One Case, and the List That Holds Three Names

Part one sets the ground. A man with HIV in Medellín had biopsies read as unmistakable cancer; the cells were the wrong size, a test that asks what organism this is returned dwarf tapeworm DNA, and sequencing found large rearrangements in the parasite's genome compatible with mutations described in cancer. We take that case apart word by word — monomorphic is not monoclonal, compatible with is not caused by, and one patient is one patient — then show it was the end of a forty-year trail running through 1976, 1996 and 2003. Then the institution: what IARC is, how a Monograph is made, who sets its priorities, why hazard is not risk, and what each of its four groups actually says. Then the parasites, with the dates corrected, the arithmetic of their attributed burden, and the field's own founding false positive — the 1926 Nobel Prize for a worm that did not cause cancer.

Show notes

A man walks into a hospital in Medellín, Colombia. He is infected with HIV, the human immunodeficiency virus, which destroys the immune cells a person needs to fight infection. He has swollen lymph nodes in his neck and abnormal shadows in his lungs. The doctors take biopsies — small pieces of tissue cut out and sent to the laboratory — one from a lymph node, one from the lung.

The pathologist puts the slides under the microscope and sees cancer.

Not a borderline case. Nests of cells, packed together, every one like every other, with none of the structure that would say what tissue they came from, pushing into the tissue around them. In the published words of the paper, the morphologic features and invasive behavior of the cells were characteristic of cancer. But their small size suggested a nonhuman origin.

That is Muehlenbachs and colleagues, writing in The New England Journal of Medicine in November 2015. Eighteen authors, from the Centers for Disease Control and Prevention, Emory University, a university and two hospitals in Medellín, a specialist center in nearby Rionegro, Asahikawa Medical University in Japan, and the Natural History Museum in London.

Somebody in that chain made the decision that is the entire reason we know about this. Instead of another antibody test to work out which human cancer it was, they ran a test that asks a different question. The published phrase is a polymerase-chain-reaction assay targeting eukaryotes. It does not ask “which human cancer is this?” It asks “what organism is this?”

The answer came back Hymenolepis nana. The dwarf tapeworm.

They confirmed it two more ways, in this order. Although the cells were unrecognizable as tapeworm tissue — the published wording — immunohistochemical staining and probe hybridization labeled the cells in situ. Then they sequenced. Comparative deep sequencing identified H. nana structural genomic variants that are compatible with mutations described in cancer.

And the conclusion the authors published, word for word: invasion of human tissue by abnormal, proliferating, genetically altered tapeworm cells is a novel disease mechanism that links infection and cancer.

Four terms first. A polymerase chain reaction, or PCR, copies a chosen stretch of DNA over and over until there is enough to read. A eukaryote is any organism that keeps its DNA inside a nucleus — animals, plants, fungi, the single-celled parasites, but not bacteria. So a PCR assay targeting eukaryotes aims at a stretch of DNA every such organism carries and reads the variable patch beside it to find out whose DNA it was. In situ hybridization applies a labeled DNA or RNA probe to intact tissue so it sticks to its match and you can see which cells contain it. Immunohistochemistry does the same with antibodies, staining one chosen protein a visible color — so it answers only the question you thought to ask.

One more thing about the cold open. Much of what gets repeated about this case is not in the record I can verify: the patient's age, his CD4 count — the standard measure of how much immune defense a person with HIV has left — and how long he lived afterward. Those figures circulate; they were not confirmed against the full text, so I will not state them. The record establishes one adult man, infected with HIV, two biopsy sites, and a tapeworm's cells behaving like a cancer inside him.

The three words that keep this honest

Word one. The paper says the cells were monomorphic. Monomorphic means all of one shape: under the microscope, the cells look alike, and that is all it says. It is not the word monoclonal, which means all the cells descend from one original cell — a claim about ancestry, needing genetic evidence of common descent, not a visual impression of sameness. Whether that population was monoclonal is not something I can state from the published record, and when you hear this case described as a “clone” of malignant tapeworm stem cells, that word has been added by someone else.

Word two. The variants are compatible with mutations described in cancer. The authors chose that phrase, and it does not say driver mutations. A driver mutation is a change shown to push a cell toward malignancy; a passenger is along for the ride, and telling them apart takes functional experiments or very large datasets, neither of which exists for a tapeworm. A structural variant, which is what they report, is a large-scale rearrangement: a deleted, duplicated, inverted or fused segment of DNA, not a single-letter change. So: large rearrangements of the kind also found in cancers. Not: the mutations that caused it.

Word three, the shortest. One. This is a case report: the description of a single patient. It cannot tell you how often something happens or who is at risk; it can prove something is possible. Before 2015 the belief was stated in this paper's first sentence: neoplasms occur naturally in invertebrates but are not known to develop in tapeworms. A neoplasm is an abnormal growth of cells, a tumor in the broad sense.

And one thing this case is emphatically not. It is not a transmissible cancer — a cancer that spreads as living cancer cells passed from one animal to another, so the tumor in the new host is made of the original animal's cells. That is real: Yonemitsu and colleagues, in eLife in 2019, record transmissible cancers identified in Tasmanian devils, in dogs and in four bivalves, shellfish. In Medellín nothing passed between two people. A parasite already living inside one man had cells that started proliferating where they should not.

Not a bolt from the blue

Start in 1976. Connor and colleagues, in Archives of Pathology and Laboratory Medicine, report the autopsy of a middle-aged man who died of Hodgkin disease, a cancer of the lymph nodes. Their description: a peculiar metazoan parasite — a multicellular animal — had proliferated and disseminated throughout his body. They could not identify it. Electron microscopy showed the structure of a flatworm, and they concluded it was an aberrant sparganum manifesting uncontrolled proliferation and dissemination. A sparganum is the tissue-dwelling larval stage of certain tapeworms. That paper's title ends: Possibly a mutated sparganum. The idea in the 2015 case is fifty years old.

1996. Santamaría-Fríes and colleagues, in The Lancet, describe a man infected with HIV whose rapidly enlarging abdominal mass — in their words, suggestive of a neoplasm — invaded his liver and killed him. The tissue morphology matched no known disease process. They ran broad-range ribosomal DNA amplification, aimed at the gene that builds the ribosome — the cell's protein-assembly machine, conserved in every organism. From 357 base pairs of 18S ribosomal DNA amplified directly out of the tissue, they concluded the agent was a previously uncharacterized cestode — a tapeworm.

2003. Olson and colleagues, in The Journal of Infectious Diseases, go back to two cases of lethal cestodiasis in which the agent had been misidentified, and run broad-range 18S ribosomal DNA PCR on both. In the first, a patient with AIDS — acquired immunodeficiency syndrome, what HIV infection is called once the immune system has collapsed far enough that ordinary infections turn lethal — they identify Hymenolepis nana as a cause of aberrant metastatic larval disease. In the second, a patient with Hodgkin disease and similar pathology, they find a larval cestode with a previously uncharacterized sequence, and note that a prior report of that case nearly thirty years earlier had suggested a sparganum. The paper does not say its two patients are the 1996 and 1976 patients, so I will not either — but the overlapping authors and the matching interval make it very likely.

And that second sequence has since been named. Deplazes and colleagues, in a 2019 review in International Journal for Parasitology: Parasites and Wildlife, reanalyzed a published nucleotide sequence from a historic human case in the United States and attributed it to Versteria, a tapeworm of the weasel family. They do not say which case, so I keep the caution I just used: that it is Connor's 1976 autopsy is my inference, not their finding. If it is, a database finally had something to match it to sixteen years after the sequence was published, forty-three after the autopsy.

What improved across those four decades was not luck; it was the method and the reference database. In 1976 you had an electron microscope and a trained eye; by 1996, a conserved gene and a sequencer. By 2013 Tsai and colleagues had published the genomes of four tapeworm species in Nature, including the laboratory model Hymenolepis microstoma, spanning 115 to 141 megabases — a megabase is a million letters of DNA, so these are genomes a twentieth the size of ours. That is why comparative sequencing was thinkable in 2015: a relative's genome existed to compare against. Reference material is the ceiling on what any test can tell you.

So the defensible uniqueness claim is this. The 2015 case is the only one in which a parasite's own cells were shown to be genetically altered and behaving like a cancer inside a human being. It is not the first case of proliferating tapeworm tissue in an immunosuppressed person. Kikuchi and Maruyama's 2019 review in Parasitology International gathers every original human case report of proliferative sparganosis over 115 years — a disease in which the larva of Sparganum proliferum multiplies through human tissue — and it carries a warning worth keeping: some cases previously cited as proliferative sparganosis were removed from the list, judged to be a different group of tapeworms. Cases struck from the record once somebody sequenced them. Sequencing, they conclude, is mandatory for a definite diagnosis.

Two notes on the record. The 2015 paper drew published correspondence: a one-page letter from David Conn, of the Harvard Museum of Comparative Zoology, in The New England Journal of Medicine in March 2016, and a reply the same day from Muehlenbachs, Mathison and Olson. Both are behind a paywall this show could not get past, and neither is indexed with an abstract — which means the only published objection to this case is one I have not read. And Soriano and Barreiro, in a 2015 commentary prompted by this case, supply the framing: the incidence of malignancies associated with cancer-causing viruses rises as the CD4 count falls. Immune collapse is the gate.

What the organism actually is

Hymenolepis nana is a tapeworm: a flat, ribbon-shaped parasitic worm that lives in the intestine as an adult, growing as a chain of segments behind a small head, and called the dwarf tapeworm because it is small. The London refugee-health report by Killington and colleagues, published in 2025, describes adults as 2 to 4 centimeters long and about 1 millimeter wide. Roughly an inch of worm, thin as a thread. One note on its name: that report is titled Rodentolepis nana, and writes the worm Rodentolepis (Hymenolepis) nana. The genus was split, both names stay in use for the same animal, and I will say Hymenolepis because that is what the 2015 case paper says. Who split it, and when, I could not verify. I am telling you the other name exists so you do not look the worm up and think you have found a second organism.

And it is strange in a way that matters. Most tapeworms need two animals: an intermediate host, in which the larva develops, and a definitive host, in whose gut the adult lives — the pork tapeworm uses a pig, then a person. This one does both in one mammal. Ito and Budke, reviewing the life cycles in 2021, call it unique in that it can complete its entire life cycle within a single mammalian host.

Which gives it the feature that matters for a man with no immune system. Autoinfection: eggs released inside the gut can hatch there and start new larvae without the person ever being exposed again. Al-Mekhlafi's 2020 Yemen study notes it as the only cestode transmissible directly from person to person, with endogenous autoinfection the author says may have implications for immunocompromised patients. In a healthy person immunity shuts that cycle down fast. In a person whose immunity has collapsed, it does not.

One more piece of biology, because it is what makes 2015 coherent rather than magical. Tapeworms grow from a reserve of undifferentiated cells. Koziol and colleagues showed in 2014, in the related tapeworm Echinococcus multilocularis, that only these germinative cells proliferate — the single engine of the larva's growth. And Tsai and colleagues' 2013 genomes found extreme losses of genes and pathways ubiquitous in other animals, including 34 families of homeobox genes — the switches that tell a developing animal which body part goes where — and several determinants of stem cell fate. So: growth running entirely through a dividing stem-cell compartment, in an animal missing several of the genes others use to tell stem cells what to become. That is plausibility, not explanation: which cell type the tumor arose from is not stated in the paper, and neither study is about H. nana.

And the species is less settled than the textbooks. Nkouawa and colleagues, in 2016, recovered an adult tapeworm of this family from a 52-year-old Tibetan woman in Sichuan, China. The segments could not be identified by shape. Sequencing of two independent stretches — the ribosomal gene again, and cox1, from the mitochondria, which carry a small genome of their own — showed it was distinct from both Hymenolepis diminuta and Hymenolepis nana, the two species conventionally held to infect people. Akira Ito is an author on that paper and on the 2015 case. A textbook claim was wrong — and any identification made from an egg under a microscope is one genetics can overturn.

Is it really the most common tapeworm?

Cabada and colleagues open their 2016 paper in the American Journal of Tropical Medicine and Hygiene with this sentence: Hymenolepis nana is the most common cestode infection in the world. A peer-reviewed tropical medicine paper from the University of Texas Medical Branch — and notice what the sentence is: a statement in an introduction. Jin's 2025 analysis opens the same way — the dwarf tapeworm affects 50 to 75 million people worldwide — and Al-Mekhlafi's 2020 Yemen paper puts the figure above 75 million. Each is citing the number, not deriving it. And hymenolepiasis — carrying this worm — does not turn up in the cause list of the Global Burden of Disease study, the standing project that estimates how much illness and death every disease causes, country by country. If a condition is in it, somebody has modeled it. This one is not — an absence I report from my own search, not from any published statement. The narrower claim a paper will actually make is Jin's: no studies have explored the disease burden of H. nana in Sudan.

Sudan, the largest dataset. Jin's 2025 secondary analysis of a nationwide survey carried out in 2017 across 189 districts and 18 states: 105,167 students selected from 1,772 schools, 96,679 stool samples collected, 4,706 of them positive for H. nana. That is 4.9 percent. Fewer than 1 percent also carried a schistosome — a blood fluke — and 0.1 percent soil-transmitted worms. Children in households with improved latrines had lower odds of infection: an adjusted odds ratio of 0.87, 95 percent confidence interval 0.80 to 0.94. An odds ratio compares the odds of an outcome in one group against another: 1.0 means no difference, 0.87 about 13 percent lower. Adjusted means other measured factors were removed statistically — only as good as the factors somebody thought to measure. A confidence interval is the range compatible with the data; if it includes 1.0, the finding is not convincing.

Northern Peru. Vilchez Barreto and colleagues, 2017: among 14,761 children aged 2 to 15, 1,124 were infected. That is 7.61 percent, and the authors volunteer in their own abstract that it is a likely underestimate, because only a single stool sample was examined by microscopy — telling you the direction of their own error. The strongest associations were lack of adequate water, adjusted prevalence ratio 2.22, the same kind of comparison as an odds ratio, and lack of sanitation in the house, 1.94.

Rural Yemen. Al-Mekhlafi, 2020: 498 children aged 2 to 17, screened by three methods — a wet mount, a drop of stool under a coverslip; formalin-ether concentration, which spins the sample so eggs collect in a pellet; and Kato-Katz, which examines a measured amount so eggs per gram can be counted. 77.5 percent carried at least one intestinal parasite, and the dwarf tapeworm was in 17.5 percent, 87 of 498. Risk was highest under age 6, adjusted odds ratio 4.28 with an interval of 2.04 to 8.98.

Now the honesty check, which keeps “most common” from turning into something false. Goudarzi and colleagues pooled the human prevalence of this parasite in Iran in a 2020 systematic review and meta-analysis — a method that pools results from multiple studies into one estimate, and inherits every weakness of the studies it pools. Their pooled figure for humans was 1.2 percent, interval 1.0 to 1.4. Schoolchildren 2.2 percent. Food handlers 0.5 percent. In rodents in the same country, 13 percent, and in rats specifically 19.6 percent.

One point two percent. Not seventeen. In a middle-income country with functioning sanitation, this worm is uncommon in people and common in the rats. Which tells you what “most common tapeworm in the world” means: a claim about the global and tropical total, and about which tapeworm turns up most often in surveys that count more than one. Not a claim that any given population carries much of it. Killington and colleagues make the point from the other end: 54 of 1,797 people screened at a London refugee health service between 2016 and 2023, 3 percent. And Wu and colleagues, across 31 rural communities in Madagascar this year, found the dwarf tapeworm at up to 10.5 percent and — the first of these datasets to give the worm a climate preference — most common in the dry regions, not the humid east.

So the sentence I will use for the rest of this series: the dwarf tapeworm is the one most often found in people in surveys that count both, it is asserted in the peer-reviewed literature to be the most common cestode infection in the world, and its global total has never been measured. All three clauses are true. Dropping the third is how a reasonable claim becomes a fake number.

The same worm, pointing the other way

Start with a matched case-control study, the only properly designed one I found. A case-control study starts with people who have a condition and compares them against matched people who do not. Chero and colleagues, in Transactions of the Royal Society of Tropical Medicine and Hygiene in 2007, worked in a northern coastal Peruvian village. They took 25 people infected with this tapeworm and nothing else — anyone with other intestinal worms or with Giardia was excluded, which is the hard part and the reason the study matters — and compared them against 46 matched controls. No significant differences in symptoms except headache, which was less frequent in the infected.

Hold that against the Yemen and Cusco findings, read exactly. In Yemen the significant associations were abdominal pain, 39 percent against 28; anal itching; and dizziness, 21 against 11. Diarrhea and fatigue were not significant there, nor was the overall count of gastrointestinal symptoms. In Cusco they were diarrhea, jaundice, headache, fever and fatigue. Note the headache: more frequent in the infected in Cusco, less frequent in the one study that excluded co-infections. These are cross-sectional associations — measured at a single moment, which cannot establish which came first — and in Yemen 72 percent of the infected children carried other parasites, which the author says makes clean attribution impossible. Strip the co-infections out and the symptom signal largely disappears. That is what a confounder looks like when it is finally removed.

The second study is stranger. Ramos-Martínez and colleagues, in Cytokine in 2019, took four groups of fourteen female mice, all of one inbred strain, BALB/c, so the animals are near-genetically identical and the only difference between groups is what was done to them. One got the solvent alone. One was infected with H. nana. One got a chemical carcinogen painted on the skin — dimethylbenzanthracene, a standard laboratory tumor inducer. One was infected first and then given the carcinogen. The mice infected beforehand developed fewer tumors, and smaller ones, than those given the carcinogen alone, with significantly more eosinophils and neutrophils — two kinds of white blood cell — in the protected animals.

Say that plainly: in the only controlled experiment on this parasite and cancer, prior infection protected against chemically induced tumors. Fifty-six mice, no effect sizes in the abstract, skin tumors from a painted chemical. It does not transfer to humans in either direction — but it is the only controlled evidence there is, and it runs against the intuition the 2015 case creates. Nor is that direction a freak: Mehdioghli and colleagues, in Discover Oncology this year, catalog antitumor activity in experimental models for Plasmodium, Toxoplasma, Leishmania and two helminths, with the honest coda that the evidence base remains largely restricted to preclinical systems; and Shemfe and colleagues, reviewing 17 human studies, found helminths shifting gut bacteria in both directions, with none of the 17 measuring colorectal cancer.

The list, and who keeps it

So, the question underneath the series — not whether a parasite can cause cancer, which is closed. Three organisms cleared the highest bar there is, and to explain that bar I have to explain the institution that sets it.

IARC stands for the International Agency for Research on Cancer, the World Health Organization's specialized cancer agency, established in 1965 and working from Lyon, in France — sixty-one years old this year. It does several jobs that get confused: it counts the world's cancers, publishes a series on what prevents cancer, and decides whether a given thing can cause cancer in people. That last job has a name that is the key to every argument about this agency: hazard identification. Asking only whether something can cause cancer at all, in anyone, under any realistic circumstance. IARC itself calls it the essential first step in cancer prevention — that phrase is from Samet and colleagues' 2020 account of the agency's procedures.

The product is a Monograph: a numbered, book-length volume in which an invited panel of outside experts reviews the published evidence on one or more agents and states whether each can cause cancer in people. Baan and Straif, two retired members of the program writing in ALTEX in 2021, date the Monographs to the early 1970s. Krewski and colleagues' 2019 survey gives the scale: more than 1,000 agents evaluated, and through volume 119, 120 of them met the criteria for the top category. By March 2024 the count stood at 128, in the Advisory Group report published in The Lancet Oncology by Berrington de González and colleagues. By 2026 IARC's own survey of its program, by Schubauer-Berigan and colleagues, counts 135, with 24 agents added since the last full review of the category — Volume 100, compiled in 2008 and 2009, so call it seventeen years.

How a rating is actually made

The standing rulebook is the Preamble: a document printed at the front of every volume setting out, in Baan and Straif's description, the objective and scope of the program, the principles and procedures used in developing a Monograph, the types of evidence considered, and the criteria that guide evaluations. It was revised in 2006 and was substantially rebuilt in 2019, following the recommendations of a 2018 expert advisory group — the first major revision in thirteen years.

That 2019 revision is the current rulebook, and four things in it matter for us. Systematic review methods were strengthened. Epidemiological studies are now scrutinized more closely for quality and informativeness, including how they measured exposure — usually the weakest link. Mechanistic evidence was given much greater weight. And the three streams of evidence are now combined in a single step. Those streams are the spine of every evaluation: cancer in humans, the epidemiology; cancer in experimental animals, the long-term laboratory studies; and mechanisms, how the agent could cause cancer at the level of cells and molecules. Before 2019, mechanisms were weighed after a preliminary verdict had already been formed from the first two. Now all three go into one integration step, under a published table.

Each stream gets its own grade, and the vocabulary is load-bearing. Human and animal evidence are each rated sufficient, limited or inadequate. Sufficient, in humans, means a positive association in which chance, bias and confounding have been ruled out with reasonable confidence. Confounding is when a third factor — smoking, another infection, poverty — produces an apparent link between two things that are not causally related. Limited means a causal reading is credible but those explanations could not be excluded. Mechanistic evidence is rated strong, limited or inadequate on its own scale.

The mechanistic stream runs on a framework with a specific origin. Smith and colleagues, in Environmental Health Perspectives in 2016, explain that reviewing the more than 100 agents in the top category had been complicated by the absence of a broadly accepted method for evaluating mechanistic data. So IARC convened two workshops, and a working group identified ten key characteristics of carcinogens: properties commonly shared by agents already known to cause human cancer. As published, in two breaths. The ability to act as an electrophile, meaning reactive toward DNA; to be genotoxic, which means to damage DNA; to alter DNA repair or cause genomic instability; to induce epigenetic alterations, changes to the chemical tags that decide which genes a cell reads; and to induce oxidative stress, a flood of reactive oxygen that damages whatever it touches. Then: to induce chronic inflammation; to be immunosuppressive; to modulate receptor-mediated effects, jamming the switches a cell uses to hear hormones and growth signals; to cause immortalization, letting a cell divide past the limit that retires it; and to alter cell proliferation, cell death, or nutrient supply.

Ten properties. Not a checklist that produces a verdict by itself, but a way of sorting mechanistic studies so a working group compares like with like. And not decorative: Krewski and colleagues examined the profiles of 86 established human carcinogens in 2019 and found the most prevalent characteristic was genotoxicity, with agents averaging four of the ten each. Characteristic six, chronic inflammation, is the one that carries the parasite classifications.

Now the room. A Working Group is the panel of independent scientists who review the evidence, reach the evaluation and write the volume. They are not IARC employees and they are not paid. The volume that evaluated the parasites is volume 61, 241 pages, and its printed front matter records the Working Group meeting in Lyon from 7 to 14 June 1994 — eight calendar days. The database record carries neither an abstract nor those dates, so that citation is to the printed book, not to PubMed. Subgroups take the three streams separately, the whole group argues it out, and they vote.

And one procedural fact, which I am giving you from the agency's sharpest critics because they are right about it. Goodman and colleagues, in Regulatory Toxicology and Pharmacology in 2020, grant that the 2019 amendments are an improvement, then note that the conclusions of the working groups are not subjected to outside, independent scientific peer review. That is true. Three of the five authors are staff of the American Chemistry Council, the United States chemical industry's trade association; the other two work at a commercial consultancy. Both facts belong in the same breath.

Who tells them what to look at

Who tells them what to look at? A separate body with an unlovely name and a real function: the Advisory Group to Recommend Priorities. It meets roughly every five years, and its job is to decide what the Monographs program should evaluate next.

Berrington de González and colleagues reported the most recent round in The Lancet Oncology in 2024. In March 2024, 28 independent scientists from 22 countries met in Lyon to recommend priorities for 2025 through 2029. They assessed a public call for nominations — anyone can nominate an agent, including a private citizen — and considered more than 200 candidate agents. Priority goes to agents where there is evidence of current human exposure and enough published evidence to support a new or updated evaluation. Any one of the three streams, alone, can justify prioritizing an agent never evaluated before.

Then the line in that report which is the reason this series exists. The group noted that the program had not evaluated pesticides since 2015, and had not evaluated infectious agents since 2012 — leaving nineteen high-priority recommendations in the first category and ten in the second.

Count the years — then count them again, because the gap has since closed once. In June 2025 Karagas, Kaldor, Michaelis and colleagues announced in The Lancet Oncology the evaluation of hepatitis D virus, human cytomegalovirus and Merkel cell polyomavirus: the first infectious agents the program had taken up in thirteen years, and not one of them a parasite. So the sentence that holds is narrower, and still damning: no parasite has been evaluated since the malaria parasite in 2012, and no fluke since 2009. Every parasite rating we are about to walk through is at least that stale. That 2024 group also judged a systematic appraisal of all 128 agents then in the top category warranted, because suggestive evidence of additional cancer sites had turned up for all ten of those nominated — the agency telling you, in its own publication, that its list is probably incomplete in ways it has not checked.

Hazard is not risk

Hazard is whether something can cause harm at all. Risk is the chance that harm actually happens at a given level of exposure. A shark is a hazard; whether one bites you depends on whether you swim where the sharks are. IARC does the first job and says so in print: Samet and colleagues state that hazard identification as conducted in the Monographs is distinct from risk assessment, in which exposure-response characterization is used to estimate cancer risk for a given scenario and level of exposure. Risk assessment is the other job, done by regulators and dose-setting committees.

The four groups, worded exactly

Group 1 means carcinogenic to humans. The evidence is strong enough to be confident the agent can cause cancer in people. Hepatitis B and human papillomavirus are in it, as Rumgay and colleagues' list confirms. So is processed meat — which is why a rating that says only can it gets misread as how much.

And one exception you need, because it is the door this series is asking about. Since 2019 the streams are integrated in one step, so a Working Group can reach Group 1 without sufficient human evidence if the mechanistic case is strong enough. It has happened twice: Schubauer-Berigan and colleagues record that of the 24 agents newly placed in Group 1, all but two had sufficient human evidence for at least one cancer type, and the two that did not are perfluorooctanoic acid and dioxin-like polychlorinated biphenyls. The bar is high, and it is not only epidemiological.

Group 2A means probably carcinogenic to humans. Group 2B means possibly carcinogenic to humans. Probably and possibly are doing real work there, and they are not interchangeable — a point we will need shortly, because one of the flukes has had those two words swapped in print.

Group 3 means not classifiable as to its carcinogenicity to humans. This is the category almost everyone gets wrong in both directions. It does not mean safe. It means the evidence is insufficient to say either way: a verdict on the state of the literature, not on the organism. The correct translation is “nobody has finished the study” — and that sentence and “it is fine” get told to the public as if they were the same sentence.

Now a limit I will state rather than paper over. There used to be a fifth category, Group 4, meaning probably not carcinogenic to humans. The story goes that in the agency's whole history it held exactly one substance, which was then moved back out. The first half is supportable: Pearce and colleagues' 2015 account records Group 4 holding one agent out of 970 evaluated, under 1 percent, as of that year — but Pearce does not name it. That it was caprolactam, a chemical used to make nylon, belongs to IARC's own agent list, and that is where the citation for the name sits. The second half is wrong. The 2019 revision abolished the category and folded it into Group 3, as Samet and colleagues record, with working groups now able to add the words “probably not carcinogenic” where the evidence warrants. The governing definitions are in the printed 2019 Preamble, and that is where a citation belongs, not in any secondary restatement.

Organism by organism, dates corrected

Seven parasites have been through this process. Six are flukes — parasitic flatworms — and one is the malaria parasite. Here is where each stands, and the dates are where the usual telling goes wrong.

Volume 61, from the Working Group that sat in Lyon from 7 to 14 June 1994, evaluated the schistosomes, the liver flukes and Helicobacter pylori together. One sourcing note before the results: that volume's own database record carries no abstract, so the group assignments below come from the later literature that restates them, named as I go. Out of that week:

Schistosoma haematobium, the blood fluke that lodges in the veins draining the bladder, went into Group 1 for squamous cell carcinoma of the bladder — cancer of flat, skin-like cells, not the bladder's usual kind. Opisthorchis viverrini, the Southeast Asian liver fluke, went into Group 1 for cholangiocarcinoma — cancer of the bile ducts, the tubes that carry bile from the liver to the intestine. Schistosoma japonicum went into Group 2B, possibly carcinogenic. Schistosoma mansoni and Opisthorchis felineus went into Group 3, not classifiable. Sourcing, as promised: Shin and colleagues date the 1994 decisions on O. viverrini and C. sinensis, and Rumgay and colleagues confirm S. haematobium and O. viverrini on the current Group 1 list. For the other three my sources give only the present rating — Leija-Montoya 2022, von Bülow 2021, Jain 2026 — so read the 1994 date on those three as mine, not theirs.

And Clonorchis sinensis, the East Asian liver fluke, went into Group 2A — probably carcinogenic — because the human evidence was judged limited rather than sufficient. Not Group 1. That is the correction this part exists partly to make, because an earlier draft of this series said all three current Group 1 parasites were classified in 1994 and 2009 and never seriously challenged, and the clean version of that sentence is wrong.

Here is what actually happened to Clonorchis. In 2007 Sripa and colleagues published a review in PLoS Medicine titled “Liver fluke induces cholangiocarcinoma,” and recorded in it that C. sinensis was still not considered a Group 1 carcinogen. Then the Working Group that produced Volume 100B met in February 2009, found the human evidence now sufficient, and raised it to Group 1. The conclusions were announced in April 2009 by Bouvard and colleagues in The Lancet Oncology; the full 441-page volume appeared in 2012. Shin and colleagues, in the Journal of Korean Medical Science in 2010, open with it: in 2009, infection with Clonorchis sinensis was classified as carcinogenic to humans, Group 1. The 1994 Group 2A rating it replaced is recorded in that same paper.

So the real story is a fifteen-year gap, and it is better than the tidy one. Two flukes in the same family, doing the same thing to the same tissue in neighboring regions, and one took an extra decade and a half to clear the bar — not because the biology changed, but because the epidemiology had not been done well enough. Shin and colleagues also give the number the 2009 group was looking at: a summary odds ratio for cholangiocarcinoma of 4.7, interval 2.2 to 9.8, with roughly 10 percent of Korean cholangiocarcinomas attributable to this fluke.

And the two ratings have held up since. Huang and colleagues published the largest pooled analysis to date in Acta Tropica in 2024: 6,488 articles screened, 22 eligible studies, 34,367 participants. The association with cholangiocarcinoma came out at an overall odds ratio of 4.24, confidence interval 3.33 to 5.39 — and species by species, 4.49 for C. sinensis and 3.69 for O. viverrini. Roughly 25 million people worldwide carry one of the three liver flukes in this family — the two in that analysis, plus Opisthorchis felineus, which we come to in a moment.

Schistosoma japonicum next, because a specific error is worth killing. I have seen it written that S. japonicum is “now” classified as a probable human carcinogen. Both halves are wrong. It is Group 2B, possibly carcinogenic — not 2A, probably — and it has been there since 1994, so “now” implies a change that never happened. The error is traceable to a secondary restatement that mixed the 2A wording with the 2B grade, but I will not quote a sentence I cannot cite, so the quotation comes out until the paper carrying it is in the source list by name. The peer-reviewed source here gets it right: Leija-Montoya and colleagues, in 2022, write “group 2B: possibly carcinogenic to humans.”

Then S. mansoni and O. felineus, where I owe you an admission about sourcing. I could not verify either group assignment from the Monographs themselves, only from peer-reviewed restatements — and the apparent conflict I started with was not one, because Group 3 is the category whose published name is not classifiable as to its carcinogenicity to humans. Unclassifiable and Group 3 are one sentence, not two. Von Bülow and colleagues, in Cells in 2021, write that infection with S. mansoni is included in Group 3, indicating insufficient evidence to determine its carcinogenicity. Jain, in 2026, writes that O. felineus is currently placed in group 3, and argues it should be moved up.

Why those two sit where they sit is a matter of record. Berry and colleagues, in a 2017 review, put it flatly: S. mansoni and S. japonicum have been suspected to be associated with liver or colonic carcinomas, but epidemiological studies have not yielded any firm evidence so far. And for O. felineus, Fedorova and colleagues went looking in the best available national data: official Russian medical statistics across all 83 units of the Federation, 2011 to 2013. The fluke ran at 24.7 cases per 100,000 a year, reaching 599.7 in the Khanty-Mansiysk district and none at all in 27 of the units; liver and intrahepatic bile duct cancer ran at 4.8 per 100,000. The fluke's incidence was not associated with the cancer's: a correlation of 0.20, p value 0.07 — a p value being the probability of seeing a result at least this extreme if nothing were really going on, with 0.05 the conventional threshold.

So: a null result, just outside the line — and I want to be careful in both directions, because this is the weakest design in the toolkit. Comparing rates between whole regions rather than individuals is an ecological comparison, and anything that differs between regions — diagnostic habits, registry quality, age structure, hepatitis prevalence — can create or erase a correlation. A null from an ecological study of routine statistics is weak evidence of absence, not proof the fluke is harmless.

And the last parasite this agency has evaluated at all: the malaria parasite Plasmodium falciparum, in 2012, for Burkitt lymphoma — a fast-growing cancer of the B cells, the white blood cells that make antibodies. Bouvard and colleagues' two-page announcement in The Lancet Oncology carries no abstract, so the group comes from elsewhere in this same source list: Rumgay and colleagues record P. falciparum as currently Group 2A, probably carcinogenic. The malaria parasite is not on the Group 1 list, and that is the fourth parasite rating people get wrong.

The arithmetic of the canon

Rumgay and colleagues published IARC's current accounting in The Lancet Oncology in 2026, verifying what is on the list today. It quantifies 12 infectious agents classified as Group 1 carcinogens by the Monographs program: Helicobacter pylori; human papillomavirus; hepatitis B and C; Epstein-Barr virus; Kaposi's sarcoma-associated herpesvirus; Schistosoma haematobium; human T-cell lymphotropic virus; Opisthorchis viverrini; Clonorchis sinensis; Merkel cell polyomavirus; and HIV. The full list now runs to thirteen: hepatitis D became the thirteenth out of that June 2025 Working Group — Haennel and colleagues confirm the 2025 classification — though, as Rumgay and colleagues note, its cancers already sit inside the hepatitis B count. That announcement carries no abstract, so what the other two received I will not state.

Three of those twelve are parasites. The same three. No others.

The totals: an estimated 2.3 million new cancer cases worldwide in 2024 were attributable to infection — 12 percent of all cancer cases. The largest contributor is H. pylori, a stomach bacterium, at 760,000 cases; then human papillomavirus at 750,000; hepatitis B at 360,000; Epstein-Barr virus at 260,000; hepatitis C at 160,000. Eastern Asia carries 990,000 cases, 42 percent of the global total, at an age-standardized incidence rate of 31.9 per 100,000 against a global average of 22.7 — an age-standardized incidence rate being one adjusted so populations with different age structures can be compared fairly. The second-highest rate in the world is sub-Saharan Africa, at 28.5, ahead of central and eastern Europe at 24.3 and southeastern Asia at 23.1. Hold that against what comes next about the registries, because it is the same continent. The tool underneath all of it is the population-attributable fraction: the share of a population's cases that would not have occurred without a given cause — a modeled quantity, not a count.

And the parasite share? In the body of that paper, not its summary, the three Group 1 parasites come to roughly 9,700 cases — about 5,900 bladder cancers for the blood fluke and about 3,800 intrahepatic bile duct cancers for the two liver flukes. Out of 2.3 million. Four-tenths of one percent of the infection-attributable total.

Now. Why that small number is not a verdict.

Those estimates are built on cancer registries, and a registry is a system that records every new cancer diagnosed in a defined population. Without one, cancers are estimated rather than counted. The paper's own limitations are blunt: the highest-quality registries it draws on cover about 19 percent of the world's population — and about 2 percent of Africa's. Africa is where Schistosoma haematobium mainly is. The authors published no uncertainty intervals, and note that methods for identifying bacterial and parasitic carcinogens are less well standardized than those for viruses.

Read those together. The parasite number is an estimate with no stated error bars, built on registry data thinnest exactly where the parasites are thickest, by methods the authors call less standardized than the ones used for viruses. That does not make it wrong; it makes it a number with unknown error in both directions, not a measured ceiling and not a proven undercount. Anyone who tells you the parasite burden is secretly enormous is going beyond this data. So is anyone who tells you 9,700 is the answer.

Who went first, told without sneering

The usual telling starts with Ferguson in 1911. Berry and colleagues' 2017 review records what that paper actually was: A. R. Ferguson, professor of pathology and microbiology at the Faculty of Medicine in Cairo, published a detailed survey from 40 autopsies and reported a likely association of bladder carcinoma with the inflamed nodules — granulomas — caused by urinary schistosomiasis. Forty autopsies — not a massive series, and not proof. What it can honestly be called is the first systematic autopsy series linking a parasite to a human cancer.

And Ferguson was not first to suspect it. Berry and colleagues state flatly that the link between urinary schistosomiasis and bladder carcinoma was first suspected by C. Goebel in 1905. Goebel's work is an inaugural dissertation for the Universität Breslau, printed in Breslau in 1905, running 165 pages, on the bladder tumors occurring in bilharzial disease with particular attention to carcinoma. Bilharzial is the older word: bilharzia and schistosomiasis are the same disease, infection with the blood fluke. I am specific because the version circulating — an 1866 Würzburg thesis with a different title — is wrong on the city, on the year by nearly forty years, and on the title. Those catalog details come from a library record, not the original.

And earlier still, with a thinner thread. Reginald Harrison is reported to have found carcinoma in four of five bilharzial bladder specimens sent to him from Egypt in 1889, and to have asked whether bilharzia, cancer and bladder stone were causally linked. That account traces to Brumpt's 1930 paper on the role of bilharzias in the production of certain cancers in Annales de Parasitologie. Harrison's own 1889 publication could not be located for this episode. Its chain of custody reaches us through a French parasitology paper written forty-one years later, and nobody in this chain has read Harrison.

What all three of those men had was an association. Berry and colleagues record how long it took to become more than that: evidence of a positive correlation between urinary schistosomiasis and bladder carcinoma was delivered only many decades later, following case-control studies adjusted for age, sex, type of dwelling and tobacco consumption. And the mechanism is still not settled — the same review calls it poorly understood, partly for lack of a convenient animal model.

In 1926 the Nobel Prize in Physiology or Medicine was awarded to Johannes Fibiger, a Danish pathologist, for the discovery of the Spiroptera carcinoma. Fibiger had found a nematode — a roundworm — in the stomachs of rats, fed infected arthropod intermediate hosts to more rats, and produced what he called gastric carcinoma in the rat forestomach. Petithory and colleagues, in Histoire des Sciences Médicales in 1997, record that he was professor of pathological anatomy from 1900 and died in 1928 of cancer of the colon. It was the first Nobel Prize given for the idea that a parasite causes cancer.

It was wrong.

And the way it came apart is the lesson. A year before the prize was awarded, Wolbach and Howe published in the Journal of Experimental Medicine what happens when you deprive rats of fat-soluble vitamin A. Normal lining tissue is replaced by stratified keratinizing epithelium — layers of flattened, hardened cells, the kind that belong on the outside of you, growing where soft lining should be — arriving through focal proliferation, cells dividing in patches. Growth activity, they wrote, is not diminished but greatly augmented. And in a few animals the number of dividing cells, and the response of the surrounding connective tissue and blood vessels, suggests the acquisition of neoplastic properties. The confounder was in print in 1925. The prize was awarded in 1926. Nobody connected them.

The systematic re-examination came in 1952, twenty-six years later, from Hitchcock and Bell in the Journal of the National Cancer Institute, in a paper titled: studies on the nematode parasite Gongylonema neoplasticum and avitaminosis A — vitamin A deficiency — in the forestomach of rats, with comparison to Fibiger's results. That record carries no abstract in the databases, so I will not quote numbers from it. What the historical record says it established is that Fibiger's rats were vitamin A-deficient and his lesions were overgrowth misread as cancer. Petithory and colleagues list the three problems found later: confusion over which Gongylonema species was involved, the diet's lack of vitamin A, and the impossibility of reproducing his results.

The institution that gave the prize has published its own correction. Stolt, Klein and Jansson, at the Karolinska Institute, titled their 2004 study in Advances in Cancer Research “An analysis of a wrong Nobel Prize — Johannes Fibiger, 1926: a study in the Nobel archives.” A wrong Nobel Prize, in the title, from the Karolinska.

Here is why I will not let this one go. Fibiger had a real worm, a real lesion, and a working experimental model — more than most hypotheses in this field have ever had. And he was still wrong, because he missed a dietary confounder and misread reactive overgrowth as malignancy. Everything this series argues about parasites being under-examined has to be held to a standard that would have caught Fibiger. That is not a concession to skeptics but the only way the claim could ever be believed.

And the honest coda, from Petithory and colleagues: the cancer-causing action of some parasitic helminths, such as Schistosoma, is now recognized. The specific claim was wrong. The general idea turned out to be right, about different organisms, on much better evidence, sixty-eight years later.

This year, 2026, is the centenary of that prize.

The frame this series carries

Three organisms have cleared the highest evidentiary bar a cancer agency sets. That is not fringe and not contested, and any physician who treats parasites and cancer as quackery is unaware of their own literature. Equally: hundreds of millions of people carry worms for decades and do not get cancer from them, and the best counter-evidence on this episode's anchor organism is a mouse experiment in which infection made tumors less likely.

The most interesting thing published on this in the last year comes from inside the field, and cuts against the loose version of the argument. Heneberg, in One Health in 2026, starts from the observation that helminth infections are among the most prevalent causes of chronic inflammation in humans while only a small subset is causally linked to cancer — and says outright that this discrepancy challenges the long-standing assumption that chronic inflammation is intrinsically carcinogenic. He proposes a pro-oncogenic inflammation threshold: carcinogenesis emerges only when several dimensions converge at once — sustained inflammatory intensity and quality, prolonged exposure, genotoxic stress, tissue-specific vulnerability, and permissive environmental cofactors. Most worms stay below it, he argues, because evolution favors host survival, because they carry no direct genotoxins, and because their immune regulation is effective. The liver flukes exceed it through five things at once: chronic mechanical injury to the bile ducts; secretions that both damage DNA and push cells to divide; synergy with dietary nitrosamines, the DNA-damaging compounds of preserved and fermented food; a disturbed bacterial community in the bile ducts; and the failure of the tissue's own repair machinery. His conclusion, verbatim: chronic helminth-induced inflammation is therefore a conditional, rather than universal, driver of cancer.

That is a hypothesis and a framework, not new data, from a single author — but it is the most useful sentence in the field, because it is falsifiable, and because it replaces “inflammation causes cancer,” which is too loose to be wrong, with five conditions that have to coincide.

So the question this series asks is not whether parasites cause cancer. Three of them do, and the record goes back to 1994. The question is whether three is the real number, or only the number anyone has looked for. The agency that keeps the list has not evaluated a parasite since 2012, or a fluke since 2009; when it did return to infectious agents in June 2025, all three were viruses. The burden estimate is modeled from incidence data whose highest-quality registries cover about 2 percent of the continent where one of the three lives. And the one case in which a parasite's own cells turned malignant inside a person was found because somebody ordered a test that asks what organism this is.

Nine parts follow. Part two takes the liver fluke and the gene-editing experiment: how strong the causal case for bile duct cancer is, and exactly what was cut out of the worm, what changed, and what that proves and does not. Part three is the blood fluke and the bladder, where the cancer the worm is famous for causing changed type as the worm was driven back, and where we ask how a parasite reaches a human cell's DNA at all. Part four walks the specimen from the body to the slide, every step at which a parasite could be lost. Part five is the tests somebody has to order, and the iodine question — the idea that iodine used on a specimen could destroy a parasite before anyone looks, which somebody asked but nobody has published, and part five says so out loud. Part six is fuel. Part seven is charge, and part eight is movement. Part nine is the drug this question keeps running into, and the arithmetic of a dose. Part ten is the animals that already show the thing we are arguing about.

The Atlas carries the full account of the institution we just walked through, on its own page for the agency, with separate pages for the dwarf tapeworm, the blood flukes, the East Asian fluke whose fifteen-year wait we traced, bile duct cancer, how cancer is diagnosed under a microscope, the chromosome studies part five needs, and the iodine question part five takes up.

Sources

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