What the Slide Cannot Say

A pathologist looks at a slide and sees unmistakable cancer. The cells are the wrong size. A test almost nobody orders identifies them as a tapeworm's. This episode walks the six steps between a lump and a diagnosis — fixation, processing, sectioning, staining, and a human being at a microscope — and asks what that pipeline was built to find, what it was not, and whether the iodine everyone worries about is anywhere near the top of the list.

Show notes

A man with HIV comes into a hospital in Medellín, Colombia. He has enlarged lymph nodes in his neck and lesions in his lungs. They take biopsies — small pieces of the tissue — and send them to pathology.

The pathologist puts the slide under the microscope and sees cancer. Not a borderline case. Nests of cells packed together, all identical to one another, with no recognisable structure, pushing into the surrounding tissue. In the authors' own description: the morphology and the invasive behaviour were characteristic of cancer.

Except for one thing. The cells were too small. Their small size, the paper says, suggested a nonhuman origin.1

Somebody made a decision at that point which is the entire reason we know about this. Instead of running another antibody test to work out which kind of human cancer it was, they ran a different test altogether — a broad eukaryotic PCR. I will explain exactly what that means in a moment, but the short version is that it does not ask "which human cancer is this?" It asks "what organism is this?"

The answer came back Hymenolepis nana. The dwarf tapeworm. The most common tapeworm in human beings — a population survey of 14,761 children in northern Peru found it in 7.6% of them, and the authors call that a likely underestimate because only one stool sample per child was examined23 — and a parasite considered essentially harmless.

They confirmed it two more ways. Antibody staining lit the cells up as tapeworm tissue. Genetic probes bound to tapeworm sequence in place, inside the cells, on the slide. And then they sequenced the whole thing and found structural changes in the parasite's genome — changes, in the authors' phrasing, "compatible with mutations described in cancer."1

The tapeworm's own cells had turned malignant. And then that malignancy had spread through a human being.

That paper appeared in the New England Journal of Medicine in November 2015, and it remains the only case of its kind ever described.

Let me be extremely careful about what it does and does not show, because this is the episode where being careless would be worst.

It does not show that parasites cause human cancer. It does not show that this can happen in a person with a working immune system — the patient had untreated HIV and was profoundly immunosuppressed, and in an intact immune system this almost certainly does not occur. It is one case.

What it shows is narrower, and I think stranger and more important. It shows that a competent pathologist, looking at human tissue through a microscope, can see unambiguous cancer and be looking at a different species entirely. And that the only reason anyone found out was that somebody ordered a test that is almost never ordered.

Which raises the question this episode is actually about. Not "do parasites cause cancer" — we will get to that, and the answer is yes, three of them, definitively. The question is: if there were more of this, would we know?

The three that are certain

Start where the ground is solid, because it is much more solid than most physicians realise.

The International Agency for Research on Cancer — IARC, the World Health Organization's cancer arm — sorts substances and organisms into groups according to how good the evidence is that they cause cancer in humans. Group 1 means carcinogenic to humans, and the evidence is beyond serious argument. That group contains tobacco smoke, asbestos, ionising radiation, hepatitis B and C, human papillomavirus.

It also contains three parasites.2

Schistosoma haematobium, a blood fluke that causes bladder cancer. Opisthorchis viverrini and Clonorchis sinensis, two liver flukes that cause bile duct cancer. Classified in 1994, 2009 and 1994 respectively, and never seriously challenged since.

So the general question — can a parasite cause a human cancer — was settled before most of today's doctors finished medical school. What is in dispute is everything else: how many more, in which organs, and whether anybody is looking.

Let me explain what a fluke is, because the word means nothing to most people and the biology is genuinely remarkable.

Flukes are flatworms — soft, leaf-shaped animals, usually a few millimetres to a few centimetres long, with a sucker at one end. They are not microscopic. They are animals you could see and pick up. And they have life cycles that require two or three completely different hosts in sequence, which is why they are geographically bound: break any link in the chain and the parasite cannot complete its circuit.

Schistosoma haematobium works like this. Eggs leave a human in urine and reach fresh water. They hatch into a swimming larva that must find a particular species of freshwater snail. Inside the snail it multiplies enormously and eventually releases a second larval form, the cercaria, which swims free and burrows directly through the skin of anyone standing in that water. It enters the bloodstream, matures, and the adult worms pair up — the female lives in a groove along the male's body — in the veins that drain the bladder.

And there they stay, for years, releasing eggs that have to force their way out through the bladder wall to reach the urine and start again. Each egg carries a spine. Each egg crossing the bladder lining does damage. Many do not make it out and are walled off in the tissue instead.

Now: the bladder cancer that follows is not the bladder cancer a Chicago urologist expects. In the West, bladder cancer is overwhelmingly urothelial carcinoma — arising from the specialised lining cells of the urinary tract, usually in older smokers. Schistosomiasis-associated bladder cancer is typically squamous cell carcinoma,3 which means it arises from cells that have transformed into a flat, skin-like type they are not supposed to be — a change the body makes in response to chronic irritation, called squamous metaplasia.

Different cell of origin, different behaviour, different age group, same organ. And a clinician who has only ever seen the smoking kind will not have the other one in mind.

Here is what the definitive clinical review says about the mechanism, and I want you to hear it exactly: several carcinogenic pathways have been proposed, "but the exact mechanism(s) are not defined yet."3

Read that again in light of the Group 1 classification. A parasite can be the most certain category of human carcinogen the WHO recognises, alongside asbestos, and nobody knows how it does it. Group 1 is a statement about the epidemiology, not about the biology. The epidemiology was overwhelming; the mechanism is still open.

Now the contrast that teaches the whole system, and it is the single most useful thing in this episode.

Schistosoma mansoni is haematobium's close relative. Same genus. Same basic life cycle, different snail, different destination — it settles in the veins draining the intestine and the eggs end up in the liver.

IARC classifies S. mansoni as Group 3: not classifiable as to its carcinogenicity in humans.4

Here is what Group 3 means, and almost everyone gets this wrong in both directions. It does not mean "shown to be safe." There is a group for that — Group 4, probably NOT carcinogenic — and in the entire history of IARC it has held a grand total of one substance, which was later moved back out of it. Proving that something does not cause cancer is so hard that the category is effectively unused. Group 3 means: we looked, and there is not enough evidence to classify it either way. It is a verdict on the literature, not on the organism.

And for S. mansoni, what exists is a real body of suggestive material — a multiplicity of case reports, animal models, and cell culture work pointing at liver and colorectal cancer, with proposed biomolecular mechanisms.4 Enough to publish a review titled with a question mark. Not enough to classify.

Two organisms in one genus. One is as established a human carcinogen as asbestos. The other has been in the not-enough-evidence box for decades. That asymmetry — inside a single genus, with the same tools available — is the shape of this entire field.

The liver flukes, and the experiment that proved it

Two of the three Group 1 parasites are liver flukes, and their story is the strongest causal case in parasite oncology. It is worth the detail because it shows what proof actually looks like here.

Opisthorchis viverrini infects close to ten million people in Southeast Asia. You get it from dinner: raw or fermented freshwater fish carrying the infective larval stage. The larvae travel up the bile ducts — the plumbing that carries bile from the liver to the intestine — attach to the lining, and feed there.

For ten to thirty years.5

That is the number that reframes everything. This is not an infection in the sense of a week of being unwell. It is a multi-decade tenancy. And for most of it there are no symptoms at all — the disease is described as long asymptomatic, only declaring itself as chronic inflammation of the bile ducts once the parasite load gets high enough.

The cancer at the end of it is cholangiocarcinoma — cancer of the bile duct lining. Khon Kaen province in northeast Thailand has the highest incidence in the world. It grows silently, especially inside the liver, is usually found late, is usually inoperable by then, and five-year survival is under 5%.5

The mechanism is three things happening at once, and they should be understood as three because each one on its own would probably not be enough.

One: physical damage. An animal is attached to the lining of your bile duct, feeding on it, for decades. The lining responds by proliferating — replacing lost cells — and every round of cell division is another chance for a copying error.

Two: chronic inflammation and oxidative stress. The immune system attacks something it cannot remove. The weapons it uses are largely chemical — reactive oxygen species, which are unstable oxygen-containing molecules that damage whatever they touch, including DNA. In a short infection that is a reasonable trade. Over thirty years it is a slow bombardment of the exact tissue that is already dividing faster than normal.

Three, and this is the one people find hardest to believe: the parasite secretes growth factors into your tissue. Not as an attack. Because a fluke living on a surface it is eroding benefits from that surface repairing itself. Its own proteins drive host cell division and suppress host cell death.5

And there is a fourth ingredient that is not the parasite at all: nitrosamines, a class of chemical carcinogen, present in the fermented raw-fish dishes that carry the infection in the first place. The meal delivers the parasite and the chemical together.

Now the experiment. This is the part I would put in a textbook.

In 2019 a team spanning Khon Kaen, George Washington University, James Cook University and Melbourne did something that had never been done in a parasitic flatworm: they used CRISPR gene editing on the fluke.6

CRISPR is a technique that lets you cut DNA at a chosen sequence. The cell's repair machinery then patches the break clumsily, which usually breaks the gene. It is the closest thing biology has to a delete key.

The gene they deleted was granulin — one of those secreted growth factors, the fluke's own protein that drives host cells to divide.

Then they infected hamsters with the edited flukes and waited.

The flukes were fine. They colonised the bile ducts and developed into adults normally, which matters, because it means the experiment is not just "sick worms cause less disease." The parasite did its whole life cycle.

But the pathology was markedly reduced. The bile duct overgrowth and the fibrosis — the scarring that precedes cholangiocarcinoma — were both attenuated.6

Delete one parasite gene. The pre-cancer largely does not happen.

That is not an association. That is a molecule, named, removed, with the disease following it out the door. If you want to know what real proof looks like in this field, it looks like the granulin knockout.

Clonorchis sinensis, the other Group 1 liver fluke, works the same way and is endemic in Korea, China, Taiwan, Vietnam and the Russian far east. Here is the scale, which nobody in the West registers: in Korea's 2012 national survey of intestinal parasites, C. sinensis had the highest prevalence of any parasite in the general population — 1.86% — and transmission is ongoing.2

Almost two per cent of a wealthy, modern, universally-insured country carries a Group 1 human carcinogen in their bile ducts. Not in 1950. In the 2012 survey.

One more thing about the liver flukes, because it is the detail that complicates every easy conclusion anyone wants to draw from this episode — including mine.

There is a drug that cures opisthorchiasis. Praziquantel works, it is cheap, it is a single course, and it has been deployed in mass deworming campaigns across the endemic region for decades.

The campaigns failed.5

Not because the drug does not work. Because people are reinfected. The transmission runs through raw and fermented fish dishes that are central to the regional cuisine, the sanitation that would break the cycle is not there, and health education has not changed a food culture. So the same person is treated, cured, and reinfected, over and over, across a lifetime.

And here is the line that should stop anyone who thinks the answer to all of this is obviously "treat the parasite": the review raises the possibility that repeating cure-and-reinfection cycles may themselves promote carcinogenesis.5

Think about why that could be true. Every time the parasite is killed, the damaged bile duct lining sets about repairing itself — a burst of cell division in tissue that has been chronically inflamed for years. Every burst of division is another set of chances to copy the DNA wrong. Then the parasite comes back and the whole cycle restarts.

It is a hypothesis, not an established finding, and I want to flag it as such. But it points at something that matters far beyond flukes: an intervention aimed at a cause is not automatically a treatment for the consequence. Removing the parasite from a bile duct that has been remodelling for twenty years does not put the bile duct back. And doing it repeatedly may not be neutral.

Hold that thought when we get to ivermectin.

And then the fourth fluke, which is the one that should make you uncomfortable.

Opisthorchis felineus is endemic across the Russian Federation, particularly Siberia, and parts of Europe. Infestation reaches 80% of the population in endemic areas.

In hamsters, infection with O. felineus alone produced hyperplastic and dysplastic changes in the bile ducts — recognised precancerous states. Infection combined with a nitrosamine produced cholangiocarcinoma in every single animal by week eighteen. The authors argued explicitly that their data should be used to assign the organism an IARC category.7

It still has none.

That is not because it has been examined and cleared. It is because the human epidemiology has not been done. A positive animal model, an 80%-infested population across a landmass, and no category.

How a cancer is actually diagnosed

Everything from here depends on understanding what happens between a lump and a diagnosis, and almost nobody outside medicine has ever been told. So we are going to walk the whole pipeline, slowly, and I am going to define every term as it arrives. It takes about fifteen minutes and it is the most useful fifteen minutes in this episode, because by the end you will be able to work out for yourself what this process can and cannot find.

There are six steps, plus two more that happen only sometimes.8

Before we walk the pipeline, one definition, because the whole thing rests on it and it is almost never said plainly.

Cancer is not a substance. It is a behaviour.

There is no molecule you can test for that means cancer, and there is no single gene. What there is, is a set of behaviours that a cell has stopped performing correctly. Normal cells divide when they are told to and stop when they are told to. They stay where they belong. They respect the boundary membranes that separate one tissue compartment from another. And when they are too damaged to be useful, they kill themselves — a genuinely astonishing process called apoptosis, in which a cell dismantles itself tidily on instruction.

A cancer cell has lost some or all of that. It divides without being told to. It ignores the signal to stop. It does not kill itself when it should. It crosses boundaries it should respect. And eventually some of its descendants travel somewhere else entirely and start again, which is what metastasis means.

Now — and this is the part that explains everything about the pipeline — those behaviours leave visible marks.

A cell dividing too often has a bigger nucleus relative to the rest of it, because the nucleus holds the DNA and it is being copied constantly. A population dividing chaotically produces nuclei of many different sizes and shapes, which pathologists call pleomorphism — many forms. Cells caught in the act of dividing show up as mitotic figures, and you can count them. Cells that have lost their instructions stop building the structures they were supposed to build, so glands stop looking like glands. And a cell that has crossed a boundary can be seen on the wrong side of it.

That is why the diagnosis of cancer is a morphological diagnosis — made by looking at shape. Not by a blood test, not by a scan, and not by a gene panel. A scan tells you something is there. A biopsy tells you what it is. And the reason two dyes from the 1870s are still sufficient is that the behaviour we care about happens to be visible in shape and colour, and a trained human eye is extraordinarily good at reading it.

Which also tells you the shape of the blind spot. A system optimised to recognise abnormal human cells by their form is not thereby optimised to recognise a different organism. Those are not the same task, and being excellent at the first does not make you even competent at the second.

Step one: something is removed from a person.

There are three ways, and the difference between them matters enormously later.

A fine needle aspiration uses a thin needle to suck out loose cells. You get cells, not tissue — no architecture, just a smear.

A core needle biopsy uses a wider, hollow needle with a cutting edge to take an intact cylinder of tissue, roughly a millimetre across and one to two centimetres long. This is what most breast, prostate, liver and lymph node biopsies are. You get architecture, from a very small piece.

A resection is surgery — the whole lump, or the whole organ. This is what you get after a diagnosis, usually, not before it.

Hold onto the core needle, because we are going to do arithmetic on it later.

There is a step between the operating theatre and the formalin that I skipped, and I should not have, because it is where more tissue is discarded than at any other point.

When a whole organ comes out — a colon, a uterus, a lung lobe — it does not go into a cassette whole. It goes to a bench, and a pathologist or pathologists' assistant does what is called grossing: examining it with the naked eye, describing it, measuring it, slicing it, and then choosing which pieces to submit for processing. A handful of blocks, each a couple of centimetres square, out of an entire organ.

Those choices are governed by protocols, and the protocols are excellent — they specify the tumour, the margins, the nodes, representative normal tissue. They exist so that no pathologist has to rely on memory for what matters in a colon cancer resection.

And they were written to answer the same question as everything else. A protocol tells you to sample the tumour and the margins because that is what staging requires. Nothing in it says "and take a block from anywhere unusual in case something is living in there."

So by the time the tissue reaches formalin, a human being has already decided, for good reasons, that the great majority of the organ will never be looked at under a microscope by anyone. That decision is invisible in the final report. The report describes what was on the slides.

Step two: fixation. This is the step that governs everything downstream.

The moment tissue leaves the body it starts to destroy itself. Enzymes that were safely locked inside compartments spill out and digest the cell from the inside — a process called autolysis, self-eating. Within hours, untreated tissue is mush.

So the tissue goes straight into a jar of formalin: a roughly 10% solution of formaldehyde, buffered so it does not turn acidic. Formaldehyde is a very small, very reactive molecule, and what it does is cross-link — it forms chemical bridges between proteins, and between proteins and nucleic acids, until the entire tissue is one enormous interlinked mesh. Nothing can move any more. Nothing can be digested. The architecture is frozen exactly as it was at the moment of immersion.

Understand what that means for our question. Formalin does not merely kill everything in the specimen — it kills everything and then holds it in place. A parasite in that tissue is dead within minutes and perfectly preserved indefinitely. Both halves of that sentence matter.

Step three: processing. Wax and water do not mix, and the tissue is about 70% water, so before it can be embedded in wax the water has to come out.

It is done in stages so the tissue does not shrivel. First dehydration: the specimen goes through a graded series of alcohols, 70%, 90%, then absolute, each one pulling out a little more water. Then clearing: the alcohol is replaced by a solvent that will mix with both alcohol and wax — traditionally xylene, which is also the step that makes the tissue translucent, hence the name. Then infiltration: the xylene is replaced by molten paraffin wax, which seeps into every space the water used to occupy.8

This takes many hours and is usually done overnight by a machine. It is also why pathology results take days rather than minutes, and why formaldehyde and xylene — both of which the researchers who built a replacement describe as causing "severe health and safety issues for humans and the environment" — are still in every hospital laboratory in the world.8

Step four: embedding and sectioning.

The wax-filled tissue is set into a small block and chilled until the wax is hard. The block goes onto a microtome, which is a very precise guillotine — a fixed blade and a stage that advances by a set distance with each stroke.

The set distance is typically four micrometres.

A micrometre is a thousandth of a millimetre. A human cell is somewhere between ten and thirty micrometres across. So a standard section is a fraction of the depth of a single cell. The slide you look at is not a picture of the tissue. It is one arbitrary plane through it, thinner than the things it contains.

This is the single most important fact about histopathology and it is the one laypeople never hear. Everything a pathologist knows about your tumour, they know from a handful of arbitrary planes a few millionths of a metre thick.

Step five: staining.

The section on the slide is still full of wax and is completely colourless. So the wax is dissolved off, water is put back in — the whole dehydration run in reverse — and then the section is dipped in two dyes.

Haematoxylin comes from the heartwood of a Central American logwood tree and has been used since the nineteenth century. It binds to acidic structures — which in a cell means nucleic acids, so DNA and RNA — and turns them blue-purple. Nuclei go blue.

Eosin is a synthetic dye named after Eos, the Greek goddess of the dawn. It binds basic structures — most proteins — and turns them pink. Cytoplasm and connective tissue go pink.

Two dyes. Pink and blue. Essentially unchanged in principle since the 1870s. And almost every cancer diagnosis made anywhere on earth is made on them.

Step six: a person looks at it.

This is the step that gets skipped in every explanation and it is the one that does the work. A pathologist puts the slide on a microscope and compares what they see with a catalogue built up over years of training and practice. Are the nuclei too big relative to the cytoplasm? Are they irregular, or all different sizes? Are there mitotic figures — cells caught mid-division — and how many? Is the normal architecture of the tissue preserved, disrupted, or gone? Are the cells crossing a boundary they should respect?

It is pattern recognition, performed by a human being, and it is astonishingly good. I want to say that clearly before anything critical, because the rest of this episode is going to examine the limits of a system that is one of the genuine triumphs of modern medicine. A skilled pathologist can name a tumour's organ of origin, its type, its grade and often its likely behaviour from pink and blue dye on a four-micrometre section. That is remarkable and it saves lives every single day.

The criteria they are judging a slide against have been written down: cellular detail, tissue arrangement, tissue integrity, stain uptake, and visual distinction of tissue structure under light microscopy.8

Notice what every single one of those criteria is about. Human tissue architecture. Not one of them asks whether something non-human is present.

Then there are the two steps that happen only if the H&E raises a specific question.

Immunohistochemistry. Antibodies — proteins that bind one specific target — are manufactured against particular human molecules, tagged with something that produces a visible colour, and washed over the section. If the target protein is there, that part of the slide goes brown. This is how a pathologist distinguishes a lymphoma from a carcinoma, or works out where a metastasis came from.

The word to hold onto is specific. An antibody panel is a list of questions, each one asking "is this particular human protein here?" A panel cannot return "something unexpected". It returns negative, and a negative panel means none of the things I asked about, not nothing.

Molecular testing. DNA or RNA is extracted from the block and examined for mutations that direct treatment — EGFR in lung cancer, BRCA in breast and ovarian, KRAS in colon. This is done with primers: short pieces of synthetic DNA designed to match and amplify one target sequence. Or with sequencing panels covering a defined list of human genes.

Again: designed to match. A primer built for a human gene will not amplify a tapeworm. It is not that it fails to notice the tapeworm. It is that it was never pointed at it.

So now you can answer the question yourself. Six steps, plus two, every one of them exquisitely designed to answer a single question — what kind of human cell is this, and is it behaving badly? — and a non-human organism falls outside that question at every stage.

Not because anything destroys the evidence. Because nothing in the workflow ever asks.

The iodine question

There is a specific hypothesis circulating about why parasites are not found in cancer specimens, and it deserves to be taken seriously and examined properly, which means giving it its strongest form first.

It goes like this. Iodine is a powerful antiseptic. Iodine kills parasites. Iodine touches cancer specimens. Therefore iodine may be destroying the evidence before anyone can look at it.

Let me establish both halves, because both halves are true.

Iodine does touch cancer specimens, by at least three documented routes before a pathologist ever sees the tissue.

First, surgical antisepsis. Povidone-iodine — the brown solution everyone has seen painted on skin before an operation — is standard skin preparation at 7.5 to 10% concentration, and it is also used to irrigate body cavities and wounds during surgery. Any resected specimen has been in contact with it.

Second, chromoendoscopy. Lugol's iodine — a solution of iodine and potassium iodide — is sprayed directly onto the lining of the oesophagus during endoscopy to find early cancer. Healthy squamous lining is rich in glycogen, and iodine stains glycogen dark brown; abnormal or dysplastic areas have used up their glycogen and stay pale. The endoscopist then biopsies the pale patches. So in that procedure, iodine is applied to living tissue moments before the biopsy is taken from it.

Third, histology reagents. Lugol's iodine has a long-standing role in the laboratory itself, historically to strip mercury pigment out of tissue fixed in mercury-containing fixatives.

And iodine is a genuinely formidable antiparasitic — considerably better than most people assume.

In 2022 a German and Hungarian group tested eight off-label anti-amoebic drugs against Acanthamoeba, a free-living amoeba that causes a sight-threatening corneal infection. Acanthamoeba forms cysts — dormant, armoured, notoriously drug-resistant. The test was whether a drug could stop cysts from waking up and becoming active trophozoites.

Of eight agents across four isolates and five biological replicates, exactly one completely suppressed reactivation in every case: povidone iodine at 1%. Chlorhexidine failed. Miltefosine failed. Dibromopropamidine failed.9

One per cent. Surgical prep is roughly ten times that.

And now the finding that turns this from speculation into a real question, because a whole specialty went looking at exactly this mechanism and found it.

In 2019 a group at Wolverhampton and Moorfields Eye Hospital tested topical ophthalmic agents against Acanthamoeba and concluded that povidone iodine's antiamoebic effect was superior to the diamidine drugs actually used to treat the infection. Then they wrote this, and I am quoting it exactly:

Substances applied to tissue before the sample is taken can kill the organism and produce a false negative diagnostic test. That is not a hypothesis. That is a peer-reviewed clinical warning, published seven years ago, in a specialty that thought to check.

So the mechanism is real. Anybody who says the idea is nonsense has not read this literature.

Now the turn, and it is the honest heart of this whole segment.

Read what kind of test they meant. The false negative is in a viability test — culture. You take the sample, you put it on a plate, and you see whether anything grows. A dead organism does not grow. So if the antiseptic killed it between the eye and the plate, the plate is blank and the patient is told they do not have an amoeba.

But histopathology does not culture anything.

Go back to step two of the pipeline. The specimen goes into formalin, and formalin kills everything in it within minutes — and then preserves it. That is the entire purpose of formalin. Every cancer specimen ever examined has been sterilised by fixation before anyone looked at it, and that has never stopped anyone from seeing bacteria, fungi or parasites on a slide, because morphology survives death.

A dead fluke in a paraffin section still looks exactly like a fluke. A dead schistosome egg is still an oval structure about 150 micrometres long with a spine on it. Killing an organism and hiding an organism are different operations. Iodine performs the first one superbly. It has no particular power to do the second.

And there is a fact that cuts directly against the simple version of the story, which is why I will not let it be told without this attached.

In parasitology, iodine mount microscopy is a routine method for finding parasites. It sits in the standard workup alongside the direct wet mount: you put the specimen on a slide, add iodine, and the iodine kills the protozoa and stains their internal structures so a human being can identify the species. In a 2019 comparison of stool concentration techniques, direct microscopy with wet and iodine mounts detected parasites with 90.1% sensitivity, and the better concentration methods reached 98.7% — with improved yield of Hymenolepis nana among others.11

In the laboratory that goes looking for parasites, iodine is the reagent that reveals them.

So where could iodine genuinely matter? Nucleic acid. Iodine is a strong oxidising agent, and oxidation damages DNA. If povidone-iodine at surgical concentration degrades parasite DNA in a specimen enough to defeat the broad eukaryotic PCR that found the tapeworm, that would be a real and important problem.

Nobody has measured it. It is one bench experiment. It has not been done.

The obstacles that are actually bigger

If parasites are being systematically missed in tumour pathology — and I want to be clear that we do not know whether they are — then here is the list of reasons, in my judgement, in order of size. Iodine is on the list. It is not near the top.

One: nobody orders the test. This is the whole of the problem in one sentence, and the case reports say so themselves. A peritoneal mass in a 47-year-old Korean man was called a gastric duplication cyst on CT and turned out to be a migrated lung fluke. The authors' recommendation was not a new reagent. It was: "clinicians should bear in mind that an intra-abdominal mass may be related to a parasitic infection."12 The bottleneck is the differential diagnosis — the list of possibilities in a clinician's head — and a possibility that is not on the list does not get tested for.

Two: the arithmetic of sampling. Let us do it. A core needle biopsy takes a cylinder about one millimetre across and one and a half centimetres long, which is roughly fifteen cubic millimetres of tissue. A three-centimetre tumour is about fourteen thousand cubic millimetres. So the core is about a thousandth of the mass. And then the slide is a four-micrometre plane through that thousandth.

For cancer, this works, because cancer cells are everywhere in the mass — any sample will do. For anything rare and focal, it works terribly. The authors of that fluke case name it precisely: detecting the parasite's eggs in biopsy specimens "may be difficult due to an insufficient amount."12 Not chemistry. Arithmetic.

Three: section thickness and training. A schistosome egg is roughly 150 micrometres long. A tapeworm segment is millimetres. At four micrometres, an organism appears as a slice through something — a fragment, a ring, an unfamiliar texture. Recognising it requires having seen it before, and a pathologist trained in Chicago may complete an entire residency without encountering a schistosome egg in tissue. It was not in the teaching set.

Four: the targeted tests are targeted. Human antibodies, human primers. Already covered, and it is a hard ceiling.

Five: formalin fragments DNA. Here the popular story is closer to right than it is about iodine. A systematic review of a hundred and eleven studies on getting usable DNA out of fixed tissue states it without euphemism: preservation in formalin "is known to damage DNA through crosslinking activity", producing "severely fragmented DNA of variable yields", which "reduces the ability to perform downstream molecular analyses."13

So the step that makes tissue readable by eye is the same step that degrades the molecules you would need to identify an organism nobody suspected. That is a genuine, quantified, structural conflict at the heart of diagnostic pathology. And it is a hundred and thirty-year-old engineering decision nobody has revisited: formalin won in the 1890s because it was cheaper than the alcohol fixatives that preceded it, and alcohol-based fixatives do not cross-link, do not compromise antibody staining, and do not damage DNA or RNA. One laboratory in Leiden has processed over forty thousand skin biopsies and a hundred thousand cervical samples that way.14

But — and this is the reason this episode does not end in despair — fragmented is not unreadable.

In 2022 a Japanese group recovered roughly seven thousand genes from a punch of tissue a tenth of a millimetre across and ten micrometres thick, taken from mouse liver blocks that had been sitting fixed for two years. They then did the same on human lung cancer blocks fixed six years earlier, and characterised the tumour microenvironment from it.15

A few dozen cells. Six-year-old blocks. Usable sequence.

And the tools to name an organism from a scrap of sequence have existed for over a decade. A PCR-and-pyrosequencing assay published in 2014 distinguishes five different fish-borne flukes from a 46-nucleotide window, and detects a single Opisthorchis or Clonorchis egg in a hundred milligrams of stool. The authors built it because the eggs are "difficult to differentiate morphologically ... even for experienced technicians."16

So: the archive exists. Every hospital on earth is holding decades of paraffin blocks at room temperature. The method exists and is routine. The assay that found the tapeworm was run on fixed tissue.

Nobody has pointed it at this.

Why a parasite is not on the list

I have said twice now that the real bottleneck is the differential diagnosis — the list of possibilities in the clinician's head. Let me say why that list looks the way it does in the United States, because it is not stupidity and it is not a conspiracy.

American medical training teaches parasitology as tropical medicine. It is framed, from the first lecture, as something that happens elsewhere: a travel history question, a returning-traveller scenario, a board exam vignette that begins "a 34-year-old returns from Southeast Asia". The implicit model is that parasites are geographic, and the geography is not here.

That model is wrong in three specific ways.

People move. The clinical review of schistosomiasis-associated bladder cancer is explicit that the knowledge gap is a problem specifically because of increased migration from endemic low- and middle-income countries to high-income ones, and that the deficiency in experience "may lead to inadequate management."3 An adult who acquired a fluke at age eight in the Nile delta and develops cancer at fifty-five in Illinois is not a tropical medicine case. They are a Tuesday.

The timescales are wrong for how we take histories. A liver fluke lives in the bile ducts for ten to thirty years before the cancer appears, and is asymptomatic for most of that time. So the exposure that matters is not in the recent travel history. It is in a childhood on another continent, decades ago, and nobody asks, because "have you travelled recently" is the question the system trained us to ask.

And some of it is simply here. Hymenolepis nana, the tapeworm in the New England Journal case, is the commonest tapeworm in humans and does not require a passport.

Put those together and you get a predictable failure. The organisms are present, the exposure window is decades wide, the clinical presentation is a mass or an anaemia rather than a parasitic illness, the differential does not contain them, and the laboratory tests only what it is asked to test. Every individual link in that chain is defensible. The chain still ends where it ends.

Ivermectin

First, what ivermectin is, because the argument about it is conducted almost entirely by people who have never been told.

It is not a laboratory invention. It comes from a bacterium. In the 1970s Satoshi Ōmura, a Japanese microbiologist, was systematically culturing soil organisms looking for useful chemistry, and one soil sample yielded a Streptomyces species nobody had seen before, since named Streptomyces avermitilis. Streptomyces are the soil bacteria that gave us a large fraction of our antibiotics; they make elaborate chemical weapons to fight the other organisms in the dirt with them. William Campbell, working at Merck, found that this one paralysed parasitic worms. The two of them shared the Nobel Prize in Physiology or Medicine in 2015.

Structurally, ivermectin is a macrocyclic lactone. Take those two words apart. Macrocyclic means the molecule's backbone is one very large ring — sixteen atoms around, rather than the six-carbon rings you may remember from school. Lactone means the ring is closed by an ester bond, which is a particular link between a carbon and an oxygen. Hanging off that big ring are two sugar units and a second, smaller ring system fused on one side. It is a large, greasy, elaborate molecule — far too big and far too fat-soluble to dissolve well in water, which is exactly the problem the nanoparticle researchers keep running into.

What it does in a worm is precise and worth knowing. It activates glutamate-gated chloride channels — pores in nerve and muscle cell membranes that invertebrates have and vertebrates do not. Ivermectin holds them open, chloride floods in, the cell cannot fire, and the parasite is paralysed.24

Now the part that matters for everything that follows, and it is not what most people believe. Ivermectin is not selective. In mammals it also activates a whole family of receptors we very much do have — GABA-A and glycine receptors, which are the brain's main brakes, and nicotinic acetylcholine receptors, which are an accelerator — along with P2X4 and farnesoid X receptors and an inwardly rectifying potassium channel.24

So why is it safe enough to give to hundreds of millions of people for river blindness? Because it barely gets into the brain. A pharmacology review puts the condition plainly: ivermectin penetrates the mammalian brain poorly and exerts no meaningful effect through those brain receptors — unless it is used at high, potentially toxic doses, or the blood-brain barrier is impaired.25

Read that condition against everything we are about to discuss. The famous safety record of ivermectin is a safety record at antiparasitic doses. The same review notes that in seizure models, the doses that produce an effect sit in the range of the dose that kills half the animals.25 The drug is not safe in some abstract way. It is safe at the dose we give it, because at that dose it stays outside the one organ that would object.

So: a natural product, from soil, with a genuinely elegant mechanism, a Nobel Prize, and one of the great safety records in medicine. None of which tells you anything at all about whether it treats cancer — and that gap, between a drug being good and a drug being good for this, is where the whole argument lives.

Which brings us, inevitably, to the drug. I am going to do this the same way as everything else: the mechanism, then the data, then the number that matters, then the thing nobody wants to say.

There is a real mechanism, and it is not the one people assume. The popular version is that ivermectin kills a parasite that was causing the cancer. That is not what the oncology research is about at all.

Ivermectin binds a human protein called TEL2, part of a three-protein assembly known as the TTT complex. The TTT complex is a chaperone — a molecule whose job is to help other molecules fold into the right shape. What it chaperones is a family of very large kinases with names like ATR, ATM, DNA-PKcs and mTOR: the machinery that senses DNA damage and the machinery that controls cell growth. Both are central to cancer.17

So ivermectin has a plausible human target, in a pathway cancers depend on. That is genuinely interesting and it is why serious laboratories are working on this.

The preclinical data is real and it is consistent. Ivermectin kills cancer cells in culture, arrests the cell cycle, triggers apoptosis through the mitochondria, generates reactive oxygen species, and shrinks tumours in mice. Dozens of papers. In one 2025 study against lymphoma cells it had an IC50 — the concentration that kills half the cells — of 10.55 micrograms per millilitre, against cisplatin's 8.32 in the same assay.18

That number, 10.55, is the whole story, and almost nobody quotes it next to the number it needs to be next to.

A person taking ivermectin at ordinary doses reaches plasma concentrations in the region of 0.02 to 0.05 micrograms per millilitre.

The concentration that kills cancer cells in a dish is two to three orders of magnitude above what a human being taking the drug actually achieves. Not a bit higher. Hundreds of times higher.

That gap is not a detail. It is the central fact about ivermectin in oncology, and it applies to almost every repurposing story you will ever hear — most molecules kill cancer cells at some concentration, and the question is always whether a person can get there without dying first.

The people who take this most seriously know it perfectly well, which is why they are not giving out tablets. In 2026 a group in Alexandria published ivermectin nanocrystals wrapped in platelet membrane, engineered to home to tumour tissue and evade immune clearance, active against triple-negative breast cancer in mice. And in their own opening sentence they state the problem: repurposing ivermectin is "hindered by poor solubility and high toxicity, restricting its parenteral administration."19

The researchers most invested in this drug's anticancer future are the ones building entire delivery systems to avoid giving it the way people are buying it.

There is one preclinical result I think is genuinely promising, and it is not about killing tumours. Lung cancer cells made resistant to paclitaxel achieve that resistance by overproducing P-glycoprotein — a pump in the cell membrane that throws chemotherapy back out. When ivermectin was given alongside the escalating paclitaxel, that pump induction was completely abolished, the drug stayed inside, and the cells never became resistant.20 If ivermectin has a place in oncology, preventing resistance alongside real chemotherapy is a far better hypothesis than replacing it.

And now the human evidence, which is where this gets uncomfortable.

In 2026 Anticancer Research published a prospective observational cohort reporting real-world outcomes of ivermectin and mebendazole in cancer patients, with a "Clinical Benefit Ratio of 84.4%" and specific percentages of tumour regression. It was, predictably, everywhere.

The journal has since issued a formal Expression of Concern and opened a Post-Publication Data Integrity and Ethical Oversight Audit — examining whether the study had institutional review board approval or exemption, whether the baseline cancer diagnoses of the 197 initial participants can be source-verified from de-identified clinical records, and whether there is objective medical documentation supporting the reported anatomical regressions.21

The editorial board's sentence is the one to carry away, and it applies to far more than this paper:

That is what the strongest publicised human evidence for ivermectin in cancer currently amounts to: a paper under audit.

Meanwhile the actual answer is being pursued properly. A registered trial — ivermectin combined with immune checkpoint inhibition — is underway, and the National Cancer Institute has an interest in the question. That is how this gets settled. Not by a cohort of self-reported outcomes, and not by anyone's conviction.

What I actually think

Here is where I have landed, and it is not where I expected.

The parasite-cancer link is not fringe. It is established, and it is underplayed. Three organisms are IARC Group 1, alongside tobacco and asbestos. For one of them the mechanism has been proven at the level of a single gene, deleted, with the pre-cancer following it out. Any physician who treats "parasites and cancer" as a conspiracy theory is simply unaware of their own literature, and that includes a great many physicians.

The evidence outside those three is genuinely thin, and thin for a specific reason. S. mansoni sits in Group 3 with case reports and animal models and no adequate human cohort. O. felineus has a positive animal model, an 80%-infested population, and no category at all. That is not a field that examined these organisms and cleared them. It is a field that ran out of funding.

The diagnostic pipeline is not hiding anything, and it is not looking either. Six steps, all superb, all aimed at one question. If you ask me whether there are parasites in tumour specimens that nobody is reporting, my honest answer is: I do not know, and the reason I do not know is that the experiment is cheap, the tissue is already in the building, the method is published, and nobody has run it.

On the iodine claim specifically, I have to tell you the premise did not survive in the form it is usually told. Iodine kills parasites — that is established and it is stronger than I expected. Iodine reaches specimens — that is documented on three routes. A specialty has published a clinical warning that pre-sampling antiseptics cause false-negative diagnostic tests — that is real and in print.

And it is a warning about culture, and cancer specimens are not cultured. They are fixed in formaldehyde, which kills everything and preserves it. A dead parasite on a slide looks like a parasite. Iodine is also, in the laboratory that goes looking, the stain used to find parasites.

I would love the simpler version to be true. It is a better story. But the version that survives contact with the literature is more interesting anyway, because it does not require anybody to be hiding anything: the reason nobody finds parasites in tumours is that nobody is asking, and the tools to ask have been sitting unused for a decade.

That is a worse indictment than a cover-up, honestly. A cover-up would at least imply somebody thought the question mattered.

Where this goes next

There is more here than an hour holds. A second episode should take up: Helicobacter pylori, which is the closest precedent we have — a chronic infection that was dismissed for a decade, is now a Group 1 carcinogen for gastric cancer, and whose discoverer had to drink a culture of it to be believed; Strongyloides stercoralis and its long-recognised travelling companion HTLV-1, where parasite and virus and leukaemia appear together often enough that a case report will note the strongyloidiasis history in the same paragraph as the T-cell lymphoma;22 the inverse question of helminth therapy, where deliberate infection is being investigated as a treatment for autoimmune disease, which is the same biology read from the other end.

And the study. Take a series of archived tumour blocks. Run the broad eukaryotic PCR — the one that asks "what organism is this" instead of "is this cancer". Report what fraction come back non-human. That is a cheap, fast, publishable experiment, and until somebody runs it, every confident answer to this question, in either direction, is a guess.

Sources

Every paper referenced on air, in the order it comes up. Links go to the publisher via DOI.