Does keto really double your risk of a heart attack?

“Keto Diet Doubles the Risk of Heart Attack, Stroke: Study,” ran a WebMD headline on March 6, 2023, and within days it was everywhere.^[1] It was clean, it was scary, and it was about a diet millions of people were actually on, so of course it travelled. I wanted to know where a word like “doubles” comes from, so I opened the study underneath it.

What I found was stranger than the headline. The people in it weren’t eating keto. The doubling came from 305 of them, and their diet was measured with a single questionnaire about one day of eating. When the full paper came out a year later, its own extra checks shrank the effect or made it disappear. None of that means keto is good for your heart (the better trials carry their own warning, and we’ll get there). It means this study can’t tell you either way.

Where the scary number came from

The headlines trace back to a talk at the American College of Cardiology’s annual meeting in New Orleans. The ACC’s own press release, from March 5, was more careful than most of what followed.^[2] Its headline said a “keto-like” diet “may be linked” to higher risk, and it noted that the study was observational and could only show an association. (Observational means the researchers looked at what people already ate. Nobody was given a diet.)

The researchers, led by Iulia Iatan at the University of British Columbia, used the UK Biobank, a large health study that recruited about half a million British adults between 2006 and 2010. They picked out people whose diet met their definition of low-carb, high-fat. They matched each of them by age and sex to four people whose diet didn’t, and then followed both groups for an average of 11.8 years.

Over that time, 9.8% of the low-carb group had a major heart or blood-vessel event, against 4.3% of the comparison group. That covered a heart attack, a stroke caused by a blocked blood vessel, unstable angina (a dangerous kind of chest pain), or a blocked artery that needed a procedure. Once the researchers adjusted for diabetes, smoking, blood pressure and weight, it worked out to about 2.2 times the risk.

That’s where “doubles” came from. Every part of that sentence turns out to deserve a second look, starting with the little word almost everyone skipped: “like.”

The “keto” group wasn’t eating keto

A ketogenic diet is defined by what it does to you. It cuts carbohydrate low enough that your body starts burning fat for fuel and producing ketones, a fuel the body makes from fat when carbs run short. The American Diabetes Association’s consensus report describes the very-low-carb pattern like this: “Often has a goal of 20–50 g of nonfiber carbohydrate per day to induce nutritional ketosis.”^[4]

The study’s own authors, citing the National Lipid Association, put ketogenic diets at 50 g a day or under, or 10% of calories or less. The study itself drew a much looser line. The press release defined “keto-like” as 25% of calories or less from carbohydrate and more than 45% from fat. The peer-reviewed paper widened that to under 100 g a day “and/or” under 25%.^[3]

So what did the low-carb group actually eat? According to the paper’s own table, an average of 78.9 g of carbohydrate a day, about 23% of their calories. That’s somewhere between one and a half and four times the ketogenic range. And the researchers measured it directly. When they tested blood ketones in a subset of people, the levels were “below levels indicative of nutritional ketosis.” In plain terms, these people weren’t in ketosis.

Figure 1What counted as “keto” in the study

By the authors’ own definitions, this was a low-carb group, not a keto group. The editorial the journal published alongside the paper said the same thing: the definition “does not represent a therapeutic ketogenic diet.”^[5] The study’s central word meant one thing, and the headline used it to mean another.

One day of eating, twelve years of follow-up

The next thing I wanted to know was how anyone knew what these people ate. For the headline result, the answer is that they were asked once.

UK Biobank volunteers filled in an online questionnaire about everything they’d eaten and drunk in the previous 24 hours. The ACC press release calls it “a one-time self-reported 24-hour diet questionnaire.” The 305 people in the low-carb group were sorted on that single day, then followed for nearly twelve years.

Picture what that means. Someone who skipped the bread one Tuesday in 2009 counts the same as someone who’s been strict keto for a decade. Someone who quit low-carb in 2011 stays in the low-carb group until 2021. The journal’s editorial notes that three of these one-day recalls are recommended to capture how someone usually eats.

Some volunteers did fill in the questionnaire again later. The paper ran a check on people who’d done it at least twice, and that’s where the story turns.

A year later, the paper arrived and the number shrank

When the headlines ran, there was no paper to read. I went back through the coverage from four big outlets, and only Fortune told readers the study was “not yet peer-reviewed.”^[6] (Peer review is when other researchers go over a paper before a journal publishes it.)

The peer-reviewed version appeared in JACC: Advances in March 2024, a year after the headlines.^[3] It reports the same headline result. But it also slices the same data several other ways, and the slices don’t tell the same story. Here they are, with the share of each group who had a heart event and the adjusted risk:

• The headline group (305 low-carb vs 1,220 others): 9.8% vs 4.3%, or 2.18 times the risk. A clear result.
• Everyone who met the definition (2,034 vs 8,136): 6.3% vs 5.0%, or 1.31 times the risk. Still clear, but much smaller.
• People whose diet was recorded at least twice: 5.6% vs 2.8%, or 1.99 times the risk. Not a clear result.
• Anyone under 50 g of carbs a day, with fat not counted: 7.8% vs 3.6%, or 2.05 times the risk. Not a clear result either.

Figure 2Same data, four ways of slicing it

Headline group (305 people)2.18×

Everyone who met the definition (2,034)1.31×

Diet recorded at least twice1.99×

Under 50 g of carbs a day2.05×

statistically clearnot clear: wide enough to include no effect1× = no difference

The headline group was the people who’d had blood taken on the same day as their diet questionnaire. That’s why the researchers focused on them for their main question, which was about cholesterol. Widen the lens to everyone in the study who met the definition, and the “doubling” becomes a 31% higher risk. The gap in plain numbers drops from 5.5 percentage points to 1.3.

(Those two groups aren’t perfectly comparable, because the researchers adjusted for a slightly different set of factors in each. The direction of the change is the point.)

The two checks that would make the finding most convincing didn’t produce statistically clear results. One looked at people whose diet was measured more than once. The other looked at anyone eating under 50 g of carbs a day, the closest the data gets to real keto (though it counted carbs only and ignored fat). Both point the same way as the headline. But the uncertainty is wide enough that they’re also consistent with no effect at all. In the measured-more-than-once group, the range of plausible results runs from about half the risk to eight times it.

One more oddity, worth noticing because the abstract (the short summary at the top of a paper) is what most people actually read. The paper’s abstract says the links with heart events were “confirmed” in the people with two or more questionnaires. Its own limitations section says “the association with MACE was not statistically significant in the ≥2 subgroup analysis” (MACE is the paper’s shorthand for those major heart events). It suggests that may be down to the smaller sample. That’s a reasonable explanation. It just isn’t what “confirmed” means.

What a study like this can’t tell you

It raises a hypothesis. It doesn’t answer the question.

Steven Nissen, cardiologist, to Medscape

You might fairly say: fine, but a 31% higher risk across 10,000 people is still something.

It might be. But this kind of study can’t tell you whether the diet caused it. People who eat very little carbohydrate differ from everyone else in lots of ways, and some of those ways matter for heart disease. Steven Nissen, a Cleveland Clinic cardiologist, made that point at the time. “It raises a hypothesis,” he told Medscape. “It doesn’t answer the question.”^[7]

Alice Lichtenstein, a nutrition researcher at Tufts, suggested to TCTMD that the type of fat people ate (largely animal fat) might matter more than the carbs they cut.^[8] That’s a plausible alternative, and the study can’t separate the two either.

So is keto fine for your heart? Not quite

This is where it would be easy to swing too far the other way. “The study was weak” isn’t the same as “keto is good for your heart,” and the stronger research has a real warning in it.

Randomized trials, where people are assigned a diet instead of choosing one, can’t follow people for twelve years. So they measure cholesterol rather than heart attacks, and on cholesterol they’re fairly consistent. A 2023 umbrella review (a review that gathers up other reviews) pulled together 17 meta-analyses of 68 trials. It found that ketogenic diets were “associated with a clinically meaningful increase in LDL-C,” the cholesterol most tied to heart disease, and it rated that evidence high quality.^[9]

Those trials were short, a median of 13 weeks. The same review found that very-low-calorie keto diets in people with overweight or obesity didn’t worsen LDL. A 2025 analysis of 174 trials found the same pattern: ketogenic diets produced more weight loss than other diets, and bigger rises in LDL.^[10] They also tend to lower triglycerides (another kind of fat in the blood) and raise HDL (the so-called good cholesterol), which is part of why the overall picture is still argued about.^[11]

There’s no trial yet that follows keto dieters long enough to count heart attacks. And the best-known recent paper arguing the other way, a 2025 KETO-CTA analysis titled “Plaque Predicts Plaque, ApoB Does Not,” was retracted in May 2026.^[12]

Figure 3From headline to footnote

1. Mar 5, 2023The American College of Cardiology’s press release says a “keto-like” diet “may be linked” to higher risk.
2. Mar 6, 2023The headline arrives: keto “doubles the risk of heart attack, stroke.”
3. Mar 2024The peer-reviewed paper is published, with extra checks that shrink the effect or make it unclear.
4. May 2026A KETO-CTA paper arguing the opposite is retracted.

So the real summary is narrower than either headline. The 2023 study can’t tell you that keto doubles heart risk. The trials suggest keto raises LDL cholesterol on average, by a modest amount for most people and a lot for some.

What to carry into your weekend

If you eat low-carb or keto, the thing to take from all this isn’t a ruling on the diet. It’s a number. The trials point at LDL cholesterol (and its close cousin, apoB) as the thing that changes, and it changes a lot more in some people than in others. That’s a blood test, and it’s a conversation for you and your own clinician rather than for a headline or a blog.

And the next time a diet study doubles something, three questions will get you most of the way there:

1. What did the study call the diet, and is that what you’d call it?
2. How many times did they actually measure what people ate?
3. When the full paper came out, did the bigger group say the same thing as the headline?

Enjoy the rest of your weekend, however many carbs are in it.

---

Sources

1. Lisa O’Mary, “Keto Diet Doubles the Risk of Heart Attack, Stroke: Study,” WebMD, March 6, 2023. webmd.com
2. American College of Cardiology, “‘Keto-Like’ Diet May Be Linked to Higher Risk of Heart Disease, Cardiac Events,” press release, March 5, 2023. acc.org
3. Iatan I, Huang K, Vikulova D, Ranjan S, Brunham LR. “Association of a Low-Carbohydrate High-Fat Diet With Plasma Lipid Levels and Cardiovascular Risk.” JACC: Advances 2024;3(6):100924. doi:10.1016/j.jacadv.2024.100924 (full text: PMC11287001)
4. Evert AB, et al. “Nutrition Therapy for Adults With Diabetes or Prediabetes.” American Diabetes Association consensus report, Diabetes Care 2019. PMC7011201
5. Bellissimo MP, Hundley WG. “Low-Carbohydrate High-Fat Diets, Lipid Levels, and Cardiovascular Risk.” Editorial, JACC: Advances 2024;3(6):100975. PMC11286988
6. Chloe Taylor, “Keto diet may not be as healthy as some nutritionists and influencers want you to believe, scientists warn,” Fortune, March 6, 2023. fortune.com
7. Sue Hughes, “‘Keto-like’ diet linked to doubling of heart disease risk,” Medscape (via MDedge), March 8, 2023. mdedge.com
8. Shelley Wood, TCTMD, March 8, 2023. tctmd.com
9. Patikorn C, et al. Umbrella review of ketogenic diets. BMC Medicine 2023;21:196. doi:10.1186/s12916-023-02874-y
10. Feng S, et al. American Journal of Clinical Nutrition 2025;122(5):1461-1478. doi:10.1016/j.ajcnut.2025.09.012
11. Chang C, et al. Endocrine Practice 2026;32(5):819-828. doi:10.1016/j.eprac.2026.01.009
12. Retraction notice for Soto-Mota A, et al. “Longitudinal Data From the KETO-CTA Study: Plaque Predicts Plaque, ApoB Does Not.” JACC: Advances 2025;4(7):101686. Retracted May 2026. doi:10.1016/j.jacadv.2026.102824