From thread from November 2019. With updated letter to the editor attached to the end.

New paper suggests that ketogenic diet makes mice more resistant to influenza infection.

I want to explain why I am disappointed by this paper.

Ketogenic diet activates protective γδ T cell responses against influenza virus infection

I was very excited and was about to retweet this paper, but then I decided to spend some time reading instead.

This paper shares the same flaws as the other ketogenic diet papers: it uses inappropriate control diets.

To understand why this is a problem, let’s take a look at the protein restriction literature.

Protein restriction has been known to inhibit the growth of or prevent cancer altogether for a hundred years.

This is beautifully reviewed in this commentary published last year in Cell Metabolism, here:

Eating the Beast: Dietary Protein and Anticancer Immunity

In this commentary, it is noted that protein restriction was shown in 1976 to exert this effect on cancer via immune system modulation.

This original paper by Good et al (1976) is outstanding and very much worth reading, but new papers with state-of-the-art techniques have explained this effect in even greater detail. One of these papers is the subject of the above commentary.

This paper demonstrated the molecular mechanism by which protein restriction caused enhanced immunosurveillance and caused the immune system to more effectively battle cancer.

Eating the Beast: Dietary Protein and Anticancer Immunity

In the above study, protein was restricted by a mere 25% (20%➡️15%), yet profound immunomodulatory effects were seen, which inhibited tumor growth.

How does this relate to this ketogenic diet & influenza paper? Here’s how.

Here is a screenshot of the diet descriptions in this study.

And below is a screenshot of the control diet used in this study: Envigo 2018S.

As you can see, 24% protein as % of kcals.

What about the ketogenic diet (D12369B from Research Diets; far right diet)? As you can see below, 10% of protein as % kcals.

That is a 60% protein restriction.

So not only is the diet ketogenic, but it’s also profoundly protein-restricted: far more restricted than in studies that have shown profound immune system-enhancing effects.

So is it the ketosis or the protein restriction driving the effect? We have some indication of this. Because investigators also used the control chow and added butanediol (BD), which converts to ketones: this helps to isolate the effect of the ketones.

(It’s not perfect though because the calories from BD also dilutes the control diet, but it’s a start.)

What did they find with this diet?

They found no difference in immune function from the control diet.

And this is exactly what we would expect if the protein restriction was accounting for the effect from the KD!

They also used another diet: the high-fat diet. It is basically the ketogenic diet without the ketosis. If there was some other impact of the KD besides ketosis that would be expected to also cause this influenza-resistant phenotype, you would see it in the HFD, right?


No protection from the high-fat diet!

What was the high-fat diet? Here is the macro breakdown.

It’s also important to note that many other aspects of the diet are also different–vitamins, minerals, etc.–so it is not a control diet in any sense of the term.

Yes, it is a high-fat, low-carb diet. But it is also different in many other ways, which makes it a highly inappropriate diet to use as a scientific control.

Now it did produce an immune phenotype markedly different from control. But why? Is it the fat? Is it the modest protein restriction (producing a phenotype intermediate between control and the ketogenic diet)? Or the other changes?

Interesting but no way to know.

So here we have:

Control diet does not protect
Keto protects
Control diet with ketones doesn’t
Low-carb diet doesn’t

Which doesn’t make sense until you point out that the ketogenic diet is the only diet that is profoundly protein restricted.

So does this study tell us about protein restriction?

No. The study cannot answer any diet question, because there were no real control diets, and thus we cannot know which components of the diets were accounting for which changes. We learn nothing definitive from this paper, except that some aspect of dietary change drives some immunologic changes that provide resistance to influenza.

I would estimate about 95% of the studies using ketogenic diets use inappropriate diet control groups, most of these inappropriately dramatically restricting protein. I talk about this in the context of cancer, here. Relevant figure, here:

This is entirely avoidable. Remember this screenshot (below)? The company sells sets of protein- and nutrient-matched ketogenic, high-fat, & control diets! They sell several sets of these, in fact.

Investigators looking at diet effects should make use of matched diets like these!

I guess I will edit this, add some more refs, and submit it as a letter to the editor. Not sure if it will be published–I have so little experience here–but if not, I will put it on my blog.

UPDATE: We submitted our letter to the editor as a preprint to NutriXiv here. The study authors responded here. Our preprint and their response are also available on PubPeer, here.

Perhaps of most interest is the following table. In brief, the diets are grossly mismatched for a large number of nutrients. Therefore, there is no reason to believe that any one nutrient in particular (say, carbohydrate, fat, protein content, various vitamins or minerals) should have been responsible for the observed effect. In other words, the conclusion that ketosis in particular was responsible for the immunologic effect is spurious and confounded by dozens of other dietary changes. As discussed above, this conclusion is also called into question by the study’s own data.

And that’s a wrap. Thanks for reading!

Protein kcal10.40.60.8
Fat kcal15.00.43.3
Carb kcal10.00.60.3
Fiber grams1*0.3*0.2*0.3
Amino acids
Aspartic acid10.40.90.7
Glutamic acid10.50.70.9
Fatty acids
Total Saturated Fatty Acids18.3-9.90.67
Total Monounsaturated Fatty Acids15.7-
Total Polyunsaturated Fatty Acids11.2-
Total Trans Fatty Acids16.4%
Calcium phosphate11.40.51
Non-Phytate Phosphorus10.5
FluorideNR9.9 ppm
CobaltNR0.76 ppm
ChromiumNR33 ppm0.01 ppm
MolybdenumNRNR2.1 ppm
Vitamin A1105.90.3102.5
Vitamin D31397.00.7384.4
Vitamin E10.30.30.3
Vitamin K31000
Vitamin B1 (thiamin)100.10
Vitamin B2 (riboflavin)
Vitamin B3 (niacin)
Vitamin B5 (pantothenic)
Vitamin B6 (pyridoxine)
Vitamin B7 (biotin)
Vitamin B9 (folate)
Vitamin B12 (cobalamin)

Table 1. Nutrient Composition (per kcal) of Experimental Test Diets Employed by Goldberg et al. All values are rounded to the nearest decimal. A minus sign (-) indicates the likely absence of the nutrient. A zero (0) indicates a value <0.05 or 1/20 of the control levels. “NR” indicates that the value was not reported but the nutrient might be present. An asterisk (*) indicates existence of additional compositional differences that are not possible to quantify from available data. Figures reported in percentages or ppm indicate presence of nutrient but absence of comparison quantities in standard chow. Figures reported in a range represent uncertainty owing to ambiguous information. BD values are given assuming caloric dilution by 1,3-butanediol of 6 kcal/g. SC- standard chow; KD- ketogenic diet; BD – 1,3 butanediol supplemented chow; HFD – 60% high fat diet

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Enjoy this content? Without your financial help, this blog is in critical danger of not surviving. Donate here: