range of 40 to 65[4] (so they were scanned like everyone else but left out of the 2024 comparison).
  • What the plan expected: to work out how many people it needed, the plan used “a conservative estimate of mean plaque volume change of 7 mm3 given the subjects are substantially younger with little to no risk factors”[18]. That was a planning assumption about a mean, not a forecast.
  • How high “high” was: the middle half of the 2025 group sat between 202 and 308; the highest among the 80 in the 2024 comparison was 591[1, 4].
  • The retracted paper’s Bayesian figure: data 6 to 10 times more supportive of “no association” than of an association, now withdrawn with the paper[1, 3].
Reading of the same 200 scansRise in soft plaque over one year
Cleerly, in the retracted paper18.9 mm³ median (mean 31.5; middle half 9.3 to 47.0), 43% above the start[1, 22]
QAngio, in the authors’ preprint (not peer reviewed)5.6 mm³ median, a 37% relative increase; regression in 15% of participants[9]
HeartFlow, on Feldman’s blog (not peer reviewed)5.5 mm³ (median or mean not stated)[8]

Each percentage is measured against that program’s own reading of the starting scans, so the two percentages can’t be compared with each other.

The next time a headline says “retracted”: Four questions worth asking

“Keto study retracted” sounds like a verdict on the diet. You’ve now seen how much room there is between those three words and what happened. The next time a headline like it turns up, four questions do most of the work:

  1. Who asked for the retraction? Here, the authors and the editors[3].
  2. What does the notice actually say? Methodology and the reliability of the data, nothing about diet[3].
  3. Which number was wrong, and who says so? The plaque reading, according to the authors, whose replacement numbers aren’t peer reviewed yet[5, 9].
  4. Could the study ever have answered the question it’s being used for? A one-group, one-year study couldn’t tell you whether diet-raised LDL causes plaque[12, 21].

So what I take from the whole saga is smaller than either camp’s headline. The 2025 paper has been retracted, and the question it was asking is still open, not answered in either direction.

If you eat low-carb, one thing for today: open a note on your phone and write down your last LDL result and roughly when you changed how you eat, so it’s ready for your next appointment. If that feels like a lot, write just the number. And if you don’t eat low-carb, keep question four for the next study you see shared as a verdict.

Sources

  1. 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 (online 7 Apr 2025; originally titled “Plaque Begets Plaque, ApoB Does Not: Longitudinal Data From the KETO-CTA Trial”). RETRACTED. https://doi.org/10.1016/j.jacadv.2025.101686 · https://pmc.ncbi.nlm.nih.gov/articles/PMC12418260/
  2. Expression of Concern: “Plaque Begets Plaque. ApoB Does Not: Longitudinal Data from The KETO-CTA Trial”. JACC: Advances 2026;5(3):102607 (online 23 Jan 2026). https://doi.org/10.1016/j.jacadv.2026.102607
  3. Retraction: Longitudinal Data from the KETO-CTA Study: Plaque Predicts Plaque, ApoB Does Not. JACC: Advances 2026;5(5):102824 (May 2026 issue). https://doi.org/10.1016/j.jacadv.2026.102824 · https://pubmed.ncbi.nlm.nih.gov/42206798/
  4. Budoff M, Manubolu VS, Kinninger A, et al. Carbohydrate Restriction-Induced Elevations in LDL-Cholesterol and Atherosclerosis: The KETO Trial. JACC: Advances 2024;3(8):101109. Not retracted. https://doi.org/10.1016/j.jacadv.2024.101109
  5. Norwitz N. “We Want to Retract Our Own Paper.” Stay Curious Metabolism (Substack), 5 Mar 2026. https://staycuriousmetabolism.substack.com/p/we-want-to-retract-our-own-paper
  6. Stanfield B. “Lean Mass Hyper-Responders See Rapid Plaque Progression.” drstanfield.com (2025). https://drstanfield.com/blogs/articles/lean-mass-hyper-responders-see-rapid-plaque-progression
  7. Leaf A. “Lean Mass Hyper-Responders Are Not Protected From Atherosclerosis.” alexleaf.com, 13 Apr 2025. https://alexleaf.com/lean-mass-hyper-responders-are-not-protected-from-atherosclerosis/
  8. Feldman D. “Keto-CTA Preliminary Data Update.” Cholesterol Code, 16 Sep 2025. https://cholesterolcode.com/keto-cta-update/
  9. Budoff M, Kinninger A, Manubolu V, Norwitz N, Feldman D, Soto-Mota A. The Impact of Sustained LDL-C Elevation on Plaque Changes: Primary Coronary plaque progression results from the Keto CTA Study. medRxiv, posted 16 Jan 2026. Not peer reviewed. https://doi.org/10.64898/2026.01.15.26343955
  10. Feldman D. “About.” Cholesterol Code. https://cholesterolcode.com/about/
  11. The Lundquist Institute, press release as reprinted by SciTechDaily, 14 Apr 2025. https://scitechdaily.com/?p=468532
  12. ClinicalTrials.gov NCT05733325, “Diet-induced Elevations in LDL-C and Progression of Atherosclerosis (Keto-CTA)”. https://clinicaltrials.gov/study/NCT05733325
  13. Norwitz NG, Feldman D, Soto-Mota A, Kalayjian T, Ludwig DS. Elevated LDL Cholesterol with a Carbohydrate-Restricted Diet: Evidence for a “Lean Mass Hyper-Responder” Phenotype. Curr Dev Nutr 2022;6(1):nzab144 (online 30 Nov 2021). https://doi.org/10.1093/cdn/nzab144
  14. National Heart, Lung, and Blood Institute. “High Blood Triglycerides.” https://www.nhlbi.nih.gov/health/high-blood-triglycerides
  15. Mindrum MR. Let’s Be Clear about Expected Cardiovascular Risk. Curr Dev Nutr 2022;6(5):nzac042. https://doi.org/10.1093/cdn/nzac042
  16. Orrall A. “Widely criticized keto diet study retracted.” Retraction Watch, 22 May 2026. https://retractionwatch.com/2026/05/22/widely-criticized-keto-diet-study-retracted/
  17. Norwitz NG, Mindrum MR, Giral P, et al. Elevated LDL-cholesterol levels among lean mass hyper-responders on low-carbohydrate ketogenic diets deserve urgent clinical attention and further research. J Clin Lipidol 2022;16(6):765-768. https://doi.org/10.1016/j.jacl.2022.10.010
  18. Javier DAR, Manubolu VS, Norwitz NG, et al. The impact of carbohydrate restriction-induced elevations in low-density lipoprotein cholesterol on progression of coronary atherosclerosis: the ketogenic diet trial study design. Coronary Artery Disease 2024;35(7):577-583. https://doi.org/10.1097/MCA.0000000000001395
  19. Alyami B, Santer M, Seetharam K, et al. Non-Calcified Coronary Artery Plaque on Coronary Computed Tomography Angiogram: Prevalence and Significance. Tomography 2023;9(5):1755-1771. https://doi.org/10.3390/tomography9050140 · https://pmc.ncbi.nlm.nih.gov/articles/PMC10514817/
  20. The Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center. “New research from The Lundquist Institute finds ketogenic diet-induced high cholesterol does not predict heart disease.” Press release, EurekAlert!, 7 Apr 2025. https://www.eurekalert.org/news-releases/1079651
  21. López-Moreno M, López-Gil JF. The KETO CTA Study [Letter]. JACC: Advances 2025;4(7):101861. https://doi.org/10.1016/j.jacadv.2025.101861
  22. Soto-Mota A, Norwitz NG, et al. Reply: The Keto CTA Study. JACC: Advances 2025;4(7):101862. https://doi.org/10.1016/j.jacadv.2025.101862
  23. Norwitz N. “KETO-CTA: Clarifications and Moving Forward.” Stay Curious Metabolism (Substack), 8 May 2025. https://staycuriousmetabolism.substack.com/p/keto-cta-clarifications-and-moving
  24. Ference BA, Ginsberg HN, Graham I, et al. Low-density lipoproteins cause atherosclerotic cardiovascular disease. 1. Evidence from genetic, epidemiologic, and clinical studies. Eur Heart J 2017;38(32):2459-2472. https://doi.org/10.1093/eurheartj/ehx144
  25. Citizen Science Foundation, “Keto-CTA” open data page (posted May 2025, updated June 2026). https://citizensciencefoundation.org/keto-cta/
  26. Oberoi S, Meinel FG, Schoepf UJ, et al. Reproducibility of noncalcified coronary artery plaque burden quantification from coronary CT angiography across different image analysis platforms. AJR 2014;202(1):W43-9. https://doi.org/10.2214/AJR.13.11225
  27. Sunday Memos, “The keto study that wasn’t about keto” (on the 2023 UK Biobank study by Iatan et al.). https://sundaymemos.com/keto-heart-attack-study/

Keep reading.

Surprise me ↻  ·  Every memo →

See you Sunday.

Want a short letter on Sunday mornings? It’s what I wrote that week, one small thing to try, and a question you can answer just by hitting reply. Free, and you can leave with one click.