I’m writing this post in response to one of the most pervasive beliefs about dietary protein that has been parroted by people of all dietary tribes, most notably people eating a vegan diet (although this belief is changing), and most recently by people eating a high-fat carnivore diet.

For example, you can see this belief being shared by a prominent carnivore diet influencer below.

The TLDR version of this blog is that the commonly cited upper limit for protein intake was extrapolated from a 1973 study whose own authors later acknowledged had underestimated the liver’s capacity for urea synthesis. And rabbit starvation doesn’t provide much evidence for a hard ceiling either.

Now, let’s address the easiest and most ridiculous part of this claim: the ceiling is 35% of calories. That’s 130 grams if you eat 1500 kcal and 350 grams if you eat 4000 kcal, which obviously makes no sense. The liver doesn’t care how many calories you are eating when it is metabolizing protein.

Moving on, from a research standpoint, I was first introduced to this protein poisoning idea by a review article published in 2006, and then later by the Dietary Reference Intake rationale for protein (beginning on page 693).

Both rely exclusively on a study by Dr. Daniel Rudman and colleagues published in 1973, who sought to determine the maximal rates of urea synthesis and excretion in 10 healthy adults and 34 patients with liver disease. Our interest is in the healthy people.

Rudman placed the participants on a controlled diet and then fed them one of five protein challenges every 4–8 days, with each challenge providing a different amount of protein spread across 8 am, noon, and 4 pm. He then measured urea excretion in the urine for three days afterwards.

Annoyingly, Rudman provided the protein doses as units of protein nitrogen per kilogram of body weight raised to the three-quarter power, which means the common g/kg bodyweight amount changes with body size.

The participant shown below, for example, weighed 47 kg. When Rudman’s doses are converted into ordinary g/kg of actual body weight, this person consumed approximately 0.64 (A), 0.95 (B), 1.27 (C), 2.24 (D), or 3.17 (E) g/kg of protein.

Anyway, progression from dose A to C increased the maximal rate of urea excretion, but there was no further increase after doses D and E. Yet, urea excretion maintained its maximal rate for a progressively longer time, so Rudman estimated this to be the maximal rate of urea synthesis.

The protein “ceiling” idea, however, came from extrapolation of this result. In the 2006 review, Bilsborough and Mann took Rudman’s estimated maximal hourly rate of urea synthesis and extrapolated how much protein nitrogen the liver could theoretically process over 24 hours. They then added an allowance for protein that would be retained for biosynthetic purposes and landed on a “ceiling” of 3.6 to 4.6 g/kg.

This is, fundamentally, the finding that has been interpreted as the theoretical ceiling for protein intake above which ammonia will accumulate and cause toxicity.

I disagree.

And to my surprise, there are several methodological flaws in Rudman’s work that have been contested by later research, including by Rudman himself, which seem to have been ignored by the research community.

For example, Rafoth and Onstad published a study in 1975 seeking to relate hourly urea production to the corresponding serum amino acid concentrations reached after the ingestion of single protein meals. They fed 18 healthy adults a single meal containing up to 240 grams of protein.

As can be seen in the image above, there was a direct linear correlation between the concentration of blood amino acids and urea production, with no evidence of a plateau over the range they tested.

To quote Rafoth and Onstad:

“Even though urea synthetic rates one and one half times higher than the maximum rate reported by Rudman et al. were achieved in normal subjects in the present study, no plateau in production was reached. In their studies, both 4-h collection periods and administration of oral protein in divided doses over 8-h periods would favor a stable rate of urea production. The former by minimizing hourly differences and the latter by causing relatively stable serum amino acid levels.”

In other words, the upper limit observed in Rudman’s original research could likely be explained, at least in part, by the fact that the doses of protein were provided as multiple doses over 8 hours.

The other explanation is likely water intake and a reliance on urinary excretion of urea to estimate urea synthesis. Rudman’s participants were restricted to a set amount of water regardless of how much protein they consumed.

This issue was addressed by Hendrik Vilstrup in a 1980 publication, who gave healthy adults a constant infusion of amino acids while allowing them to drink as much water as they wanted up to 0.5 liters per hour (plus water provided through the IV).

As you can see above, there was once again a direct linear correlation between plasma amino acid concentrations and urea synthesis. Moreover, in these participants, the highest urea synthesis rates obtained were double the maximal values observed in Rudman’s original work. Vilstrup summarizes this issue succinctly:

“One explanation of this discrepancy may be that the criterion for saturation of the urea synthesis rate used by Rudman et al. was that the urinary excretion of urea did not rise further when the load of protein or amino acids exceeded a certain level… Thus the phenomenon may reflect a maximum urinary urea excretion rate at a given urine flow.”

I also want to point out that every participant in this study experienced nausea and more than half of them vomited. According to Vilstrup,

“It is reasonable to assume that hepatic urea synthesis is a saturable process, but apparently maximum rates require a-amino nitrogen concentrations that are unfeasibly high in clinical studies because of the nausea and vomiting that they cause.”

If something is “unfeasibly high in clinical studies” that use amino acid infusions, it seems unlikely that it would be achievable in daily life.

Anyway, all of this follow-up data spurred Rudman’s group to develop a kinetic urea tracer method in 1980 that confirmed that the maximal rate of urea synthesis was not obtained in their earlier investigation. In this updated study, they measured substantially higher rates of urea synthesis than in their original investigation (about 28–72% higher). To quote Rudman himself:

“Our earlier conclusion that the normal urea synthesis capacity was maximal at 55-74 mg urea nitrogen/kg BW^3/4 /hr must now be revised… With the tracer method, we found the average rate of urea synthesis in normals… was 95 mg urea nitrogen/ kg BW^3/4 /hr, instead of 55-74 as previously calculated… The synthesis rate we achieved in normals should not be considered maximal; it approached the infusion rate and was not associated with progressive accumulation of AAN [alpha-amino nitrogen].”

Basically, Rudman himself concluded his original finding was wrong and that his more recent data, despite achieving substantially higher rates of urea synthesis, still did not reach a demonstrable maximum.

Plus, we have to consider that the liver is adaptive. A 1992 study coauthored by Vilstrup showed that the liver’s capacity to synthesize urea increased as protein intake rose over two weeks. This has also been demonstrated in animals.

Obviously, we don’t know how high this adaptive potential goes, but it shows that you can’t really treat the liver as some fixed-capacity object.

It’s worth noting that Dr. Jose Antonio has published a series of studies investigating the effects of eating 4.4, 3.4, and 3.3 g/kg of protein per day, all without any apparent evidence of harm over 8 weeks.

What Rabbit Starvation Actually Tells Us

I feel as though I need to also address the concept of rabbit starvation in this post because it seems to be thrown around casually as evidence of why high-protein diets can be harmful. I think relying on the direct investigations into ammonia metabolism and urea production within the liver are more insightful, but it is worth addressing this belief around protein poisoning too.

Now, there is no actual scientific investigation into rabbit starvation that I’m aware of. Rather, the concept was popularized by Arctic explorer Vilhjalmur Stefansson in his book, The Fat of the Land:

“The groups that depend on the blubber animals are the most fortunate, in the hunting way of life, for they never suffer from fat-hunger. This trouble is worst, so far as North America is concerned, among those forest Indians who depend at times on rabbits, the leanest animal in the North, and who develop the extreme fat-hunger known as rabbit starvation.

Rabbit eaters, if they have no fat from another source – beaver, moose, fish – will develop diarrhoea in about a week, with headache, lassitude, a vague discomfort. If there are enough rabbits, the people eat till their stomachs are distended; but no matter how much they eat they feel unsatisfied. Some think a man will die sooner if he eats continually of fat-free meat than if he eats nothing, but this is a belief on which sufficient evidence for a decision has not been gathered in the north. Deaths from rabbit-starvation, or from the eating of other skinny meat, are rare; for everyone understands the principle, and any possible preventive steps are naturally taken” (pgs. 30-31).

Stefansson spent 12 years living with the Eskimos in the early 1900s, before they had been pervaded by European dietary habits. He was an observer who took copious notes and made every effort to be an Eskimo among Eskimos. Accordingly, he consumed a diet almost exclusively based on animal flesh and fat.

The idea of rabbit starvation gained further support from a single case study. When Stefansson returned home from his Arctic adventure, he conducted a year-long experiment on himself at Bellevue Hospital in New York, where he consumed an all-meat diet under medical supervision.

The experiment began with a diet primarily of lean meat at the request of the doctors. After three days on this diet of 3.8 g/kg protein, or about 45% of energy intake (fat being the remaining 55%), Stefansson developed nausea and diarrhea. In response, Stefansson radically reduced his intake of protein to about 1.4 g/kg, or about 20–25% of energy intake, which was associated with recovery. (Data from this publication)

It is impossible to establish the cause of Stefansson’s GI distress from this case study. Period.

The symptoms appeared when he abruptly increased his protein intake and went carnivore, and this seems to happen to many carnivore dieters anecdotally, but there is no investigation as to why this happened. It’s incredibly unlikely that he was at some theoretical ceiling for protein intake given everything we’ve just finished discussing.

Honestly, I’d bet that the GI issues were just from the abrupt dietary change. Stefansson when from his regular diet to pure meat with zero fiber basically overnight, and that’s a huge shift in the substrates reaching the large intestine. Plus, after he added fat, he was constipated for 10 days, so its not like the fat “normalized” anything. To quote:

“At our request he began eating lean meat only, although he had previously noted, in the North, that very lean meat sometimes produced digestive disturbances. On the 3rd day nausea and diarrhea developed. When fat meat was added to the diet, a full recovery was made in 2 days. This disturbance was followed by a period of persistent constipation lasting 10 days.” (citation)

So, to me, this looks much more like his gastrointestinal system was having a rough time adapting to an extreme dietary change. It wasn’t an issue with the lean meat or high protein intake per se.

So, what was this “rabbit starvation” that sprinkles historic accounts?

Rather than being “protein poisoning”, it was likely a combination of factors as outlined in a 1983 publication by Speth and Spielman:

• High thermic cost of metabolizing protein, meaning that people needed more total food to obtain the same net energy
• Difficulty consuming enough lean meat to meet high energy demands
• Possible appetite suppression
• Inefficient use of protein when nonprotein energy was scarce
• Nutrient deficiencies (essential fatty acids, fat soluble vitamins, and minerals)

In other words, rabbit starvation is just a term used to describe the predictable consequences of sustaining on exceptionally lean and nutritionally limited foods, particularly when super active with a high energy expenditure.

But there is remarkably little evidence that it represents “protein poisoning” caused by exceeding a fixed hepatic capacity for urea synthesis.

So, Is There an Upper Limit to Protein?

I think there is, but I don’t know what it would be. Certainly, it is going to be ridiculously high, possibly impractically so outside of laboratory conditions.

The notion that we have an upper limit actually achievable in every day life comes from a combination of two things: (1) an apparent ceiling in urinary urea excretion observed in a small study from 1973, and (2) historic accounts of “rabbit starvation”.

The first was shown to be an artifact of methodological limitations, with subsequent research unable to find a plateau in urea synthesis. Even the original research group later improved their methods and acknowledged that their original estimate was not maximal.

The second is just an unverified guessing game. There is no definition or diagnosis for “rabbit starvation”. It was simply a catchy term that got stuck to observations of people trying to survive on exceptionally lean meat who likely couldn’t get enough calories or essential nutrients to sustain normal function.