A Keto Cholesterol Study Was Retracted in 2026: What the Evidence Shows Now
What Happened
The KETO-CTA study was one of the most widely shared pieces of research in the low-carb world, and in 2026 it was retracted from the journal that published it.
That sentence needs unpacking, because "retracted" gets used loosely online and the details decide what the event actually means. Here is the timeline, with sources.
- 2025: KETO-CTA publishes in JACC: Advances, reporting on 100 people following a ketogenic diet who had markedly elevated LDL cholesterol alongside otherwise favorable markers. The headline finding was that ApoB and LDL-C exposure were not associated with coronary plaque progression over one year.
- 2026: JACC: Advances issues a retraction notice. The stated reason was methodological concerns that could not be resolved by a correction notice.
- January 16, 2026: The same investigators post a new preprint on medRxiv, reanalyzing the same cohort using the originally pre-specified endpoint (Budoff MJ, Kinninger A, Manubolu VS, Norwitz NG, Feldman D, Soto-Mota A; medRxiv preprint, DOI 10.64898/2026.01.15.26343955).
The British Dietetic Association summarized the sequence in its cardiovascular nutrition review for 2026, and its one-line verdict is the most careful summary we have seen: the reanalysis reports plaque progression that looks less severe than the retracted paper's headline implied — but it is a preprint, not yet peer-reviewed, from a group that already has one retraction on this exact dataset.
Both of those clauses matter, and neither cancels the other. A study can be retracted for procedural reasons without its underlying question being answered in either direction. That is the situation here, and it is worth understanding before you repeat a headline in either camp.
Why This Study Carried So Much Weight
To see why a 100-person observational study generated years of argument, you have to know what it was trying to test.
Some people who eat very low-carbohydrate diets see their LDL cholesterol rise sharply — sometimes to levels that would normally trigger urgent clinical attention — while their triglycerides, HDL cholesterol, blood sugar and body composition all look favorable. The low-carb community named this pattern the lean mass hyper-responder, or LMHR, phenotype: lean, physically active people on ketogenic diets with very high LDL-C, high HDL-C and low triglycerides. We looked at one line of research into why the rise happens in the first place in our read of the Stanford genetics study.
The unresolved question is what that combination means over time. The conventional view, supported by decades of evidence linking cumulative ApoB exposure to atherosclerosis, is that very high LDL is a risk regardless of how good the rest of the panel looks. The LMHR hypothesis proposes that in this specific phenotype, the elevated LDL does not behave the way it does in a typical dyslipidemic patient.
KETO-CTA was the direct test of that hypothesis. If high LDL in lean, metabolically healthy dieters did not drive plaque formation over a year, that would be evidence — not proof, but real evidence — that the phenotype behaves differently.
That framing is why the publication mattered, and it is also why the retraction carries weight. A study designed to answer a contested question, funded and promoted by one side of the argument, needs to survive scrutiny more cleanly than most. This one did not.
What the Retraction Notice Says
The journal's notice cites methodological concerns that could not be fixed with a correction. That is the formal language, and it is deliberately narrow.
A retraction is a statement about a paper's reliability, not a finding about its topic. When a journal retracts a study, it is saying the published record can no longer be cited as evidence. It is not saying the opposite conclusion is now true. Readers routinely collapse those two things, and the collapse runs in both directions — critics treat a retraction as a verdict against the diet, while supporters treat the reanalysis as vindication. Neither reading survives contact with what was actually published.
There is one detail from the sequence that deserves attention on its own, because it is a general lesson rather than a point about any one diet. The original paper's primary endpoint and the endpoint it led with were not the same measure. Pre-registration exists to prevent exactly that gap: you declare what you will measure and how, before you see the data, so the analysis cannot quietly migrate toward the most favorable number. When a paper's emphasis and its pre-specified primary endpoint diverge, that is a flag worth noticing whether or not you care about the diet being studied.
That is a claim about how the paper was reported, and it sits alongside the imaging concerns rather than replacing them. Both are reasons the record was pulled.
What the January 2026 Reanalysis Found
The preprint posted in January 2026 re-ran the same cohort's imaging with a different analytical method and reported on the originally pre-specified primary endpoint.
Per the British Dietetic Association's summary, the result showed plaque progression that was less severe than the retracted paper's headline had implied. That is a meaningful correction of the original picture — the alarming version of the story does not hold up on the reanalysis.
It is also not the same as showing that LDL is harmless. Three limits are worth stating plainly, because the preprint's authors state them too:
- It is a preprint. It has not been through peer review. Anyone who has watched this dataset change twice in two years has a reason to wait for the reviewed version.
- There is no control group. A single-arm observational cohort can show that plaque changed while people ate a certain way. It cannot show that the diet caused the change, or that the change would have been different under another diet.
- The observation window is one year. Atherosclerosis accumulates over decades. A twelve-month window in a cohort that spent most of its life at normal LDL levels is a short exposure period for the question being asked.
The British Dietetic Association's conclusion on the whole affair is worth quoting in substance: this data is informative but not definitive, and the causal weight of evidence for LDL-C and ApoB has not changed.
That is the honest state of play. The specific study that was being cited as proof that LMHRs are protected is no longer citable. The general question remains open, and open is not the same as answered.
Where the Evidence Stands — and What Good Evidence Looks Like
The contrast that makes this story useful is sitting right next to it in the same year's research.
In August 2026, a randomized controlled trial from Washington University School of Medicine published in Cell Metabolism put 55 adults with obesity, prediabetes and excess liver fat on one of three diets — ketogenic, Mediterranean, or very-low-fat plant-forward — with all food provided and calories individually adjusted so that every group lost roughly the same 10 percent of body weight. Our walkthrough of that trial covers the numbers in full.
The structural differences between the two studies are the actual lesson:
| Feature | KETO-CTA (2025, retracted 2026) | WashU liver fat trial (2026) |
|---|---|---|
| Design | Single-arm observational | Randomized, controlled, three arms |
| Participants | 100, self-selected keto cohort | 55 enrolled, all food provided |
| Comparison group | None | Mediterranean and plant-forward arms |
| Key variable controlled | None | Weight loss matched across arms |
| Primary endpoint | Pre-specified, but not the one emphasized | Pre-specified and reported |
| Peer review status | Retracted; reanalysis is a preprint | Peer-reviewed |
Sources: JACC: Advances retraction notice (2026) and medRxiv preprint DOI 10.64898/2026.01.15.26343955; Cell Metabolism, August 2026, Washington University School of Medicine. Table compiled by NetCarbGuide from the cited sources.
A randomized trial with food provided and weight loss matched across arms cannot prove that a diet is safe for everyone over a lifetime either — it ran about five months. Its limits are real and its authors say so. But it can do something an observational cohort cannot: it can hold a variable constant and see what changes. That is why the liver fat trial will still be citable next year and KETO-CTA will not. Our TAKEOFF trial walkthrough runs the same source-checking exercise on a study whose results have not been reported yet.
One useful detail from the controlled trial, since the retracted study was about lipids: despite the ketogenic arm consuming more saturated fat, the trial reported no relative worsening in LDL cholesterol, apoB or 24-hour triglycerides against the comparison diets. That is one trial, in a specific population, over five months. It is not a license for anything. It is an example of the kind of evidence that moves a question forward — and notice that it moved it on liver fat and glycemic control, not on the LMHR claim, which it was never designed to test.
How to Read a Retraction Without Taking a Side
This story will keep circulating, so here is the checklist we apply to our own sourcing. None of it is specific to keto.
- Separate the retraction from the question. A pulled paper removes a citation. It does not install a conclusion.
- Check what was pre-registered versus what was emphasized. When a paper leads with a different measure than its declared primary endpoint, the reporting deserves scrutiny regardless of the result.
- Ask whether the study could have detected the effect at all. A one-year window cannot measure a decades-long process, and a single-arm cohort cannot separate the diet from everything else its participants do.
- Watch for funding and affiliation structures. Research on contested questions is often funded by people with a stake in the answer. That does not invalidate it — it raises the bar it has to clear. Both supporters and critics of low-carb diets fund and promote studies, and the standard should be the same for both.
- Prefer the upstream source. If you cannot reach the journal notice or the preprint itself, treat the summary as a claim about someone's summary. That includes this page: every factual statement above traces to a journal notice, a preprint with a DOI, or a named professional body, and none of them is a substitute for the original.
What This Does Not Prove
Two things get over-read every time this story resurfaces, and both are worth refusing.
It does not prove that ketogenic diets damage arteries. The study that was being cited for that claim was pulled, and the reanalysis does not support its alarming version.
It also does not prove that high LDL on a ketogenic diet is safe. The evidence for a causal role of cumulative ApoB exposure in atherosclerosis is old, broad and independent of this dataset, and nothing published in the last year has displaced it. The British Dietetic Association's reviewers, who have no stake in the low-carb argument, put it flatly: the causal weight of evidence has not changed.
What is genuinely new is narrower and more useful. A contested claim lost its flagship citation. The question is back to being open. Anyone who tells you this dataset settled it — in either direction — is telling you about their position rather than about the evidence.
This page summarizes published research and public statements, with sources named throughout. It is not medical advice. Cholesterol levels and cardiovascular risk are individual matters; if you are on a ketogenic diet and your LDL cholesterol has risen, that is a conversation to have with your own clinician, who can see your full history and your other risk factors.
Was the KETO-CTA study retracted?
Yes. JACC: Advances issued a retraction notice in 2026, citing methodological concerns that could not be resolved by a correction. The study had reported that ApoB and LDL-C exposure were not associated with coronary plaque progression over one year in 100 people following a ketogenic diet with high LDL cholesterol. A reanalysis using the originally pre-specified endpoint was posted as a medRxiv preprint on January 16, 2026. A retraction removes the paper as a citable source; it does not establish the opposite conclusion.
What is a lean mass hyper-responder (LMHR)?
LMHR describes a pattern seen in some lean, physically active people on ketogenic diets: markedly elevated LDL cholesterol alongside high HDL cholesterol, low triglycerides and otherwise favorable metabolic markers. The term comes from the low-carb community rather than from a formal clinical definition, and the British Dietetic Association notes there is no absolute definition of the phenotype. Whether the pattern carries the same long-term cardiovascular risk as elevated LDL in other people is the open question KETO-CTA was designed to address.
Does keto raise LDL cholesterol?
In some people, yes — sometimes substantially. That is why the LMHR question exists at all. What is not established is whether a large LDL rise in an otherwise metabolically healthy, lean person carries the same risk as the same LDL level in someone with a conventional risk profile. A 2025 network meta-analysis summarized by the British Dietetic Association found ketogenic diets performed better than Mediterranean-style patterns on short-term weight and triglyceride outcomes but worse on LDL and total cholesterol, with long-term cardiovascular outcomes still uncertain. Individual response varies, which is why this is a conversation to have with a clinician rather than a verdict to take from a headline.
Why do retractions matter for diet research?
Because the same study gets cited for years, and most people encounter it through a headline or a social post rather than the paper. When a study is retracted, the citations usually keep circulating anyway. Knowing the publication status of a study — peer-reviewed, preprint, expression of concern, retracted — changes how much weight it can carry. For this cohort, the publication status has changed twice in two years, which is a good reason to treat any current summary of it, including this one, as provisional.