Ammonia and cancer

The Molecule We Called Waste

Ammonia is the most familiar nitrogen compound in biology and one of the least studied in cancer. It is a weak base, which means it drifts through cell membranes and makes acidic compartments less acidic — and that single, unremarkable property turns out to switch off the killing machinery of three separate arms of the immune system. Almost everything we know about this has arrived in the last nine years. Most of it in the last four.

Start with the chemistry, because the whole argument rests on it and it is not complicated.

Ammonia, NH₃, carries no electrical charge. Uncharged molecules pass through cell membranes freely. Once inside an acidic space it picks up a spare proton and becomes ammonium, NH₄⁺, which is charged and therefore cannot get back out. It is trapped — and the proton it consumed is no longer available to keep that space acidic.

So ammonia does one thing reliably and without any biological machinery at all: it finds acidic compartments and makes them less acidic. No receptor, no transporter, no gene required. Just a molecule small enough to slip through a membrane and basic enough to grab a proton on the other side.

The reason this matters is that several of the immune system's most important weapons are built, loaded and fired inside acidic compartments.

A molecule produced constantly by every tissue, excreted by a liver pathway many tumours switch off, and measured in hospitals for a century, may be acting as a tumour-promoting agent — not by damaging DNA, but by disabling the cells that would otherwise remove the damaged ones. The hypothesis is old enough to be embarrassing and new enough that the decisive experiments are still being published.

What follows is a case, not a conclusion. It is assembled from work most of which was done for other reasons by people asking different questions.

The same measurement, asked three different questions. Almost everything known about what ammonia does has arrived since 2017, and most of it since 2022.

One

Ninety-Three Years of Asking a Different Question

In 1933 a physician named Fuld published a short paper in Klinische Wochenschrift reporting that patients with leukaemia had raised ammonia in their blood. The title of the paper is the thing to notice: it asks whether ammonia measurements are diagnostically usable. From the very first observation, the question put to the molecule was whether it told you anything — not whether it did anything.

What followed was thirty years of argument about whether the finding was even real. In 1960 and 1961 two researchers proposed that the elevation was an illusion created by the test tube: leukaemia patients have enormous numbers of white cells in their blood, those cells go on metabolising after the sample is drawn, and the ammonia they release accumulates in the tube before anyone measures it. A reasonable objection, and for three decades it stood.

It was settled in 1967 by a group in Milan who simply designed the objection out of the experiment. They took arterial rather than venous blood, moved every sample straight into ice, separated the plasma at four degrees, and — the decisive touch — chose an anticoagulant that happens to block the specific enzyme reaction their critics had invoked. Twenty healthy subjects came in at 14.2 units. Thirty leukaemia patients came in at 47.1. A 3.3-fold elevation, and real.

A result the authors did not claim

On a single page of a journal's short-communications section, they printed their entire dataset: white cell count against plasma ammonia for every patient. Those numbers contain more than the paper says.

Ammonia tracks leukaemic cell mass almost perfectly — a correlation of about 0.9 in each of the three disease subtypes they studied. The authors reported two of those and wrote that in acute leukaemia "no correlation was observed." In fact the acute correlation is the same size as the other two; it simply fails a significance test because there were only four patients in that group.

More telling: extrapolate each of those relationships back to zero tumour burden and they land almost exactly on the healthy average. That is the signature of a tumour adding ammonia to an intact system — not of a clearance system that has broken down.

That paper has been cited twice in fifty-eight years, according to the Web of Science core collection. Different databases will give slightly different counts; none of them gives a number that changes the point.

Three honest limits belong with it. These patients were not ill from ammonia in any clinical sense — none had the neurological symptoms of liver failure, and the authors say so. The study reports no liver function data, so leukaemic infiltration of the liver cannot be ruled out as a rival explanation. And leukaemia is a disease in which the cancer cells and the immune cells are the same kind of cell, which makes this good evidence about where the ammonia came from and poor evidence about what it did.

But notice what the sub-clinical part means. A three-fold elevation that makes nobody obviously sick is exactly the range nobody investigates — and, as it turns out, exactly the range in which immune cells stop working.

Two

What a Weak Base Does to an Immune Cell

The clearest finding came from a question nobody in cancer metabolism was asking: why do natural killer cells inside tumours behave correctly in every observable way — find their target, attach to it, release their granules — and then fail to kill it?

The answer is a protein called perforin, which does what its name suggests: it punches holes in the target cell. Perforin is manufactured as an inactive precursor and has to be cut down to its working form, and that cutting happens inside an acidic storage compartment. Raise the pH of that compartment and the protein never matures. The cell arrives, engages, and discharges granules that cannot do anything.

Ammonia raises the pH of exactly that compartment. At concentrations of two to five millimolar — the range measured in real tumours — it impairs natural killer cell killing in proportion to dose. At four to five millimolar it abolishes killing altogether.

And it blunts several major cancer drugs

The same work found that ammonia suppressed the activity of rituximab, daratumumab and trastuzumab — three antibody therapies whose mechanism depends on natural killer cells doing the actual killing — and suppressed engineered CAR-NK cells too. If this holds in patients, a tumour's nitrogen chemistry would be quietly setting a ceiling on an entire class of treatment.

Dendritic cells fail by the same physics in a different compartment. These are the cells that collect fragments of a tumour and present them to the rest of the immune system — the step that starts an immune response at all. That processing also happens in acidic compartments. At five millimolar, ammonia leaves dendritic cells swollen, poorly able to engulf material, leaking reactive oxygen, and markedly worse at stimulating lymphocytes. Dendritic cells taken from patients with cirrhosis — a disease of chronically raised ammonia — showed the same defect as healthy cells poisoned in a dish.

T cells complete the set, and here the story took a turn that nobody predicted. In 2024 a group showed that effector T cells accumulate ammonia inside their own lysosomes, that this raises the lysosomal pH from about 4.5 to 6.2, and that once the storage capacity is exceeded the ammonia floods back into the mitochondria and the cell dies.

The detail that makes this hard to dismiss

The ammonia killing those T cells is their own. Activated T cells divide furiously, and rapid division runs on glutamine, and burning glutamine releases ammonia. The harder a T cell works against a tumour, the more of the poison it makes.

And it has no way out, because T cells do not have a urea cycle — the liver pathway that converts ammonia to harmless urea. They possess no CPS1, the enzyme that starts it.

The proof is the rescue experiment. Give a T cell the missing enzyme — force it to express CPS1 — and it survives, in culture and in animals. Several other interventions worked too: blocking the glutamine-burning enzyme, blocking the transporter that loads ammonia into lysosomes, or simply giving the animals phenylbutyrate, a drug used for decades to lower ammonia in children born with urea cycle defects.

That result carries a caveat, and it is the caveat that runs through this whole literature. The in vitro experiments used ten millimolar ammonium chloride — above the one-to-three millimolar range measured in living tumours, and above the range the paper itself says it is modelling. The mechanism is convincing; whether it operates at the concentration a T cell actually meets inside a solid tumour is the question the dish cannot answer. The in vivo rescue work, where the animals received phenylbutyrate or a glutaminase inhibitor, is the more persuasive part of the paper for exactly that reason.

Three arms of anti-tumour immunity. Three different compartments. One piece of physical chemistry.

And it is reversible

This is the part that matters most for whether any of it is useful. Perforin production returned to normal levels sixteen hours after the ammonia was washed away. Dendritic cell function recovered when ammonia was removed. These are not damaged cells. They are functional cells in a disabling medium — which is a problem suited to clearance rather than to blockade, and clearance is something medicine already knows how to do.

Three

We Have Been Measuring It in the One Form It Is Least Often In

Here is a problem that sits underneath all of this and is rarely stated.

A hospital ammonia test reports free ammonia. That is a real number and a useful one. But the nitrogen atom that ammonia carries does not spend much of its life as free ammonia. It travels as glutamine, the body's main nitrogen shuttle. It travels as alanine and aspartate. It gets built into the bases of DNA and RNA, into glutathione, into the polyamines that dividing cells need. It leaves as urea, or — if you give the right drug — as a compound the kidney can excrete instead.

Every one of those reads as zero on the assay.

Nine molecular identities for one nitrogen atom. Only the centre is what a clinical ammonia assay reports — the carriers, the incorporation products and the exits all read as zero. A tumour running its nitrogen economy well therefore looks, on the instrument, like a tumour with no nitrogen problem at all.

The practical consequence is uncomfortable: a tumour running a brisk and well-managed nitrogen economy — taking ammonia in, fixing it into useful molecules, keeping the free pool low — looks, on the instrument, like a tumour with no nitrogen problem at all. Every estimate anyone has made of how much ammonia matters has been made with a measurement blind to most of where the nitrogen goes.

Four

Why the Tumour Survives What the Immune Cell Cannot

An obvious objection: if ammonia is so damaging, why does it not kill the tumour too?

Part of the answer is that at high enough concentrations it does. Above roughly five millimolar, compiled studies find ammonia switching from promoting tumour growth to inhibiting it, and above ten millimolar it starts killing cancer cells outright. Tumours measured in living animals sit at roughly one to three millimolar — comfortably in the promoting band, and just underneath the ceiling.

The more interesting part of the answer is that the tumour builds itself a shield, and the shield is the acidity everyone already knew about.

ammonia → stalls the mitochondria → cell switches to glycolysis → lactate

lactate → acid outside the cell → ammonia trapped as ammonium → tumour protected

Ammonia interferes with the mitochondrial machinery that burns fuel with oxygen. A cell in that state has oxygen it cannot fully use, and glycolysis becomes its remaining option — which produces lactate, which is exported along with protons, which makes the space outside the cell acidic. Tumour exteriors typically run at pH 6.5 to 6.9 against a normal 7.4.

And in that acidic space, ammonia sits mostly as charged ammonium, which cannot cross membranes. The acid rind is a shield. Recent work states it plainly: the acidic microenvironment protects cancer cells from ammonia toxicity. The tumour has generated the conditions that make its own waste tolerable to itself.

Which creates a genuine problem for a promising treatment

Neutralising tumour acidity — with something as simple as oral bicarbonate — improves responses to immunotherapy in animal models. That is well demonstrated, and it is a real line of research.

But raising the pH outside the cell converts trapped ammonium back into diffusible ammonia, and diffusible ammonia is precisely what stops perforin maturing and antigen processing working. Buffering could relieve one immunosuppressive mechanism while amplifying another. It could equally strip the tumour of its own protection and work better than expected. Nobody has measured both effects in the same experiment, and until somebody does, this is an open question rather than an obviously good idea.

That is the kind of question this whole area is full of: not a dispute about whether something is true, but a pair of well-established findings that nobody has yet put in the same room.

One piece of evidence sits slightly apart from all of this and is easy to overlook. Cirrhosis — a disease of chronically raised ammonia, often for decades, because the liver's urea cycle has been structurally destroyed — is one of the strongest known risk factors for hepatocellular carcinoma. The standard explanation is inflammation and fibrosis, and the standard explanation is probably right. But the dendritic cells taken from cirrhosis patients and the dendritic cells poisoned with ammonia in a dish show the same defect, and the timeline — decades of exposure, then cancer — is exactly the timeline a chronic immunosuppression mechanism would predict. Cirrhosis is the closest thing this hypothesis has to a natural experiment, and it deserves to be studied as one.

Five

It Is Not Only pH. Ammonia Is Reported to Bind Proteins.

Everything so far treats ammonia as a blunt chemical agent. More recent work suggests it may also be something considerably more specific — though this is the part of the case where the evidence is thinnest, and the reader should weight it accordingly.

Ammonia has been reported to bind a protein called βII-spectrin at defined positions. That binding is said to trigger the protein's destruction, and because it is the scaffold a signalling molecule called SMAD3 depends on, SMAD3 never reaches the nucleus. The pathway it serves — TGF-β signalling — is one of the body's main brakes on inappropriate cell growth. In colorectal cancer that brake is classically lost through mutation. Here it is lost through chemistry, with every gene in the pathway intact.

And it is not an isolated oddity. Ammonia is also reported to bind a pocket on a protein called SCAP, at three identified residues, releasing a switch that turns on fat synthesis — an effect that can be blocked by a cholesterol derivative competing for the same site. Competition at a defined site is the behaviour of a drug, not of a waste product.

Both of these claims come from primary papers I have read only in abstract, or not at all; the reference notes say which. I include them because they point in the same direction as the better-verified material above, not because they are settled. If they are wrong, the rest of the article still stands.

Two proteins, then. Two reported binding surfaces. Two unrelated consequences. If they hold up, ammonia is a molecule that binds things on purpose.

Six

The Evidence Arrived While Nobody Was Looking

A hypothesis that has been available for ninety years and ignored usually deserves its neglect. This one has a different shape: the supporting evidence is almost all recent, and it has been accumulating faster each year.

In 2017, tumours were shown to take their own ammonia back up and build it into amino acids — recycling waste as a nitrogen source, at about 57% efficiency. In 2019, a laboratory working on inherited metabolic disease proposed that tumours switch off a urea cycle enzyme deliberately, so that ammonia accumulates and keeps their interior comfortably alkaline while the outside turns acidic. In 2022, ammonia in the tumour microenvironment was shown to exhaust T cells by stripping their antioxidant defences — and lowering ammonia, with an ordinary amino acid supplement, restored them.

Then 2024 brought the finding that T cells poison themselves and have no urea cycle to stop it, and with it a new named category: ammonia-induced cell death, now with its own review literature and its own prognostic gene signature in lung cancer.

And in January 2026, cause

Everything above is association and mechanism. A study published at the start of this year did the experiment that tests direction, and did it both ways.

Working in mice with liver cancer, researchers silenced each of four urea cycle enzymes in turn — removing the body's capacity to clear ammonia. All four significantly shortened survival and increased tumour burden, with blood ammonia rising as expected.

Then they went the other way. Mice fed a low-protein diet — less nitrogen coming in — had markedly lower ammonia, markedly slower tumours, and markedly longer survival.

Their conclusion was a recommendation about prevention rather than treatment: that it may be important to lower dietary protein in people at high risk of both urea cycle deficiency and malignant transformation.

By April 2026 a review was proposing an "ammonia signalling network" as a unified account of tumour metabolism, epigenetics and metastasis — describing ammonia as having "evolved from a traditional concept of metabolic waste into a critical signalling molecule." The same review states that ammonia drives the Warburg effect by blocking a key mitochondrial enzyme, that it alters the chemical marks on DNA and on the proteins that package it, and that tumour stem cells actively import it through a dedicated transporter which is now a drug target.

None of this is a fringe position any more. The reclassification has largely already happened — it has simply happened quietly, and in several literatures at once that do not often read each other.

Seven

Where This Leaves Us

The argument in this article is short enough to restate in a paragraph.

Ammonia is a weak base, so it drifts into acidic compartments and makes them less acidic. Several of the immune system's weapons are built and loaded inside acidic compartments. At the concentrations actually measured inside tumours, ammonia stops natural killer cells from maturing the protein they kill with, stops dendritic cells from presenting the fragments that start an immune response, and kills the T cells that are working hardest — which turn out to manufacture the poison themselves and to lack the liver pathway that would dispose of it. The tumour, meanwhile, is partly shielded from the same molecule by its own acidity, and uses the nitrogen to build. None of this requires a single mutation.

What should stop anyone from treating that as settled is the state of the evidence rather than the logic of it. Most of these findings are a few years old. A great deal of the work is in mice, or in cells held at concentrations an experimenter chose. The effect reverses at higher doses — above roughly five millimolar ammonia stops helping tumours and starts killing them — and nobody has yet mapped exactly where that line falls, or whether it falls in the same place for a tumour cell as for the lymphocyte beside it. A promising intervention that does not know which side of that line it is on could easily do harm.

Three things that would move this forward

Look at immune cells, not tumours. Asking whether ammonia exposure predicts who gets cancer requires an enormous study that may come back empty even if the idea is right. Asking whether people with high ammonia burdens have visibly disabled immune cells in their tissue is a far smaller question, and a far sharper one.

Settle the buffering question. Two well-supported bodies of work currently predict opposite outcomes for the same intervention, and they have never been tested against each other in one experiment.

Try clearance before disease, not after. Every ammonia-lowering result so far is in tumours that already exist — the setting where this model expects the smallest benefit. The drugs have been given safely to children with inherited urea cycle defects for decades. The people worth studying are the ones who are not ill yet.

There is one tissue where the question could be asked particularly cleanly. The kidney's proximal tubule is the one part of the body that manufactures ammonia continuously and deliberately — it is how we excrete acid. It is also where the large majority of kidney tumours begin. A European research fellowship proposed testing whether those two facts are connected, after finding that cells from that tissue accumulate mutations faster than their neighbours and carry an unusual pattern of them. That project has since closed, and I have found no published result.

A closing thought about categories

Biology has a small, respected category called oncometabolites — ordinary metabolites that build up abnormally and drive cancer by interfering with the enzymes that read oxygen levels and maintain the marks on DNA. Three molecules are usually named, and each one requires a specific mutation in order to accumulate at all.

Ammonia reaches comparable endpoints with no mutation whatsoever. It accumulates from ordinary metabolic traffic in any tissue whose disposal capacity has not kept pace with its appetite, and it disables three arms of immune surveillance by a mechanism that needs nothing but a difference in pH.

It was left out of that category for a reason that has little to do with evidence. It was classified a long time ago, as waste — and once a thing has a name, the questions people think to ask about it tend to be the questions that name invites. For ninety-three years the question put to ammonia was whether the number meant anything. Hardly anyone asked what the molecule was doing to the cells suspended in it.

A version of this argument was set out on this blog four years ago, and revised since, under the title The Ammonia Hypothesis. It is stated there in considerably bolder terms than here, and some of its claims rest on single cases rather than trials. But several of the mechanisms it proposed — ammonia pushing cells toward glycolysis, stabilising the low-oxygen response, switching on fat synthesis through SCAP, impairing dendritic cells and T cells, and the case for repurposing ammonia scavengers — have since been reported independently in the literature summarised above. Readers who want the stronger version of the claim, and its weaker evidence, will find both there.

Sources

Where a claim rests on something I could not read in full, or could not verify independently, the note says so rather than smoothing it over.

1. Ammonia suppresses the antitumor activity of natural killer cells and T cells by decreasing mature perforin.
   Cancer Research — DOI. Preprint: bioRxiv 2023.11.20.567708. The source of the perforin findings in section two: the 70 kDa precursor and 60 kDa mature form, suppression of killing at 2–3 mM and complete inhibition at 4–5 mM, loss of rituximab, daratumumab and trastuzumab antibody-dependent killing, loss of CD19 and PD-L1 CAR-NK activity, and recovery of perforin to control levels after sixteen hours of washout. I verified these from the preprint full text, not the journal version; the two may differ. I was unable to confirm the published author list.
2. Ammonia-induced lysosomal and mitochondrial damage causes cell death of effector CD8+ T cells.
   Zhang H, Liu J, Yuan W, et al. Nature Cell Biology 26(11):1892–1902 (2024) — DOI. T cells poisoning themselves, the lysosomal pH shift from 4.5 to 6.2, the absence of CPS1, and the rescue by forced CPS1 expression, glutaminase inhibitors, transporter knockdown and phenylbutyrate. I read the abstract and figure titles plus a detailed commentary on it, not the full paper. The in vitro work used 10 mM ammonium chloride, which is above the tumour range quoted elsewhere in this article — a limitation now flagged in the body text of section two.
3. Microenvironmental ammonia enhances T cell exhaustion in colorectal cancer.
   Bell HN, Huber AK, Singhal R, et al. Cell Metabolism 35(1):134–149.e6 (2022) — DOI. Ammonia stripping T cells of glutathione, and ornithine restoring them. Read. Two results from this paper cut against the enthusiasm: lactulose did not significantly reduce tumour size, and ornithine on its own did not significantly improve survival in their hardest model (P = 0.07) — only the combination with checkpoint blockade did.
4. Metabolic recycling of ammonia via glutamate dehydrogenase supports breast cancer biomass.
   Spinelli JB, Yoon H, Ringel AE, Jeanfavre S, Clish CB, Haigis MC. Science 358(6365):941–946 (2017) — DOI. The 2017 entry on the timeline, and the source of the 0.8–3 mM figure for ammonia measured inside living tumours. Read. Breast models in mice only. The authors concede the enzyme they propose normally runs in the opposite direction in cancer cells, so the mechanism is inferred from the tumour setting rather than demonstrated in it.
5. Impaired nitrogenous waste clearance promotes hepatocellular carcinoma.
   Han X, Shen J, Yan J, et al. Science Advances 12:eaec0766 (9 January 2026) — Open access. The causal experiment in section six, run in both directions: silencing each of four urea cycle enzymes shortened survival and increased tumour burden, while a low-protein diet lowered ammonia and prolonged survival. Read in full. Mice, not people. The authors state plainly that while their results are consistent with a causal contribution, "additional studies would further strengthen the causal link." One figure appears to label ammonia in nanomolar where micromolar is meant.
6. Ammonia signaling network: the intersection of tumor metabolism, epigenetics, and metastasis.
   Yuan Y, Zhang W, Li X, Ge H, Zhu H. Frontiers in Endocrinology 17:1803224 (27 April 2026) — Open access. The April 2026 entry, and the source for ammonia driving the Warburg effect, altering histone and DNA methylation, stabilising the low-oxygen response, and being actively imported by tumour stem cells. Read. A review rather than primary work, so every mechanism in it is quoted at second hand. It is also the source of the honest limit on the environmental-exposure argument: gut-derived ammonia's effect on distant tumours is described as "largely speculative," while its local effect in gastrointestinal tumours is called robust.
7. Ammonia in cancer: dual roles and therapeutic strategies.
   Wei J, Chen Y, Li Q, et al. Cancer Cell International 25:420 (2025) — Open access. The source of the roughly 5 mM turning point, and of the figure of 10 mM above which ammonia kills cancer cells. A review; every concentration is compiled from other people's experiments rather than measured in one system, nearly all of them in cell culture. The authors themselves flag conflicting findings at 10 mM and say the exact threshold remains unclear. For that reason the 5 mM figure is the weakest load-bearing number in the article.
8. Ammonia stimulates SCAP/Insig dissociation and SREBP-1 activation to promote lipogenesis and tumour growth.
   Cheng C, Geng F, Li Z, et al., Guo D. Nature Metabolism 4:575–588 (2022) — DOI. Ammonia binding a defined site on SCAP, with 25-hydroxycholesterol competing for the same site. Abstract and figure titles only; the full text is paywalled. I could confirm the residue D428 — mutating it to alanine prevents ammonia binding — but not the two further residues sometimes cited alongside it, so only D428 is relied on here. The prose in section five has been softened to match what I could actually verify.
9. Microbial metabolite ammonia disrupts TGF-β signaling to promote colon cancer.
   PMID 40311681 (2025) — PMC12155590. The βII-spectrin and SMAD3 account in section five. Located but not read; I have the title and indexing only. This is the newest and least checked of the mechanisms described here, and the reader should weight it accordingly — which is why the prose in section five now says "reported to" rather than "binds."
10. SoMuKT: somatic mutagenesis in kidney tubule is enhanced by specific metabolic pathways.
    Marie Skłodowska-Curie Individual Fellowship, grant agreement 896832, Ospedale San Raffaele, Milan — CORDIS. The kidney proposal referred to at the end of the article. The project describes a "unique mutation pattern" in proximal tubule cells and names three candidate metabolic pathways, with ammonia production listed first. €183,473 over two years; started 1 September 2020 and closed 26 September 2022. The project page describes preliminary data only. I found no published results, which is not evidence of a negative finding.
11. The historical thread: Fuld (1933); Larsen and Manning (1960–61); Bockel, Imler and Stahl (1964); Rezzonico, Gervasini and Secchi (1967).
    The ninety-three-year sequence in section one, including the Milan group's experimental design, the control mean of 14.20 and leukaemic mean of 47.13, and the reanalysis of their published table. The citation count of two is from the Web of Science core collection, queried 12 October 2026; Scopus and Google Scholar give slightly different numbers and none of them changes the point. These are carried from earlier work rather than verified by me directly against the original papers, several of which are not available online. The reanalysis of the 1967 dataset — the correlations near 0.9 and the extrapolation back to the healthy baseline — is a reading of their printed figures, not a claim the authors made.
12. The dendritic cell findings, the 2019 account of ammonia as the tumour's own buffer, the cirrhosis parallel, and the bicarbonate work.
    Dendritic cell dysfunction at 5 mM and the matching defect in cells from cirrhosis patients; the 2019 proposal from a laboratory specialising in inherited urea cycle disease that tumours suppress argininosuccinate synthase so that ammonia accumulates and alkalinises the cell interior; and the animal work showing that oral bicarbonate raises tumour pH and improves checkpoint-inhibitor responses. All are carried from prior work and not independently verified here. The cirrhosis parallel in section four is my own suggestion of a natural experiment, not a claim made in any of the papers cited. The bicarbonate experiments in particular began buffering before the tumours were implanted, which is a long way from treating an established human cancer.

If I wanted to show this hypothesis was wrong, the first thing I would examine is the concentrations. A great many of the striking numbers above were chosen by an experimenter rather than found in a tissue, and the gap between a mechanism that works at a convenient dose and one that operates at the dose a living tissue actually experiences is where a great many promising cancer mechanisms have quietly died.

Last updated: 12 October 2026. Literature searched through 12 October 2026. Prepared with AI assistance.