Is Butter Really Dangerous to Eat?
Ken D. Berry, MD, family physician
Published 11/16/26
In my view, no. The 2025 study that set off the "butter is dangerous" headlines is an observational study. It can never establish causation, and the hazard ratios it reports are very small. Humans have eaten butter for thousands of years, and this study does not give you a good reason to replace it with oils humans have only eaten for a few hundred.
What did the news headlines say about butter?
You have probably seen the news reports. Newsday ran "Butter can shorten your life, increase cancer risk, study finds." That wording implies causation, no ifs, ands, or buts. Fortune ran a headline saying one simple dietary change can help you fight back against cancer and heart disease. NBC News said eating less butter may improve health and protect against heart disease and cancer. At least they included the word "may."
The study itself was published in JAMA Internal Medicine, one of the journals of the American Medical Association. The title is "Butter and Plant-Based Oils Intake and Mortality." My friend and colleague Dr. Gil Carvalho has already made a video about it. I greatly disagree with his dietary recommendations, but I believe he is earnest and honest. Here is my step-by-step breakdown of why I think this study tells you nothing.
Can an observational study prove butter causes death?
No. This is a prospective cohort study, which is a type of observational study. It cannot ever establish causation. When you see a headline that says butter causes something, or butter can lead to something, that is poor reporting on its face. You can ignore that article.
There was also no randomization. People were not randomly assigned to eat butter or oil, and that raises the risk of confounding.
How was butter intake measured?
By self-reported food frequency questionnaires, given every four years. They ask questions like how many rashers of bacon you eat in a week. Some of these questions are very prone to recall bias, and it is well known that food frequency questionnaires can lead to underestimating or overestimating what people eat.
The study then calculated cumulative averages of butter and plant oil intake, which may not reflect what people actually ate over time. In science we want actual data we can crunch. This is not that.
Can researchers adjust away healthy user bias?
The study adjusted for BMI, smoking, alcohol, and physical activity. But adjusting is a mathematical formula in a computer model. It can never truly remove those factors. It is a good faith effort, nothing more.
Healthy user bias is a big one. People who try to be healthy, who do not smoke or drink and who exercise, have been trained by the media to avoid saturated fat, butter, lard, and bacon grease. Many of the people eating the most butter, bacon, and beef are people who are not trying to be healthy at all. They eat whatever, smoke whatever, and drink whatever. That bias is baked into the cake, and in my view it is impossible to adjust for.
Socioeconomic factors are not fully controlled either. People with lower incomes are doing the best they can, without a personal chef, a nutritionist on call, or extra refrigerators and freezers full of healthy food.
What did the study lump together?
In the video I list several problems with how the foods were grouped:
- Butter was not separated from other dairy intake patterns.
- Butter and plant oil intake were not isolated from the rest of the diet, so it is hard to see any direct effect.
- The headline plant oil result lumped olive oil, canola, soybean, corn, and safflower oil together into one group.
- There was no difference made between cooked and uncooked oil. Cooking soybean or canola oil at high temperatures causes oxidation, which can cause inflammation.
- The industrial processing of soybean, corn, and canola oil was not looked at. These oils have to be refined, bleached, and deodorized, and in my view that processing can create pro-inflammatory compounds.
How reliable are death certificates?
Cause of death in this study was classified using death certificates, along with medical records. I have been a doctor for 22 years and have filled out hundreds of death certificates. Doctors do not enjoy it and do not get paid for it. Many times there was no autopsy and I had no idea what really killed the person, so the honest truth is that doctors often just write heart disease or heart attack even without evidence.
I remember many situations where the person died at home, the family found them days later, there was no autopsy, and they were already buried by the time the death certificate reached me. In my view, death certificates are a terrible way to learn what actually killed someone.
How strong were the results?
The hazard ratio for the highest butter intake was 1.15. A hazard ratio of 1.0 means no difference in risk. The hazard ratio for the highest plant oil intake was 0.84. These are very small numbers that suggest a weak association, and in my opinion they are not clinically significant. My colleague who believes in a plant-based diet would say they are. That is two opinions.
In my view, the study does not give a convincing mechanism for how butter would cause cancer death. We have eaten butter, lard, and beef tallow for thousands of years. There was no subgroup analysis of omega-6 versus omega-3 ratios, and there were many subgroup analyses, which raises the risk of false positive findings.
Who was in the study, and how old is the data?
The data comes from long-running cohorts of nurses and health professionals. They were mostly upper middle class white men and women. I think it is a valid question whether these results apply to poor Hispanic or African American populations.
The data goes back to 1990. The headlines in 2025 make it sound like new research, but it is built on data from the 1990s, 2000s, and 2010s.
Some of the authors are at the Harvard T.H. Chan School of Public Health, an institution that in my view has historically promoted plant-based eating. With no blinding and no randomization, it is easy for researchers' beliefs to filter into how the results are interpreted.
What should you do with headlines like this?
In my experience, almost every scary headline like this is based on an observational study with very weak hazard ratios, done by researchers who believe in a plant-based diet. It is not randomized, controlled, blinded research. At most, a study like this suggests that maybe more controlled research is needed. In my opinion, it is worse than worthless for deciding what to eat.
If you were thinking about low carb, keto, or carnivore and this study gave you pause, try it for 90 days. In my opinion, your own results, how much weight you lose, how much your labs improve, and how much better you feel and look, are more valuable to you than the data from a study like this one.
This is not medical advice. Work with your own doctor to monitor your health and medications.
Frequently asked questions
Did the 2025 study prove butter shortens your life?
No. It is an observational study, and observational studies can never establish causation. Headlines saying butter causes early death misread the study design.
What were the hazard ratios for butter and plant oils?
The hazard ratio for the highest butter intake was 1.15, and for the highest plant oil intake it was 0.84. In my opinion these are very weak associations that are not clinically significant.
Why does healthy user bias matter here?
People trying to be healthy have been told to avoid butter, while many people eating the most butter are not trying to be healthy at all. In my view no computer model can fully adjust for that difference.
Is the data in the butter study new?
No. The data goes back to 1990, with questionnaires given every four years. The 2025 headlines make it sound new, but it comes from the 1990s, 2000s, and 2010s.
Should I switch from butter to vegetable oils?
In my view, no. The headline plant oil result lumped olive oil in with canola, soybean, corn, and safflower oil, and it did not account for cooking or industrial processing of those oils.
Research referenced in the video
I break down this study in the video and link it in the video description. This is the link I could verify. Do not read more into a paper than its title and abstract support.
Zhang Y, Chadaideh KS, Li Y, et al. (2025), "Butter and Plant-Based Oils Intake and Mortality": https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2831265 (PubMed: https://pubmed.ncbi.nlm.nih.gov/40048719/)
About Dr. Berry
Dr. Ken Berry is a board-certified family physician who has cared for over 25,000 patients in more than two decades of practice in rural Tennessee. He grew up in Hohenwald, Tennessee, and earned his MD at the University of Tennessee Health Science Center.
He is the author of the #1 bestselling Lies My Doctor Told Me and the creator of the Proper Human Diet.
He lives on a farm in Holladay, Tennessee, with his wife Neisha and their two children.