Carnivore diet: the most radical elimination diet there is

A Spanish research group brought together metabolomic, archaeological and microbiological findings in 2025 on how early hominins adapted to animal foods. Humans are placed there as flexible omnivores with a partial carnivorous specialisation.

The evolutionary line works neither as proof for nor as proof against. It works for one sentence: we are built for flexibility, not for a single food source.

A team around Belinda Lennerz at Boston Children's Hospital surveyed 2,029 adults through social networks in 2020 who had been eating purely animal foods for at least six months.

Median duration 14 months. 95 percent reported better overall health, unwanted symptoms ranged between under 1 and 5.5 percent, the median body mass index fell from 27.2 to 24.3. Participants with diabetes reported an HbA1c lower by 0.4 percentage points and a reduction in their diabetes medication.

What this means for you: people who keep this up for a long time mostly feel good. That number says no more than that, and no less either.

Nicholas Norwitz and Adrián Soto-Mota collected ten cases in 2024 with confirmed ulcerative colitis or Crohn's disease who stayed stable on a ketogenic or carnivore diet without medication. Six cases of ulcerative colitis, four of Crohn's disease, the improvement in the quality of life questionnaire ranged between 72 and 165 points.

Courses like that exist. They are selected success stories without a control group. For Crohn's disease and colitis, guideline based therapy remains the foundation.

A group around Gregory Austin gave people with diarrhoea dominant irritable bowel syndrome a diet with 20 g of carbohydrate a day for four weeks in 2009. Of 17 people enrolled, 13 completed, and all 13 met the responder definition. Stool frequency fell from 2.6 to 1.4 per day, independently of weight loss.

So the effect on the gut is real. But it probably comes from dropping the fermentable carbohydrates, not from the meat.

David Weigle had nineteen people go through three phases in 2005: two weeks at 15 percent of energy from protein, two weeks isocaloric at 30 percent, then twelve weeks of the same as much as they liked.

In the free phase, spontaneous energy intake fell by 441 kcal a day and body weight by 4.9 kg. And that although leptin fell and ghrelin rose, so the hunger hormones were pushing the other way.

Important for placing this: in that study carbohydrate intake stayed constant at 50 percent of energy, what was lowered was fat. So it describes the protein effect and not a carnivore situation.

What this means for you: plain meat sits very high on protein density. Whether the same effect follows for you is not shown by this. More on that under satiety and protein and under appetite regulation.

Kevin Hall had twenty weight stable adults live as inpatients in 2019: two weeks each of ultra-processed and unprocessed food, in random order, matched for calories, macronutrients, sugar and salt.

On the ultra-processed food the same people ate 508 kcal more per day and gained 0.9 kg. On the unprocessed food they lost 0.9 kg.

What this means for you: carnivore takes care of this effect along the way, and a large part of what people experience on carnivore probably sits in there. The physiology is described under unprocessed food and under quality instead of quantity.

Alice Gibson brought together studies in 2015 that measured appetite before and during ketosis. On a ketogenic low carbohydrate diet people were less hungry. The changes were small, but they appeared in a state of energy restriction, which otherwise raises appetite.

So ketosis does not take hunger away. But it may keep hunger from rebounding while weight is going down.

The polar explorer Vilhjalmur Stefansson and his companion Karsten Andersen lived for a year on meat alone at Bellevue Hospital in New York, under medical observation.

Walter McClellan and Eugene Du Bois published it in 1930 in the Journal of Biological Chemistry. The title names the focus: kidney function and ketosis. A companion paper examined nitrogen, calcium and phosphorus, a third publication appeared in the journal of the American dietetic association.

To this day that is the only documented, medically supervised year of purely animal eating that I know of. About populations it proves nothing: two people, and one of them promoted this way of eating himself.

A New Zealand team around Sylvia Goedeke calculated four carnivore day plans against the official reference values in 2025. These are model plans, not participants.

Above the recommendation were zinc, vitamin B12 and selenium among others, sodium even by fifteen to twenty fold. Consistently below it were thiamine, magnesium, calcium and vitamin C, and depending on the plan also iron, folate, iodine and potassium. Fiber intake came out below 1 percent of the guideline value in all four plans.

In a controlled depletion experiment by the Iowa group, five healthy volunteers received no vitamin C for three months, after which it was repleted over four months. François Abboud and colleagues studied the vascular and circulatory responses in them in 1970.

After three months the plasma level was 0.178 mg per 100 ml, and all five showed signs of scurvy. After four months of repletion it was 1.68 mg per 100 ml, and the complaints had gone.

What this means for you: scurvy developed here over months of complete withdrawal, not over weeks. Fresh meat and organ meats contain small amounts of vitamin C. Whether ketosis might lower the requirement is a hypothesis without testing in humans.

In practice that means: watch for bleeding gums, for pinpoint bleeding or bruises without cause, for wounds that close more slowly, for new joint or muscle pain and for unusual fatigue. If any of that appears, have your vitamin C level measured and end the phase. And have it measured anyway if you stay with this longer than a few weeks.

Lawrence David had volunteers eat purely animal and purely plant based food for five days each in 2014, with a baseline before and a washout phase afterwards.

The gut flora shifted significantly, a single day after the food reached the colon. 22 bacterial clusters changed, on the plant based food only 3. Two days after the end of the study the microbiome was as it had been before.

That return figure is the most reassuring piece of information on this topic. About months or years it says nothing.

In a study in mice, a team around Mahesh Desai colonised germ free animals with a defined human gut microbiome in 2016 and fed them without any fiber.

The bacteria switched to the host's mucus glycoproteins. The mucus layer became thinner, a mucosal pathogen reached the epithelium more easily, and the colitis ran a lethal course in these animals.

That is the strongest objection to permanently leaving out all fiber, but it is a mouse study and does not transfer one to one. Why the mucus layer counts is described under intestinal permeability and zonulin.

Hyon Choi followed 47,150 men without gout over twelve years. Across 730 new cases, the relative risk in the highest versus the lowest fifth of meat intake was 1.41, for seafood 1.51. Purine rich vegetables and total protein intake were not linked to a higher risk.

In the same observational study, gout risk sat lower in the highest fifth of dairy intake (RR 0.56). That is a statistical observation in a cohort of men and not a recommendation to use dairy against gout. Why it is so has not been clarified.

So it is not purines in general that are the topic, it is meat and seafood. Anyone who has gout or has had it should not try this on their own.

Out of the L-carnitine in red meat the gut microbiome can form trimethylamine, which the body then converts into TMAO. In a study in mice, continuous carnitine intake accelerated vascular calcification, but only with an intact gut microbiome. In humans, in the same piece of work, carnitine levels among 2,595 people undergoing cardiac work up were linked to cardiac events, but only when TMAO levels were high at the same time.

The authors conclude from this that the gut microbiome could contribute to the well documented link between high red meat consumption and cardiovascular risk. In the same piece of work, people on a mixed diet formed markedly more TMAO after a carnitine dose than vegans and vegetarians did.

That is not proof of causality in humans, but it is more than nothing. The same portion can produce different amounts of TMAO in two people, and that is one of the most interesting trails toward individualisation in this field.

Nicholas Norwitz and colleagues analysed survey data from 548 adults on a carbohydrate restricted diet in 2022.

Body mass index was inversely linked to the LDL change: the leaner, the steeper the rise. 100 of the 548 met the triad of LDL from 200 mg/dl, HDL from 80 mg/dl and triglycerides up to 70 mg/dl.

So if your LDL climbs steeply although you are lean, that is a described pattern. This too is a survey with self report.

Norwitz and William Cromwell documented a single person with this pattern in 2024 who ate twelve Oreo cookies a day for 16 days, meaning 100 g of additional carbohydrate. Six weeks of rosuvastatin followed.

LDL fell from 384 to 111 mg/dl, so by 71 percent. After the washout it rose to 421, and on the statin it fell to 284, so by 32.5 percent.

One single person, and the authors stress that no health advice follows from it. But it does suggest how closely this LDL can hang on carbohydrate intake.

Because a medicine appears here, the placing belongs with it: rosuvastatin is a prescription only active substance from the statin group. It can cause muscle complaints among other things, in rare cases rhabdomyolysis, and changes in liver values. It is not approved in pregnancy and while breastfeeding, and it can interact with other medicines, for example with certain immunosuppressants, fibrates or antivirals. I deliberately give no dosages here.

More important still: if a statin has been prescribed for you, do not stop it because of this single case. Statins have data on hard endpoints, an LDL lowered through carbohydrate intake has no such data. If you want to raise this question, raise it with your treating doctor.

Matthew Budoff compared eighty people with carbohydrate related LDL from 190 mg/dl by coronary CT in 2024 with eighty matched participants from a large cardiac cohort.

Mean LDL in the keto group 272 mg/dl, mean duration 4.7 years. Plaque burden did not differ significantly from the comparison group, whose LDL sat 149 mg/dl lower.

That is a snapshot, not a longitudinal study. A reason to keep the question open, and not a reason to file an LDL of 272 mg/dl away as harmless.

Important for placing this: the longitudinal analysis of the same cohort was retracted in 2026 because methodological concerns arose that were too large for a correction. This snapshot is formally not affected by that. It does however stand on thinner ground than it first appears to.

The Institute for Health Metrics and Evaluation reassessed six endpoints for unprocessed red meat with a meta-regression in 2022: weak evidence for colorectal cancer, breast cancer, type 2 diabetes and ischaemic heart disease, no evidence for stroke. Risk was lowest at 0 g per day, but the uncertainty interval ran from 0 to 200 g per day.

So the most honest answer is: not known precisely enough. This body of data does not yield a clear number.