Can the tumour burn ketones, and can her immune system be taught to see it

Written 14 September 2026, after her partner asked whether brain tumours can use ketones, whether sarcoma-like tumours are more metabolically adaptable, how fast they switch, and whether anything can make the tumour more visible to the patient's own immune system.

Short version, and it changes something we are doing right now.

her partner is right to be worried, and my earlier position was too soft. Until today the line in these files was that ketogenic diets have no evidence of benefit in cancer. That was true but incomplete. There is now evidence of a mechanism by which a ketogenic diet could actively help a tumour, and the clearest demonstration of it is in chondrosarcoma, which is the main alternative diagnosis on the patient's own differential.

The practical conclusion is narrow and I want it stated before the detail, because it should not be read as "stop the food plan": keep the fat, drop the goal. Everything she is eating is fine. What should stop is treating ketosis as a treatment.

1. The theory the ketogenic diet rests on, and why it is wrong

The idea is Warburg's, popularised by Seyfried. Cancer cells are said to have damaged mitochondria, to depend on burning glucose, and to be unable to use ketones. Starve the body of carbohydrate, the healthy brain switches happily to ketones, and the tumour is left with nothing.

That premise has now failed in several tumour types.

The paper that matters most for the patient is in chondrosarcoma. Published in 2024. In conventional chondrosarcoma cells:

• Ketone burning switches on when the cells are starved. The machinery appears exactly when you take the food away.
• The cells make the enzymes for it, and adding the ketone itself re-induced the enzymes and the transporter that brings it into the cell. So the diet supplies both the fuel and the signal to build the equipment that burns it.
• They can also make ketones, not just consume them.
• The authors' own conclusion is that starved chondrosarcoma cells use ketone bodies for energy and for proliferation.

The reason is worth understanding. Conventional chondrosarcoma grows inside dense cartilage, which is already short of oxygen and nutrients. It evolved in a starvation environment. A tumour that grew up starving is a poor target for a starvation strategy.

Glioblastoma does the same thing and adds a warning. A 2020 study from UCLA found brain tumour cells burning fatty acids throughout their metabolism, and found that blocking fatty acid burning slowed tumour growth. But the unrestricted ketogenic diet did not slow tumour growth, and in some models significantly shortened survival. The paper's title says it outright: glioblastoma uses fatty acids and ketone bodies for growth, allowing it to progress during ketogenic diet therapy.

And the one proper trial was negative. ERGO2 randomised 50 patients with recurrent malignant glioma to reirradiation with or without a calorie-restricted ketogenic diet plus fasting. The diet worked as a diet: 17 of 20 completers were in ketosis by day six and their glucose fell. Progression-free survival at six months was 20 per cent with the diet and 16 per cent without. No difference in any survival measure.

2. Are these tumours more adaptable, and how fast do they switch

More adaptable: on what we have, yes. These tumours do not have one fuel. They run several and move between them depending on what is available.

There is nothing published on ketone use in chordoma itself. That is a gap, not a reassurance. Chondrosarcoma is chordoma's nearest neighbour and its main lookalike, and chondrosarcoma does this. The sensible assumption is that chordoma can too, until somebody checks.

Speed, honestly, in two halves. In the dish it is fast: the enzymes came up within the span of a starvation experiment, which is hours to a few days. This is a regulatory response, a cell turning genes on, not a slow evolutionary process. In a person, nobody has measured it. Anyone who gives you a number is extrapolating from a laboratory.

The uncomfortable implication is that the window in which a ketogenic diet could restrict the tumour may be very short, or may not exist.

3. Making the tumour visible to her immune system

This is the good half of the question, and chordoma has an unusually good target.

Brachyury is a protein switched on in essentially every chordoma. It is what the pathologist stains for to make the diagnosis. It is almost absent from normal adult tissue. That combination, everywhere in the tumour and nowhere in the patient, is exactly what a cancer vaccine wants.

It has been tried properly, three times, and here is the honest scoreline.

• A phase 1 yeast-based brachyury vaccine was safe and showed a signal of activity.
• A randomised, double-blind, placebo-controlled phase 2 followed, 24 patients with locally advanced unresectable chordoma, vaccine or placebo with radiotherapy. One partial response in each arm. Median progression-free survival 20.6 months with the vaccine and 25.9 months with placebo. No benefit.
• A different brachyury vaccine with radiotherapy in 29 patients gave two lasting partial responses and a high rate of stable disease, but missed its own target of four responses.

Checkpoint inhibitors such as pembrolizumab have been used in chordoma, but the literature is case reports and small series in refractory disease, not first-line treatment.

So: the idea is sound, the target is genuinely good, it has been tested, and the randomised test did not show benefit. Immunotherapy in chordoma is a real research field and is not yet a treatment. It is for later, and it is not an alternative to surgery and proton therapy.

4. The link her partner drew, and one correction

Her partner connected this to the worry about opioids suppressing innate immunity. The instinct is reasonable but the two are different parts of the immune system.

• Opioids hit natural killer cells hardest, and those are innate immunity.
• Brachyury vaccines and checkpoint drugs work through T cells, which are adaptive immunity.

Morphine suppressing NK function is not the reason a vaccine would or would not work. They are different arms doing different jobs.

Where the point does survive: NK cells are the main defence against tumour cells in the bloodstream, and NK function matters most around surgery. That is a real concern with a proper home, and it is already question 21 on Wednesday's sheet.

5. What this means for what the patient eats tomorrow

Two things have been happening at once in her food plan and they need separating.

One, getting calories and protein into a 51 kg woman before major surgery. This is evidenced, it is urgent, and it is working. The butter, the cream, the coconut oil and the cheese have been doing the heavy lifting all week. Nothing here argues against any of it. European clinical nutrition guidance is explicitly against restricting the diet of cancer patients, and being underweight going into skull base surgery is a risk she can actually do something about.

Two, aiming at ketosis in the belief that it starves the tumour. That is the part that has got worse rather than better today.

So: keep the fat, drop the goal. A high-fat, high-calorie, low-sugar diet is an excellent way to get 2,200 calories into someone with no appetite, and that is reason enough to carry on exactly as she is eating. What should stop is counting ketosis as a win and treating carbohydrate restriction as treatment. If a potato or a piece of fruit helps her eat more overall, that is now the better choice and not the worse one.

I have been recording "clearly ketogenic day" in her food diary this week as though it were an achievement. That framing was wrong and I have corrected it.

This belongs on Wednesday's list rather than being settled in the house. It is the same shape of problem as the antioxidant question: a well-meant thing with a mechanism that might cut the wrong way.

Where the numbers come from

• Vargas-López M and colleagues. The ketone body beta-hydroxybutyrate as a fuel source of chondrosarcoma cells. Heliyon 2024. PMID 38694129.
• Sperry J and colleagues. Glioblastoma utilizes fatty acids and ketone bodies for growth allowing progression during ketogenic diet therapy. iScience 2020. PMID 32861192.
• Voss M and colleagues. ERGO2 randomised trial of calorie-restricted ketogenic diet and fasting with reirradiation for malignant glioma. International Journal of Radiation Oncology Biology Physics 2020. PMID 32619561.
• DeMaria PJ and colleagues. Randomised, double-blind, placebo-controlled phase 2 of yeast-brachyury vaccine with radiotherapy in unresectable chordoma. The Oncologist 2021. PMID 33594772.
• Heery CR and colleagues. Phase 1 of GI-6301 targeting brachyury. Cancer Immunology Research 2015. PMID 26130065.
• Cote GM and colleagues. Phase 2 of a brachyury-targeting vaccine with radiotherapy in advanced chordoma. Cancer 2024. PMID 38985843.
• Wang X and colleagues. Immunotherapy for advanced chordoma, a systematic review. Cancers 2022. PMID 36612259.

Full working is in research thread 24.