“When we talk about resistance to chemo and radiation, it’s really just a way of blaming the victim.”
— Sayer Ji, The Truth About Cancer, filmed in Naples, Florida, 2014
Story at a Glance
- “Resistance” is the wrong word. Chemotherapy and radiation do not merely fail against a stubborn subpopulation of cells. Under a growing body of evidence, they help produce it — and the term quietly relocates that failure onto the patient.
- Radiation oncologists found it themselves. A UCLA team showed that radiation reprograms ordinary breast cancer cells into induced cancer stem cells up to 30 times more tumorigenic — and that stem cells reappear even in populations that had been stripped of them first.
- At ASCO 2026, oncology said it out loud. A UCL trial of 4,000+ patients confirmed that the standard of care had been giving chemotherapy to 100 women to benefit 10. The BBC called it a breakthrough. It is a confession.
- We built a pharmacopeia against the wrong model. Taxol is potent against flat monolayers of cells and shows no activity whatsoever against three-dimensional spheroids — the model that contains the stem compartment — even at 10,000-fold higher concentration.
- The industry did not ignore nature’s answer. It patented it. Cyclopamine from the corn lily became vismodegib and glasdegib. Paclitaxel came from the Pacific yew. The debate was never natural versus synthetic.
- The real finding is not “food instead of chemo.” Curcumin appears to sensitize chemotherapy while targeting the compartment chemotherapy enriches. That combination is mechanistically obvious, clinically unbuilt, and unfundable — because no single party can own enough of it.
By Sayer Ji
Disclosure and medical caution: This is educational, not medical advice. Nothing here is a reason to decline, delay, or discontinue the care you be under. Most compounds below rest on preclinical evidence — cell and animal models — not human survival data. Where human data exist, I say so. Several of these compounds interact with chemotherapeutic drugs in both directions. Make treatment decisions with a qualified clinical team, and disclose every supplement you take.
The Word That Blames the Patient
What “resistance” actually describes
For fifty years, when a cancer failed to submit to chemotherapy or radiation, the literature called it resistance. The tumor resisted. The patient’s disease proved resistant.
The phrasing is so ordinary it passes without notice. That is what makes it worth noticing. Resistance implies a stubborn opponent and a sound weapon. It puts the failure in the disease — and, once the word leaves the journals and enters the exam room, in the patient.
The published mechanism describes something close to the opposite.
In 2014 I sat down with Ty Bollinger in Naples, Florida, for The Truth About Cancer. His first question was about chemo and radiation resistance. My answer was blunter than the literature of the day supported.
The argument had three parts. That radiation is ionizing by design and chemotherapy is genotoxic by design — meaning both operate through the exact mechanism by which we classify a substance as carcinogenic. That “resistance” therefore misdescribes what happens, because everyone exposed sustains damage. And that these therapies can shrink the visible tumor while enriching the population beneath it that is actually capable of regenerating the disease.
I made that claim on camera without a citation in hand. Here is what has been published since.
You can watch my full interview .
What the Radiation Oncologists Found
The experiment that should have ended the argument
A tumor is not a uniform mass of equally dangerous cells. It is a hierarchy. A subpopulation well under 1% of tumor mass carries the properties of stemness: self-renewal, the ability to differentiate into the bulk, and a survival apparatus the bulk lacks. Kill 99% and leave that fraction intact, and you have produced an impressive scan and a biological irrelevance.
This is no longer heterodox. A 2025 review in MedComm — Oncology calls cancer stem cells a fundamental barrier to effective therapy, listing their machinery: drug efflux pumps, enhanced DNA repair, metabolic plasticity, ALDH detoxification enzymes.1 Every item is a mechanism for surviving cytotoxic insult.
And in Signal Transduction and Targeted Therapy, a Nature journal, the key point is stated plainly: ionizing radiation upregulates CD133+ cancer stem cells in glioblastoma, and cancer stem cells are enriched in breast cancer following radiation.2 Not “survive.” Enriched.
The effect is now so well established that it has been inverted into a laboratory technique. A 2025 paper in Cancer Letters lists chemotherapy-induced stress, alongside hypoxia and high glucose, as a reliable method for generating cancer stem cells when researchers need more of them.3
Researchers who need cancer stem cells apply chemotherapy to make them. It works well enough to be a protocol.
Evidence Note — UCLA Department of Radiation Oncology
Frank Pajonk’s group reported that radiation, while killing roughly half of all tumor cells per treatment, transforms surviving cells into treatment-resistant breast cancer stem cells with more than a 30-fold increased ability to form tumors.4
A companion study isolated the mechanism: radiation drove pluripotency genes Sox2 and Oct3/4 in non-stem cells. And in the detail that ought to be famous — irradiating a population already stripped of stem cells caused sphere-forming cells to emerge anyway.5
They removed the mother cells. They applied radiation. The mother cells came back, generated fresh from cells that had never been stem cells.
That is not resistance. That is induction.
Read Pajonk honestly: his conclusion was not that radiation should be abandoned, but that preventing this transformation would make radiation more effective. That reading is more damaging than the alternative, not less. The problem was identified inside a comprehensive cancer center, published, and then not built into the standard of care.
Ninety Women to Benefit Ten
May 2026, and the confession reported as a breakthrough
Mechanism can be argued with. A number delivered from a podium at the American Society of Clinical Oncology cannot.
At ASCO 2026, a University College London team presented results from an international trial of more than 4,000 newly diagnosed breast cancer patients across six countries, testing a gene-activity assay that identifies who will not benefit from chemotherapy. More than two-thirds could be safely spared it, with virtually identical survival.6
We used to give chemotherapy to 100 women to benefit 10, knowing that 90 didn’t need it.— Professor David Miles, University College London, quoted by BBC News, May 2026
90%of breast cancer patients given chemotherapy received no benefit from it
The BBC reported it as a triumph of precision medicine. It is an indictment. And the cancer stem cell literature is precisely why it should surprise no one: a therapy optimized against the bulk of a tumor, in a disease driven by a fraction that the therapy enriches, was always going to help a minority while harming nearly everyone.
I took that admission apart in full — including the mammography-to-mastectomy pipeline that delivered those women to the chemotherapy chair — here: “We Gave Chemotherapy to 90 Women to Benefit 10.”
None of this establishes that these therapies never help. They demonstrably do, in defined settings — testicular cancer, many pediatric leukemias, Hodgkin lymphoma, several adjuvant contexts with randomized survival data. Those results are real, and they were bought at enormous cost by people who volunteered.
What it establishes is narrower and worse. The standard of care is optimized for tumor shrinkage, which is not the endpoint patients care about, and the two can move in opposite directions. Response rate measures the bulk. Recurrence is driven by the fraction.
We built an approval pathway, a reimbursement system, and fifty years of practice around measuring the part of the tumor that was never the problem.
Two Dishes, One Drug
How a pharmacopeia gets built against the wrong model
In 2015, researchers publishing in PLoS One tested 6-shogaol — the pungent compound ginger makes when dried or cooked — against breast cancer in two models, running paclitaxel alongside as comparator.7
The first was a monolayer: cells growing flat across the bottom of a dish. This is the classical preclinical screen, and it is how most chemotherapeutic agents earned their reputations.
The second was a spheroid: cells in three dimensions, reproducing tumor architecture — including the stem-like subpopulation.
Taxol performed exactly as advertised in the monolayer. Against the spheroids it showed no activity at 10,000-fold higher concentration.
A correction to my own reporting. I have headlined this finding as ginger being “10,000x stronger than chemo.” That is wrong, and I am fixing it. You cannot compute a potency ratio against an agent with no measurable activity. The accurate statement is stronger than the headline it replaces: taxol failed categorically against the three-dimensional model, at concentrations four orders of magnitude above where the ginger compound worked. Not weaker. Absent.
The monolayer is the bulk. The spheroid is the stem compartment. One model resembles the endpoint we measure. The other resembles the endpoint that kills people.
We got very good at killing flat sheets of cells.
V. The Corn Lily Tells on Everyone
Plant, molecule, patent — three times over
For years I quoted a 2015 review’s conclusion that no drug in clinical use specifically targeted cancer stem cells. That is no longer true, and how it became untrue is the most revealing thing in this subject.
Cyclopamine comes from the corn lily, Veratrum californicum. It was identified in the 1950s when Idaho ranchers noticed ewes grazing on it bore lambs with a single central eye. The teratology was the clue: cyclopamine blocks Hedgehog signaling, which patterns embryos and, in adults, maintains stemness — including the persistence of cancer stem cells.8
Cyclopamine itself was never going to be a drug; its potency and solubility were inadequate. So the industry built patentable analogs. Vismodegib was approved in 2012. Sonidegib followed. In 2018 glasdegib was approved for AML on data showing median overall survival of 8.3 months against 4.3.8
The cancer stem cell hypothesis was validated — not by a dietary study, but by a drug approval. The industry read the science, identified the target, found the lead compound in a wild lily that poisons sheep, and modified the molecule enough to own it.
Then the epilogue. Vismodegib and sonidegib met acquired resistance in clinic — medulloblastoma relapsing within months via a single point mutation in Smoothened.9 One mutation, one escape. A single-target inhibitor against a cell population defined by plasticity gets outmaneuvered, because routing around blockades is what these cells are for.
Which is the argument for eight nodes instead of one
A review of curcumin’s asymmetrical effects on cancer versus normal stem cells catalogued eight distinct mechanisms at once:10
- IL-6 — downregulated; the cytokine linking chronic inflammation to malignant progression
- IL-8 — downregulated; released after tumor cell death, it stimulates stem cells to regrow the tumor
- IL-1β — downregulated
- CXCR1 / CXCR2 — binding reduced, blocking not just cytokine release but receptor engagement
- Wnt/β-catenin, Notch, Hedgehog — all three self-renewal pathways modulated
- FAK/AKT/FOXO3A — inhibited
Glasdegib blocks Smoothened, one node in Hedgehog, and one point mutation ends its usefulness. Curcumin touches Hedgehog as one of eight simultaneous actions.
There is no single mutation that routes around eight nodes at once. That is not a claim that curcumin is more powerful — at any individual target it is certainly weaker, and glasdegib has survival data curcumin does not. It is a claim about escape architecture, and it is the entire pharmacological case for food.
It is testable. It has largely not been tested.
The Twenty-Five, Sorted Honestly
Because a toxic veterinary antibiotic and a culinary spice do not belong in the same list
The 2015 Anticancer Research review identified 25 natural compounds with reported anti-cancer-stem-cell activity.11 I previously published them alphabetically, flat, under a headline about foods. That was a mistake, and I am fixing it.
Tier 1 — dietary, with human pharmacokinetic and clinical trial data (cancer stem cell effects still preclinical): EGCG (green tea) · curcumin (turmeric) · sulforaphane and isothiocyanates (cruciferous vegetables) · quercetin (onion, capers, apple) · resveratrol (grapes, berries) · genistein (soy, red clover) · lycopene (tomato) · vitamin D3 (sunlight, cod liver oil) · silibinin (milk thistle) · piperine (black pepper)
Tier 2 — dietary and botanical, preclinical evidence only: 6-gingerol and 6-shogaol (ginger) · baicalein (Chinese skullcap) · delphinidin (blueberry) · guggulsterone (myrrh) · linalool (mint, basil) · parthenolide (feverfew) · perillyl alcohol · ursolic acid (thyme, oregano, apple peel) · withaferin A (ashwagandha) · platycodon saponin · psoralidin
Tier 3 — not foods. Do not consume. On the list only because they prove such compounds exist in nature: cyclopamine (corn lily, teratogenic) · gossypol (cottonseed, toxic) · salinomycin — a veterinary ionophore antibiotic used in poultry feed, one of the most potent anti-CSC agents known and acutely neurotoxic in humans, with documented severe poisonings. This appeared in my original list with no warning. That was an error.
And one compound removed entirely
Evidence Note — Why β-carotene came off the list
Two large randomized trials tested β-carotene supplementation for lung cancer prevention. ATBC randomized 29,133 Finnish male smokers and found a 16% increase in lung cancer incidence and 8% increase in overall mortality at 20 mg daily.12 CARET found 28% greater lung cancer incidence and 17% more deaths from all causes.13 Both were halted early. A 2022 USPSTF review of six trials found a 20% increased risk and recommended against it.14
Now hold that against the observational epidemiology, which consistently finds that people who eat more β-carotene-rich food get less lung cancer.
Same molecule. Opposite results. The investigators’ own explanation: the harm tracks the pharmacologic dose and the resulting supra-physiological serum concentration.
I have spent twenty years arguing that a nutrient is not an input but a signal — that the body reads food as information, in matrix, in ratio, at physiological concentration, alongside the hundreds of compounds that arrived with it.
β-carotene is the cleanest demonstration of that thesis anyone has produced. It was produced by accident, by researchers trying to prove the opposite.
The carrot was never the active ingredient. The carrot was the context.
Which is the warning label on this entire list. Extract any compound above, concentrate it forty-fold, give it as monotherapy at pharmacologic dose, and you have not intensified the food. You have built a drug — and inherited every liability drugs carry.
The Choice They Never Offered
What happens when watchful waiting is actually put to the test
I once wrote that chemotherapy, radiation, and even surgery may no longer be justified as first-line standard of care.
I have received a lot of pushback for that, but I still believe it to be true.
The question is not whether cytotoxic therapy ever works. It is whether the men and women who receive it are ever shown the numbers that would let them decline. And when that question has been put to a randomized trial, the answer has been the same every time.
Evidence Note — The ProtecT Trial, New England Journal of Medicine
Between 1999 and 2009, 1,643 men with localized prostate cancer were randomized to active monitoring, surgery, or radiotherapy. Not a low-risk cherry-pick: 24% had intermediate-risk disease and 10% had high-risk disease.
At a median 15 years of follow-up, death from prostate cancer was 3.1% under active monitoring, 2.2% after prostatectomy, and 2.9% after radiotherapy.15
P = 0.53. No significant difference. Fifteen years.
Surgery and radiotherapy halved metastasis, local progression, and long-term androgen deprivation. None of it translated into a single additional year of life.
Now read what the trial investigators wrote in their own conclusion, in the New England Journal of Medicine:
Depending on the extent of side effects associated with early radical treatments, more aggressive therapy can result in more harm than good.— Hamdy et al., ProtecT Study Group, NEJM 202315
That is not my sentence. That is Oxford, in the most prestigious journal in medicine, reporting the results of a fifteen-year randomized trial they ran themselves.
And there is a detail inside ProtecT that ought to end the conversation. A quarter of the men assigned to active monitoring were alive at fifteen years having received no treatment at all.15 Not delayed treatment. None. Under the standard of care, every one of those men would have been cut or irradiated, and every one of them would have been counted afterward as a life the treatment saved.
The men who were treated paid for it in incontinence, erectile dysfunction, and bowel injury at rates the monitoring group did not experience.16 They bought those harms with no survival in return.
The same result, in the other direction
This is not an isolated finding. The COMET trial, presented at San Antonio, found that women with DCIS who chose active monitoring were no more likely to develop invasive cancer than women who underwent surgery.17 GreenMedInfo argued that DCIS should not be classified as cancer starting in 2008. It took sixteen years and a randomized trial for the position to become respectable, and nobody has apologized to the women treated in the interval.
And at ASCO 2026, as we saw in Section III, the profession conceded that ninety women in a hundred received chemotherapy that did nothing for them.
Three cancers. Three randomized answers. Each one says the same thing: a large fraction of the people being treated were never going to benefit, and the system had no mechanism for telling them so.
What the harm column actually contains
The other half of the ledger is worse documented than it should be, because for most of the history of oncology nobody counted it.
When England finally did — the first national dataset of its kind in any country — it tracked deaths within 30 days of systemic anticancer therapy, on the explicit premise that early mortality is an indicator of avoidable harm from treatment. Among breast and lung cancer patients treated in 2014, 1,384 died within a month of receiving chemotherapy given with either curative or palliative intent.18
The trust-level analysis found hospitals sitting as statistical outliers beyond the 99.8% confidence boundary. Same drugs, same protocols, wildly different rates of patients dying within weeks of treatment.18
And in the accompanying commentary, the acknowledgment that matters most: because these drugs have a narrow therapeutic index, some degree of harm is inevitable at any effective dose — and real-world harm runs worse than the trials report. One Canadian analysis found hospital admissions during routine lung cancer treatment running more than three times higher than the rates published in the phase 3 trials that won the drugs their approvals.19
Sit with that. The toxicity figures a patient is quoted at diagnosis come from trial populations that were younger, fitter, and more closely monitored than she is. The number she is given understates by a factor of three what will actually happen to her.
And even the harm reduction was never built
Here is the part that turns negligence into something harder to name.
The mechanism does not only indict cytotoxic therapy. It points at the fix, and the fix has been sitting in the literature for a decade.
The Cancer Letters review on curcumin and cancer stem cells reports that curcumin synergizes with conventional chemotherapy — making it more effective, and in some cases less harmful.20 Chemotherapy hits the bulk and enriches the stem compartment. Curcumin targets the stem compartment and sensitizes the bulk. On the radiation side, radiation-induced stemness was blocked outright in mice by disulfiram, a cheap, generic, FDA-approved drug that has been on the market since 1951.21
Lower doses. Fewer secondary malignancies. Less of the 30-day mortality that England had to build a national dataset to see.
None of it has been built. Not one adequately powered trial. And the reason is not that the hypothesis failed — the hypothesis was never tested. There is no sponsor for a trial whose active ingredients are a spice and a seventy-year-old generic, because there is nothing at the end of it to own.
So the precise version of my claim is this, and I will stand behind every clause of it:
For a substantial and knowable fraction of the people receiving it, cytotoxic therapy delivers no survival benefit, imposes permanent harm, and is chosen on their behalf without the randomized evidence ever being placed in front of them. Where that evidence exists, it has repeatedly favored doing less. And the interventions that would make the treatment safer for everyone else have gone unfunded for a decade because no one can patent them.
That is not a softer claim than the one I made before. It is the same claim with the evidence attached, and it is considerably harder to dismiss.
The failure here is not only that the treatments harm. It is that the choice was never offered. A man handed a prostate cancer diagnosis is rarely told that fifteen years of randomized data show no mortality difference if he does nothing. A woman with DCIS is rarely told that monitoring performed as well as surgery. Ninety women in a hundred were never told that the chemotherapy would do nothing for them, because until 2026 nobody had built the test that would say so — and nobody was looking for it.
That is not informed consent. It is compliance under manufactured consensus.
DO NOTHING: The Peer-Reviewed Prescription BIG PHARMA Doesn’t Want You To Know
Please comment on and share this Substack post on X: https://x.com/sayerjigmi/status/1950581120212353351
Read full story
The Regenerative Turn
A cell that has lost its context is not the same as a cell that must be destroyed
One more finding, and it is where this has been heading.
Among the explanations offered for curcumin’s selectivity is that it may induce cancer stem cells to differentiate into more benign cell types rather than killing them outright.10 That is a categorically different therapeutic logic.
A cancer stem cell is not an alien invader. It is a cell of your own body running an ancient, conserved developmental program — self-renewal, plasticity, stress resistance, the capacity to rebuild tissue from almost nothing — in the wrong context, without the regulatory signals that tell it when to stop. Every property that makes it lethal is a property that makes wound healing possible.
The problem is not that the cell is defective. The problem is that it has lost its context.
That is not a metaphor. Michael Levin’s work at Tufts has shown that bioelectric patterns, not genes alone, carry the instructions determining cellular identity — and that manipulating those patterns can normalize cells that have departed from them. If identity is maintained by a field the cell is embedded in, then a cell that has fallen out of that field has a recoverable problem, not only a destroyable one.
It is also why spontaneous remission — thousands of documented cases, largely unstudied because they embarrass the model — is theoretically important rather than merely curious. Under irreversible genetic mutation, remission should be impossible. Under adaptive program plus lost context, it is exactly what you would predict when the context is restored.
I develop that whole architecture — cancer as metabolic adaptation rather than genetic accident, terrain over target, and the overdiagnosis industry built on confusing the two — in The Cancer Deception, and at book length in Chapter 4 of REGENERATE.
Cytotoxic therapy asks the first question with great force and cannot ask the second at all. A genotoxic agent has no vocabulary for context — and the reason is historical, not scientific. Chemotherapy descends directly from chemical warfare research, from the observation that mustard gas destroyed lymphoid tissue and suppressed bone marrow. The founding insight was that a poison might kill the tumor before it killed the patient. Every term still in use carries that inheritance.
This is not an argument for abandoning it where it works. It is an argument for recognizing what a medicine organized exclusively around destruction will keep producing: shrinking tumors, enriched stem populations, recurrences reported as resistance, ninety women in a hundred treated for the benefit of ten, and patients quietly blamed for the failure of a frame they never chose.
Two dishes. One drug. The model where it fails is the one that resembles the disease.
The word was always doing the work. Resistance put the fault in the patient. The mechanism puts it in the model.
Your food is not fuel but information. Your biology is not defective but contextual. What you eat, how you sleep, what light you see, and what you carry emotionally are not adjacent to the science of cancer — they are load-bearing within it.
The body has not forgotten how to regulate itself. It has been asked to do so under conditions no living system was built to normalize, and then blamed when it could not.
Regeneration begins when we stop treating that failure as the patient’s, and start treating it as the frame’s.
