A Quarter of the Dogs, and the Five Resemblances

Part eight leaves human medicine for the one species where the experiment has half been run. In dogs, the nematode Spirocerca lupi walks up the wall of three arteries, scars the thoracic aorta, crosses into the caudal esophagus and builds itself a residence out of the dog's own actively dividing fibroblasts. In 120 dogs with spirocercosis from South African and Israeli records, 32 of those residences were sarcomas. By the time the tumor exists the worm is usually gone. We take the case apart: the life cycle as it was measured rather than drawn, the three histological stages in which the inflammation recedes as the cancer arrives, the denominator that makes "a quarter of infected dogs" the wrong sentence, the sequenced genome and the catalogued secretions that contain no growth factor, and the sister worm that gives 163 foxes nodules and no cancer at all. Then the two reasons nobody has finished the job — one of them biological, because the worm has no laboratory host. Then the five resemblances offered in the experiment's place: nutrient theft, cachexia, syndromes at a distance, the efflux pumps and self-renewal. Each is scored on its own, because resemblances do not add up.

Show notes

A dog brings its dinner back up — not quite vomiting: the food returns undigested, and in seventeen client-owned Israeli dogs that sign was present in ninety-four percent (Ranen 2004). In the chest is a mass at the lower end of the esophagus, the muscular tube from throat to stomach. A roundworm built it — a nematode, the group that also holds hookworm, Ascaris and Toxocara; flukes and tapeworms are different animals, which matters later. This one is Spirocerca lupi. It built its residence out of the dog's own cells, and a 2007 review puts the histology in one line — the mature esophageal nodule, a nodule being a firm lump of inflamed fibrous tissue, is mostly dividing fibroblasts, the cells that make collagen and do repair work (van der Merwe 2007). The worm's house is the dog's repair tissue, already dividing.

About a quarter of those residences become cancer — a quarter of the ones in dogs sick enough to reach a referral hospital, which is a sentence I will have to come back to. A sarcoma is a malignant tumor of connective tissue, bone or cartilage; a carcinoma arises from a lining epithelium. This worm makes sarcomas; the human liver flukes make carcinoma of the bile duct. All seventeen of those dogs had an esophageal sarcoma — nine osteosarcomas, five fibrosarcomas, one too disordered to classify — yet eggs were found in two of the eight whose stool was tested, and an adult worm in one of the eleven that went to necropsy (Ranen 2004). By the time the tumor exists, the cause has usually gone. Looking inside the tumor for the worm is the wrong test.

Two jobs this hour. The first is this dog: the strongest real-world evidence anywhere that a worm causes a cancer in a mammal, and a narrow piece of it: one nematode, one site, one lesion, and nobody has produced the tumor on purpose. A review in Trends in Parasitology is titled “Insights on Spirocerca lupi, the Carcinogenic Dog Nematode” (Rojas 2020), while a mechanism review partly by the same authors says the process remains poorly understood (Porras-Silesky 2021): a claim about an association, not a pathway. The second job is the five resemblances offered alongside it, which score differently: one is shared mechanism, two are real biology with the causation running backwards, two are vocabulary.

The life cycle, measured rather than drawn. Adults live in those nodules; the female burrows an opening through the lining into the lumen — the open channel of the tube — and lays eggs there that already contain a larva, and they leave in the feces (Bumby 2017). A dung beetle eats them: the beetle is the intermediate host, where a required developmental step happens, and the dog, where the worm matures and reproduces, is the definitive host. A year of dog dung pads collected in two Israeli parks, every beetle dissected, found one scarab the only species infected, the infective third-stage larva needing at least seven days inside it (Gottlieb 2011). The list is not closed, though a cricket and a cockroach fed infective eggs yielded none (Doanh 2023, Mukaratirwa 2010).

Then the dog. The larva leaves its cyst in the stomach, penetrates the gastric lining, and travels — not loose in the bloodstream but inside the artery walls themselves, to the caudal thoracic aorta, where it molts to the fourth larval stage, matures, and crosses to the lower esophagus (Bumby 2017, van der Merwe 2007). The aorta pays in scar. The prepatent period — infection to eggs in the feces — is three to six months, which is why a stool test comes back empty in a dog already very sick.

The clock on the migration was run by a drug trial, not by watching. Twenty-one purpose-bred beagles were infected with about forty third-stage larvae each and given milbemycin oxime, a dewormer, from day twenty-eight. A single dose was 79.8 percent effective at preventing establishment in the esophagus and repeat dosing a hundred percent effective, but none of it reduced the aortic damage, and the authors say why: dosing began only after the larvae had completed the first stage of migration (Kok 2011). By day twenty-eight the arterial leg is over. A controlled experimental infection in dogs, the strongest design here, run by a contract research organization testing a commercial dewormer.

The stages between nodule and sarcoma have been counted: sixty-two nodules, forty-two non-neoplastic and twenty neoplastic, ten non-overlapping high-power fields in each. The mitotic index first — how many cells are caught in the act of dividing per fixed area of tissue, conventionally per ten high-power fields, the pathologist's crude speedometer. It ran 1.31 in the non-neoplastic nodules against 42.85 in the sarcomas, p less than 0.01, and multinucleated giant cells from 0.9 to 13.9. A score built from fibroblast numbers and activation minus collagen then split the non-neoplastic nodules in two, same inflammation and necrosis but mitotic indices of 0.26 against 1.89: an early inflammatory stage, and a pre-neoplastic stage of activated fibroblasts and reduced collagen (Dvir 2010).

And here is the arrow pointing the wrong way. The inflammation score fell as malignancy rose, 1.91 to 0.97, while necrosis rose and the lymphocyte-and-plasma-cell infiltrate gave way to neutrophils (Dvir 2010). Chronic irritation is the story usually told about this lesion; in the tissue, the irritation recedes as the cancer arrives. One caveat covers the whole staircase: these are nodules from different dogs at different moments, sorted by appearance and arranged into an order. No study follows a single nodule forward.

Now the number, which comes from records, not a cohort. Of a hundred and twenty dogs diagnosed with spirocercosis at referral centers in South Africa and Israel from 1991 to 2008, thirty-two had neoplastic nodules — 26.7 percent — fifteen osteosarcomas, fifteen fibrosarcomas, two anaplastic sarcomas, and the literature rounds that off to a quarter of cases, a quarter of infections, a fifth (Mukorera 2011, Dvir 2011, Porras-Silesky 2021, Bumby 2017).

Then the correction this show has to make out loud. Those are shares of referred cases: about a quarter of the dogs sick enough to reach a teaching hospital, never a quarter of infected dogs, and nothing in this literature estimates risk in an unselected infected population. On Reunion Island, eighty-eight dogs were recruited from a clinic and a municipal shelter regardless of clinical status; twenty, 22.7 percent, were positive by at least one method, and the methods disagreed — endoscopy fifteen of thirty-one, flotation six of forty, a stool DNA test six of eighty-eight (Malenfant 2026). The first denominator here that is not a referral denominator, though the paper does not report how many of the twenty positives were incidental.

Two findings teach test performance better than any textbook. The first, serum alkaline phosphatase, does not separate the two things you need separated, P equals 0.18 with substantial overlap (Mukorera 2011). The second is in the dog's legs: across thirty-one benign and thirty-one malignant cases, hypertrophic osteopathy — painful new bone laid along the long bones of the limbs, driven at a distance by disease in the chest — appeared in 38.7 percent of the malignant dogs and none of the benign, very specific but relatively rare in the authors' words (Dvir 2008). Every other sign in that series, spondylitis — inflammation of the vertebrae the worm has passed — among them, also appeared in benign cases. And the plain film is weaker than it looks: in 59 Australian dogs, radiographs showed the esophageal mass in 59 percent, and in 94 percent of those only after sedation or endoscopy had filled the esophagus with gas (Thong 2025, retrospective, multicenter) — which also puts this worm on a fourth continent.

What nobody knows, and what nobody has done

Now the part where the honest answer is that nobody knows. A 2024 draft genome — a first, incomplete assembly of the worm's DNA — gave 13,627 predicted protein-coding genes, and what it turned up was ordinary: the targets dewormers already hit, a few vaccine candidates, and more predicted only by software (Rothmann-Meyer 2024). Nothing oncogenic fell out.

And the negative results are the load-bearing ones. The worm's secretome, catalogued by mass spectrometry, comes to two hundred and eleven proteins across the stages, rich in collagen trimers — equipment for moving through tissue, and nothing that tells a cell to divide (Rojas 2019). The mechanism review reports that those secretions were not mitogenic for mouse fibroblasts — mitogenic meaning able to make cells divide — and that no growth factor or pro-oncogenic molecule has been detected at any stage (Porras-Silesky 2021). Genome in hand, secretome in hand: no mitogen, no oncogene, no pathway. A short list of candidates tested in glassware is not proof of absence, and it is still the most important thing known here.

The host's own signals rise, and are not specific. Stained by immunohistochemistry — which marks one chosen molecule and leaves the rest unstained — those nodules showed fibroblast growth factor climbing across the stages, 0 to 8 to 34.5; then the controls ruin it, ordinary sarcomas with no worm behind them scoring 118 (Dvir and Clift 2010). A feature of being a sarcoma, not of being a worm's sarcoma. Plasma interleukin-8, the signal that calls neutrophils into tissue, rises with the disease — 634 picograms per milliliter where the nodule had turned malignant, 429 where it had not, 150 in controls, p equals 0.001 across three groups — but that paper's own post-hoc test separates infected dogs from controls, not malignant from benign, and the one cytokine that did separate the two conditions ran the other way: interleukin-18 was higher in the non-neoplastic dogs, p equals 0.05 (Dvir 2012). Inside the nodule the authors state in print that the triggering factors warrant further investigation (Dvir 2011).

Which brings up an error to correct out loud, by its name. A chatbot transcript in this show's research files offers bone morphogenetic protein signaling as the mechanism. Those proteins are real, and a tempting answer for a tumor that arrives as osteosarcoma. There is no such work: the term does not appear in the mechanism review, a PubMed search for the two together returns nothing, and the only growth factors ever studied here are the three just named (Porras-Silesky 2021, Dvir and Clift 2010). That is a fabricated mechanism: a real pathway from elsewhere, attached to a question it has never been tested on. One more category belongs beside it: in the same files a model supplied a funder and a grant purpose for a paper whose full text carries no funding statement at all, only a declaration of no conflict of interest, and no author at the university named. Call that fabricated provenance.

The one real clue to the bone is histological, not molecular. In the South African jackal survey, sixteen of ninety-three animals had Spirocerca-damaged aortas, eleven showing fibroblast metaplasia with mineralization, cartilage and/or heterotopic bone in the muscular layer of the artery wall (Bumby 2017). In a lesion that is not yet a tumor, in a host where no sarcoma was found, the tissue around this worm is already making bone: a clue, not an answer.

Then the twist that breaks the simple story. It is the host, not the worm. Of thirty-nine coyote necropsy cases from Costa Rica, 1989 to 2024, thirty-three — 84.6 percent — had S. lupi-associated lesions, with a significant association between the nematodes and both esophageal granulomas, walled-off nodules of chronic inflammation, and aortic aneurysms, balloon-like weak spots in the artery wall. One animal had larvae in its spinal cord, the first such case reported in coyotes, and the worm's cox1 sequence, a mitochondrial species barcode, matched S. lupi from the Andean fox in Peru (Romero-Vega 2025). Eighty-five percent of the coyotes scarred; one fibrosarcoma, logged at under one percent, and the paper says that except for that one case no other neoplasia was diagnosed.

One correction here is internal to this show's files,, running the opposite way from the way I first had it. An earlier research summary says the first coyote fibrosarcoma appeared in a 2025 necropsy series. The fibrosarcoma is in that series. The word first is not: the paper claims a first for spinal-cord invasion in coyotes, not for the tumor (Romero-Vega 2025). Call that summary drift — a paraphrase that keeps a record's shape and swaps its content.

The other hosts say the same. Of ninety-three jackals, sixteen had aortic lesions, one an esophageal nodule, none a sarcoma (Bumby 2017). In Spain, of 163 red foxes infected with Spirocerca vulpis, a sister species split from S. lupi in 2018, 96.9 percent had nodules in the stomach region and no neoplastic forms were seen in the seventy-one examined histologically (Martín-Pérez 2022). A near-identical worm, a near-identical lesion, no cancer. On the foxes and the jackals, where neoplasia genuinely was not seen, the worm is not the carcinogen by itself; something about being a domestic dog is doing part of the work. There is a published frame for that: a 2026 conceptual review argues chronic inflammation alone does not suffice, and that the flukes which do cause cancer add mechanical injury, genotoxic secretions, dietary nitrosamines and a disturbed biliary microbiome on top of it (Heneberg 2026, review) — one reason the search here comes back empty, since Spirocerca has the inflammation and the proliferation and no detectable genotoxin.

Now the history. Act one, 1926: Johannes Fibiger's Nobel Prize for a rat carcinoma he attributed to a worm. What undid it: confusion over which species had infected the rats, diet, lack of vitamin A, and the impossibility of reproducing the results (Petithory 1997). In 1952 Hitchcock and Bell set the worm beside vitamin A deficiency in the rat forestomach — the control experiment he had not done, twenty-six years late — and in 2004 a paper from the Nobel archives at Karolinska was titled “An analysis of a wrong Nobel Prize” (Stolt 2004). The exact citation wording and award date could not be checked against the Nobel Foundation, so neither is read on air. And one thing the error did not do is found a theory. Chronic irritation was Virchow's, set out decades earlier; Fibiger's worm was offered as support for it, not as its origin. The line that lasted was chemical — Yamagiwa's coal-tar carcinomas in 1915, and Fibiger's own later tar work — and no record I retrieved credits his mistake with the later science of infectious cancers. Those dates come from the standard histories, not from a paper I opened.

Act two, 1946, is the harder experiment: “Multiple peritoneal sarcoma in rats from intraperitoneal injection of washed, ground Taenia larvae” (Dunning 1946) — dead, ground-up worm producing malignant tumors on purpose, which nothing done with Spirocerca comes close to (Dunning 1953, Hanes 1995, Mahesh Kumar 2006). One item I am labeling rather than reading as sourced: Bullock, Curtis and Rohdenburg's 1920 report of five hundred rats with liver tumors is not in PubMed and was not opened. Act three, 1955, and the title is the correction: “Observations on the possible relation of malignant esophageal tumors and Spirocerca lupi lesions in the dog” (Seibold 1955). Possible relation. Four more followed from the same circle through 1972 (Ribelin 1958, Thrasher 1963, Bailey 1963, Bailey 1972). So the claim that this worm was the first definitive proof in mammals that a worm could directly cause a malignant tumor is wrong twice over. Not first: Fibiger's prize came twenty-nine years earlier, the deliberate rat sarcomas nine. Not proof: the founding paper says possible relation in its own title, and its author called his own retrospective a continuing inquiry. What holds is the geography — the founding papers really are American, and a prevalence of forty-seven percent in rural Alabama dogs is cited in that literature, as cited rather than as opened.

Which leaves why the experiment has never been run. Experimental infection works: twenty-one beagles given about forty L3 larvae each got aortic and esophageal lesions on schedule, and every dog was euthanized at day 168 or 169 (Kok 2011), long before a sarcoma could appear. Koch's postulates are the nineteenth-century checklist for proving an organism causes a disease: find it in every case, grow it in pure culture, give that culture to a healthy animal and reproduce the disease, recover it again. The third is unmet — nobody has given this worm to a healthy dog and produced the sarcoma — and the first fails too, since the worm is usually gone by the time the tumor exists. So the strongest worm-causes-cancer case in any mammal rests on an association plus a staged histological progression: a great deal, and not a fulfilled experiment.

The reason is substantially biological, not institutional. In paratenic hosts the third-stage larva re-encysts and develops no further, so mice, rats and rabbits are dead ends and the cycle cannot be run on a bench (Porras-Silesky 2021). It would take purpose-bred dogs, deliberately infected, kept for years — and the field's own reviews call this an excellent and under-utilized spontaneous model (Dvir 2011). Under-utilized, not forbidden. One absence to state straight alongside it: PubMed holds no human case of Spirocerca infection. What the search does return is relatives — human cases of Thelazia callipaeda, the eyeworm in the same superfamily, turn up regularly (Tian 2022, Li 2025). For Spirocerca itself there is nothing: an absence of evidence in an indexed literature, not proof that it cannot happen.

What actually moves

The comparison survives at the chemistry. Ac-MTP-1 is an astacin-like metalloprotease — a protein-cutting enzyme needing a metal ion to work — secreted by infective hookworm larvae, and antiserum against it cut collagen digestion 85 percent and larval migration through tissue in a dish 70 to 75 percent (Williamson 2006): recombinant protein in a dish, not an animal experiment. Which barriers a worm's enzymes can cut is what sets its invasiveness (Hotez 1990), and the tumor side of that chemistry is on the Atlas page about collagen and invasion.

Travel by blood is the second resemblance, and so is arrest. Schistosome larvae circle the circulation several times before “chance delivery in cardiac output” toward the liver (Wilson 2009), and a review describes the same random delivery and attrition for circulating tumor cells — cancer cells loose in the bloodstream — calling metastasis “highly inefficient” (Massagué 2016). Toxocara canis larvae accumulate in the brains of mice, paratenic hosts, and two inbred strains show that where the larva ends up depends on the host's genetics, though the authors warn their cerebral counts varied between experiments, so larval burden may not be a reliable index of susceptibility (Hamilton 2006). In a survey of 602 former Far East prisoners of the Second World War, 88 had Strongyloides stercoralis and another 25 unexplained eosinophilia; Gill and Bailey went back to that second group nine to nineteen years later and, of the 11 still alive and traceable, five had strongyloidiasis (Gill 1989, Verburg 1990) — which rhymes with the “latency periods that range from years even to decades” of cancer dormancy (Aguirre-Ghiso 2007).

The one human record of hypobiosis — a larva pausing its own development — is a single page with no abstract indexed (Prociv 1995): one case report suggests it, no more. And dormant larvae reawakening at pregnancy is the dog's biology smuggled into the cat — “there is no evidence for the existence of arrested somatic larvae in the adult cat” (Coati 2003). Now where it breaks, and the break is purpose. Larvae in genera such as Ascaris “undergo extensive migrations which begin and end in the same location,” and worms with a tissue phase grow bigger than their gut-only relatives, so since egg output tracks size the detour pays — “analogous to the pre-spawning migrations of salmon” (Read 1995). A larva migrates for its own fertility, to a fixed address, and does not divide there.

Metastasis, by contrast, has no program and is grossly inefficient: of melanoma cells injected into the portal vein of mice, eighty percent extravasated — got out of a vessel into tissue, the far end of intravasation, which is getting into one — but only one in forty formed a micrometastasis and one in a hundred of those a macroscopic tumor (Luzzi 1998). One cell line, one mouse strain, 1998, not human metastatic efficiency. Three textbook claims nearby are also wrong: “Are 90% of deaths from cancer caused by metastases?” found 66.7 percent (Dillekås 2019); the supposedly obligatory pulmonary route for Strongyloides is one of several (Schad 1989); and Ascaris suum larvae invade the pig cecum and colon, not the small intestine (Murrell 1997).

The sharpest image is in the dog. Esophageal nodules turn neoplastic with metastasis to lungs, kidneys, stomach, spleen, heart and tongue (Rojas 2020, Porras-Silesky 2021). Of 284 dogs whose larvae migrated into the spinal cord, 46 percent recovered completely at one month and 75 percent at three; retrospective, not a trial (Casoy 2024). Call that aberrant migration — a larva lost rather than waiting, turning up in the spinal cord, lung, kidney, bladder or skin. The lost one also perforates gut: of 49 dogs with septic peritonitis, 21 were this worm's doing, with eosinophils in the abdominal fluid in 80 percent against 5.6 percent of the other cases, and survival to discharge better rather than worse (Berger 2026, retrospective). A larva in the wrong place does damage and stops. A cell in the wrong place may build a colony. Clonal expansion is the whole of the difference.

Cats keep reservoir and victim apart. They are called “the most important reservoir” for the liver fluke the World Health Organization's cancer agency calls carcinogenic to people — and the same review states that the cholangiocarcinoma it causes, cancer of the bile ducts, “has not yet been reported in the reservoir animals” (Tangkawattana 2018). Two claims around it need correcting. That the fluke's first description in a Paris zoo cat confirms a long evolutionary history with cats is a non sequitur; what exists is population genetics, eggs from people and cats in one Thai sub-district forming largely separate populations (Sota 2022). And in the feline fluke life cycle, lizards are paratenic rather than obligatory hosts, with a snail first intermediate host in South America and isopods the natural second (Pinto 2014). Feline bile-duct cancer is documented for other flukes in small numbers, none of it measuring how often infection leads to cancer; these are necropsy series (Hou 1964, Li 2024, Andrade 2012, Braga 2016), and the flukes belong to the liver-fluke and cholangiocarcinoma Atlas pages.

The nearest thing to a cell that really moved between hosts is not a parasite in the ordinary sense. The canine transmissible venereal tumor is genetically distinct from its host in every case examined, tumors from 40 dogs on five continents are one clone, and it is the oldest known somatic cell lineage, with no convergent vincristine resistance despite routine treatment since the 1980s: “neutral genetic drift is the dominant feature of long-term cancer evolution,” drift being what happens when neither kind of change is being selected for or against (Murgia 2006, Murchison 2014, Baez-Ortega 2019). The Tasmanian devil tumor is the same thing as a graft — an allograft, tissue from another individual of the same species, which histocompatibility, the molecular system of self-recognition, normally rejects — and there the first explanation was corrected by its own authors: the loss of self-recognition molecules turned out to be reversible epigenetic change, and a second devil cancer then appeared that expresses them, “demonstrating that loss of MHC is not necessary” (Siddle 2007, Pye 2016, Caldwell 2018).

In the sea it has happened repeatedly, and once across species: every cancer cell in the golden carpet shell clam came from a clam living alongside it, a lineage now above 75 percent at two Puget Sound sites against a 1 to 5 percent Atlantic background, with a fourth type added in catfish (Metzger 2015, Metzger 2016, Weinandt 2026, Curd 2026, Stammnitz 2023). All of this is the host's cells moving, not the parasite's — the opposite direction from the Medellín case in part one. And the SCANDAL hypothesis, that some simplified animal lineages descend from ancient transmissible tumors, has no direct test, so I am naming it a hypothesis, which is what its authors called it (Panchin 2019). The karyotype Atlas page carries the numbers, and transmissible cancer has a page of its own beside it.

Hunger, wasting, and the sign that cuts both ways

Test one is nutrient theft, where the parallel is real and the usual telling has the arrow backwards. Auxotrophy means being unable to make something you cannot live without, so you take it in. Heme is the case: the iron ring at the center of hemoglobin and the oxygen-handling enzymes. Caenorhabditis elegans, a free-living soil nematode that parasitizes nothing, cannot make heme and must eat it (Rao 2005). Its importers are conserved far enough that worm HRG-1 rescued blood formation in heme-starved zebrafish, and the same system runs in parasitic nematodes (Rajagopal 2008, Luck 2016, Zhou 2020).

The dates carry the argument. Sterol dependence was shown in 1968 and heme dependence in 1970, in a sister species, Caenorhabditis briggsae, the heme work in a chemically defined axenic medium with no living organism but the worm (Hieb 1968, Hieb 1970). The shopping list is ancestral to nematodes: parasitism exploits a dependency rather than creating one. That is the causation error. Two more claims need correcting by name. No published hypothesis explains nematode auxotrophies by superoligotrophy; the only indexed use of the word is a 2026 paper on nematode-trapping fungi losing a quinone-synthesis gene in a cold “superoligotrophy” period (Wu 2026) — a different organism, a different question — so as this corpus states it, the hypothesis has no source. And no paper totals any worm's required nutrients in the low to mid-forties: what is published is nine amino-acid pathways lost in animal genomes and ten requirements measured for one nematode (Payne 2006, Hata 1994).

One nuance before the tumor side, and it is about sterols rather than heme: in Ascaris, squalene — the oily precursor a sterol ring is built from — oxygen and a reducing enzyme assemble on that worm's oxygen-avid hemoglobin, which may run part of sterol synthesis itself (Sherman 1992). So nematodes cannot make heme; for sterols here the picture is unfinished.

The tumor side mirrors the dependency, and it became medicine. Asparagine dependence traces to 1953: guinea-pig serum made transplanted lymphomas regress in mice and stalled a rat lymphosarcoma, while mammary carcinomas and a fibrosarcoma grew unimpeded, and its active ingredient became asparaginase (Kidd 1953). Arginine cuts both ways: in 249 patients with non-epithelioid pleural mesothelioma, whose tumors lack the enzyme that makes arginine, depletion with chemotherapy gave 9.3 months against 7.7, hazard ratio 0.71 — how much faster an event happens in one arm than the other, 1.0 meaning no difference and under 1 slower — while in 635 patients with liver cancer it gave 7.8 against 7.4 (Szlosarek 2024, Abou-Alfa 2018). Both trials carry co-authors from the company that makes the drug. Some lymphomas must likewise import cholesterol (Garcia-Bermudez 2019). And the gate: tumors needing more leucine than they can make pull it in through a transporter called LAT1, and blocking it reached patients once — nanvuranlat, 105 randomized patients with refractory biliary tract cancer, progression-free hazard ratio 0.56, no survival benefit (Furuse 2024). The sensing inside the cell is not what it is usually said to be: leucine does not switch on mTORC1 by docking into it, but by binding Sestrin2 (Wolfson 2015, Han 2012).

One thread to dispose of. Pork is thiamine-rich, and the claim in this corpus is that the pig evolved a muscular thiamine buffer against parasite drain. It did not: pigs are monogastric, muscle thiamin tracks the feed (Shaw 2002), and no study in any species links muscle thiamine to parasite defense. Nor does allithiamine form in the gut from garlic plus seeds — none of the reactive sulfur compounds forms below pH 3.6, and stomach acid kills the garlic enzyme irreversibly (Lawson 1992). The real human tapeworm vitamin story is B12 (Scholz 2009).

Test two is cachexia, where the parallel is emotionally irresistible and factually wrong. The 2011 international consensus defines it as ongoing loss of skeletal muscle, with or without fat loss, not fully reversible by conventional nutritional support (Fearon 2011): feeding does not fix it, and that clause is the definition. The mechanism history starts in parasitology. Rabbits infected with Trypanosoma brucei ran high blood triglycerides in 1980 because the clearing lipase was deficient (Rouzer 1980); five years later that wasting factor, named cachectin for the syndrome, proved to be the same protein as tumor necrosis factor (Beutler 1985). What follows is mouse work: antibody to interleukin-6, but not to tumor necrosis factor, suppressed wasting in the colon-26 model, and removing the tumor let the mice regain weight (Strassmann 1992, from an Otsuka laboratory), with a fat-mobilizing lipase and a hormone that makes fat burn itself as heat filling in the chain (Das 2011, Kir 2014). The obvious drug failed twice in people (Jatoi 2007, Wiedenmann 2008).

Then the first causal reversal in people: blocking GDF-15 — growth differentiation factor 15, a signal that rises in cancer and suppresses appetite — with ponsegromab in 187 patients over 12 weeks gave a median weight gain over placebo of 2.81 kilograms, 95 percent credible interval 1.55 to 4.08, the Bayesian cousin of a confidence interval and not the same object. Pfizer-funded (Groarke 2024), and a phase 2 result: the phase 3 trials are only now starting and no cachexia drug is approved anywhere (Muscaritoli 2025). Now the undercut. If worms drained systemic nutrients into wasting, deworming at scale would move weight. It does not. The 2019 Cochrane review pooled 51 trials — 84,336 children across fifty of them, plus a mortality analysis of over a million — and found no improvement in height, hemoglobin, cognition, school performance or death; on weight it says plainly that it does not know, two trial sites from more than twenty years ago pulling the estimate with nothing since 2000 reproducing them (Taylor-Robinson 2019).

A Bayesian network meta-analysis of 1,108,541 children put the proportion stunted at eight fewer per 1,000, credible interval 48 fewer to 32 more, on high-certainty evidence — the top rung, meaning further research is very unlikely to change it (Welch 2017). That stunting figure belongs to the network meta-analysis, not the Cochrane review; the two get run together. So, by name: cachexia is not the tumor eating the patient's groceries but a host catabolic program, and the helminth cachexia analog does not exist as measured. What exists is mechanical — hookworm blood loss and iron-deficiency anemia, nitrogen absorption improving after deworming — plus one association, schistosome intensity going with lower albumin and higher inflammatory markers (Loukas 2016, Brown 1980, Coutinho 2006). Inflammation-linked undernutrition, not the tumor's program.

Test three is paraneoplastic syndromes: two tracks hold, a third collapses. One is exactly two things — a molecule the tumor secretes, or immune cross-reactivity with normal tissue (Pelosof 2010) — and parasites do well on one. The best case is schistosomiasis-associated pulmonary arterial hypertension, where the eggs lodge in the lung vessels and the adult worm is nowhere near the lung. Echocardiographic screening of every hepatosplenic patient in one Brazilian clinic — hepatosplenic meaning the ones whose liver and spleen are already scarred by the fluke — found raised pressures in 18.5 percent, where catheterization confirmed pulmonary hypertension, the raised pressure itself, in 7.7 percent and pulmonary arterial hypertension, the subset in which the small lung arteries themselves are the problem, in 4.6 percent (Lapa 2009): the screening test more than doubling the disease. Pooled hemodynamics are indistinguishable from idiopathic disease but survival is better (Knafl 2020), and the mechanism is mouse (Kumar 2017).

The purest analogue is tropical pulmonary eosinophilia — an eosinophil being the white cell that stains pink with a particular dye and is the immune system's standing answer to worms, eosinophilia being too many of them. Severe lung disease from an immune reaction to filarial worms, the thread-like ones carried by biting insects, that are nowhere near the lung; diagnosed partly by the organism's absence: eosinophils above 3,000 per microliter, no microfilariae — the worm's larvae in the blood — anywhere in it, IgE above 1,000 units per milliliter, and a response to diethylcarbamazine (Ong 1998). The control case is the rest of that literature: in Loeffler's syndrome and visceral larva migrans the larva really is in the lung or the tissue (Chitkara 2006, review). Taught as benign, it leaves reduced diffusion capacity, and about 20 percent relapse (Mullerpattan 2013).

The hard test the cancer side sets: does removing the cause end the syndrome? Both answers exist. Across 89 papers on hyper-reactive malarial splenomegaly, every treated patient with no further exposure improved on a short antimalarial course — the parasitic equivalent of a syndrome remitting after resection (Leoni 2015). The cancer-side mirror is a single case: an eosinophilic fasciitis that appeared with a myelodysplastic syndrome turning into leukemia, resolved when that was treated, then returned with a second and separate malignancy and resolved again (Haddad 2014, case report). Against both, benznidazole in 2,854 patients with Chagas cardiomyopathy cut detectable parasite sharply and changed nothing in the heart (Morillo 2015): some of this damage becomes host-driven and self-sustaining.

One parasite molecule does behave like an ectopic hormone. The plerocercoid — the larval stage that sits in muscle — of the tapeworm Spirometra mansonoides releases a cysteine proteinase with no homology to any growth hormone that binds the growth-hormone receptor, mimics it, and suppresses the host's own growth hormone — all rodent (Phares 1996, Salem 1989, Phares 1986). Convergent receptor agonism, not ectopic synthesis. And where the resemblance is only a turn of phrase, here is where it is. There is no parasitic inappropriate-antidiuretic-hormone syndrome: in 171 adults with severe malaria, antidiuretic hormone did not differ between hyponatremic and normonatremic patients, and the low sodium went with lower mortality (Hanson 2009). The only parasite evidence for PTHrP, the hypercalcemia molecule, is two dogs whose authors call it a first (Fradkin 2001), and canine hypercalcemia in American schistosomiasis shrank from 11 of 22 to 4 of 60 once a stool DNA test caught milder disease (Fabrick 2010, Graham 2021). Ascertainment, not calcium.

The reversal this topic needs. A 2017 paper proposed that nodding syndrome in Ugandan children is autoimmunity to a human protein, leiomodin-1, cross-reacting with Onchocerca volvulus — striking mimicry, called initial evidence by its authors (Johnson 2017). The antibody then turned up in 6 of 52 patients against 14 of 61 controls, p equals 0.113, with the protein in cerebral vessel walls rather than in neurons (Hotterbeekx 2021, Kodja 2023), and in 240 children six weeks of doxycycline changed no autoantibody level and no seizure outcome at two years (Idro 2024). A beautiful mimicry result is a hypothesis, not a finding.

Now the clinical service. Eosinophilia here means opposite things in different compartments, so the compartment has to be named every time. A woman with 6,530 eosinophils per microliter: in one telling, a worm; in the published case, a recurrent clear-cell kidney cancer (Zalewska 2021). In 14,298 returning travelers, eosinophilia had limited predictive value (Schulte 2002), while inside a tumor it reads the other way, tumor-infiltrating eosinophils predicting better survival across 26 studies (Hu 2020). Its absence can be the bad sign: in 339 published cases of severe strongyloidiasis, eosinophilia went with a better outcome (Olga Rojas 2023, a different author from the Alicia Rojas of the Spirocerca papers). And a normal count does not exclude organ damage: across 179 published, histologically proven cases of eosinophilic myocarditis gathered out of 443 screened manuscripts — case reports, the authors warning that published cases skew severe — 75.9 percent had blood eosinophilia, a quarter none (Brambatti 2017).

The asymmetry, then. Paraneoplastic neurology has named antibodies sorted by cancer risk, three grades of certainty, and a care score combining phenotype, antibody, cancer and follow-up (Graus 2021). The parasite side has one refuted antibody, one failed antibiotic trial, and real syndromes with no nomenclature. That gap is not evidence they are unreal: schistosomal pulmonary hypertension, tropical pulmonary eosinophilia and hyper-reactive malarial splenomegaly are real, and so is immune-complex glomerular disease in schistosomiasis, though treatment made it rare — in one Brazilian renal-biopsy survey only 4 percent of patients were positive for Schistosoma mansoni at all, and eleven of those twenty-four had the glomerular patterns associated with it (dos-Santos 2011). The famous counterexample, quartan malarial nephropathy, did not hold (Ehrich 2007). The gap is evidence of where the committees and the money went.

The pumps, the stem cells, and two promises

Test four is the efflux pumps, the strongest shared claim in this season. An efflux pump is a protein in a cell's membrane that spends energy pushing molecules back out. The parallel sits in one 1994 sentence: delete one mouse gene, mdr1a, and the animal looks normal but becomes 100 times more sensitive to ivermectin and three times more sensitive to vinblastine (Schinkel 1994). Now the collies: a four-base-pair deletion in the dog's MDR1 gene truncates P-glycoprotein, and dogs with two copies are the ivermectin-sensitive ones (Mealey 2001). They still have a blood-brain barrier; they lack the protein that works it. No record I opened gives an allele frequency, so I will not.

The protein has three names — P-glycoprotein, MDR1 and ABCB1. It belongs to the ATP-binding cassette family, ATP being the cell's energy currency and the cassette the part that spends it; humans have 48 of these transporters, most moving fats, bile and metals (Dean 2001). Multidrug resistance means resistance to several chemically unrelated drugs at once, because the pump recognizes a shape, not a molecule. Named in 1976 in colchicine-resistant hamster cells, the P standing for permeability (Juliano 1976), it is a protein whose mouse crystal structure holds a cavity of about 6,000 cubic angstroms with several drug-binding sites (Aller 2009) — in vitro. In normal human tissue it sits at excretory surfaces: bile canaliculi, kidney tubules, gut lining (Thiebaut 1987). Plumbing, not cancer.

The borrowing ran from oncology into parasitology,. A 1998 paper cloned a full-length P-glycoprotein from Haemonchus contortus, at 61 to 65 percent homology with the mouse, human and C. elegans proteins, citing the cancer field's knockout mouse in its abstract and reaching for verapamil, a cancer-ward pump blocker, to explain ivermectin failing in sheep (Xu 1998). On the tumor side the association is real: in 211 patients over 55 with acute myeloid leukemia, 71 percent of leukemias expressed the pump, and across the extremes — pump-negative new-onset leukemia with favorable chromosomes against pump-positive secondary leukemia with unfavorable ones — remission ran 81 percent against 12 (Leith 1997). The same group later found the pump predicting remission but not survival (Leith 1999), and 31 pooled breast cancer studies gave a relative risk of chemotherapy failure of 3.21, 1997 vintage, from assays changing mid-stream (Trock 1997).

Then the trials. Zosuquidar shut off more than 90 percent of pump activity in 16 patients' leukemia cells within two hours (Lancet and colleagues, 2008 — the first author's surname, not the journal; a phase I carrying co-authors from the company developing the drug), then changed nothing in 449 randomized older adults, 7.2 months against 9.4, p equals 0.281 (Cripe 2010). Valspodar did harm, one arm closing early with deaths at 44 percent against 20 (Baer 2002), because blocking the pump blocks the bile ducts, kidney tubules and gut lining that clear the drug. Tariquidar gave one partial response in 17 women (Pusztai 2005); its phase III lung trials are widely said to have been halted for toxicity, which I could not source anywhere, so I am not saying it.

One trial was positive, and omitting it would be the thing this show criticizes: intravenous cyclosporine in 226 poor-risk leukemia patients raised two-year survival to 22 percent from 12, p equals 0.046, concentrated in pump-bright patients — with that paper's own caveat that cyclosporine also raised daunorubicin levels (List 2001). Twenty-two years later parasitology reached the same place: targeting nematode P-glycoproteins “is hampered by the lack of selective inhibitors” (Lespine 2024). The protein that protects the host is the one you would have to block. A shared failure, not a shared cure — and the pump itself, 1976 to the modulator trials, has its own Atlas page.

The worm side gets its own reversal. Six pump genes were deleted one at a time in an ivermectin-resistant C. elegans strain; only deleting pgp-9 restored sensitivity, the Haemonchus version rescued the resistance, and a fluorescent ivermectin showed drug accumulating again once the pump was gone (Blancfuney 2026), across two further macrocyclic lactones and one unrelated drug, tunicamycin (Blancfuney 2025) — causal work, in a free-living laboratory worm that parasitizes nothing. In a real parasite in a real host it is murkier: the pump changes in resistant H. contortus “could not be enough to explain” the resistance (Maté 2018), and in a genetic cross the gene raised in every resistant population was a transcription factor (Laing 2022). Then the asymmetry: drug selection of pump alleles is documented in nematodes (Xu 1998, Prichard 2007, Williamson 2011), while on the tumor side I could not retrieve one paired before-and-after sequencing study showing clonal ABCB1 amplification in a patient.

Two distinctions keep this clean. Flubendazole is not a pump inhibitor: cells made resistant to vinblastine by overexpressing P-glycoprotein “remained fully sensitive” to it (Spagnuolo 2010). Evading a pump and jamming a pump are opposite pharmacologies, routinely collapsed into one sentence about antiparasitics beating drug resistance. And the ivermectin-reverses-resistance claim traces to a 1996 dye-retention assay in two leukemia cell lines, no concentrations in its abstract, whose last sentence says ivermectin “may” be a substrate and an inhibitor (Didier 1996) — while the strongest modern paper contradicts that mechanism, ivermectin working “not by directly inhibiting P-gp activity” but by reducing how much the cell makes (Jiang 2019). So I am narrowing what I said the first time: the promising result is cells in a dish, and what was abolished was expression during gradual drug exposure (Hayashi 2024) — prevention of induction, not a pump blocked.

The dosing arithmetic belongs to the ivermectin part, and the vertebrate-sparing half is on the ivermectin Atlas page, with the blood fluke's own pump. Where it is only rhetorical: not the same protein, nobody has put a nematode PGP-9 and a human ABCB1 in one assay, and the shared binding-site architecture is postulated rather than shown (Mani 2016). Test five, then, is self-renewal, where the architecture is parallel and the machinery is not — self-renewal meaning a cell divides and at least one daughter stays the same kind of cell, so the population keeps going. A neoblast is the dividing stem cell of planarians, a germinative cell the tapeworm equivalent, an EdU pulse a labeled DNA building block given for a fixed window so that whatever copies its DNA then lights up.

In the larva of Echinococcus multilocularis, the fox tapeworm that grows through a liver like a tumor, germinative cells were the only cells ever seen in mitosis — a fifth to a quarter of all cells, 24 percent labeled after a five-hour pulse — and hydroxyurea cut them to 3 to 5 percent, survivors regrowing the population in patches (Koziol 2014). The rest of the worm side is less tidy: 43,642 single schistosome cells sorted into 68 populations, including stem cells restricted to the gut (Collins 2013, Wendt 2020) — renewal in several local pools, not one tumor apex. And the line that all flatworms have neoblasts is refuted: the earliest-branching flatworm regenerates with division spread across differentiated tissues (Gąsiorowski 2025).

The cancer side at its real and contested strength. The claim is that a tumor is not a uniform mass but a hierarchy, and that a small subpopulation is the part able to rebuild the whole thing — a contested model, not settled fact. As few as 100 breast cancer cells with the right surface markers formed tumors where tens of thousands of others did not, and 100 brain tumor cells did it where 100,000 marker-negative cells did not (Al-Hajj 2003, Singh 2004, Bonnet 1997, Schepers 2012). Then the reversal: in more deeply immunocompromised mice about 25 percent of unselected melanoma cells taken straight from patients formed tumors, and 28 percent of single cells did, against prior estimates of one in a thousand down to one in a million, with none of 22 markers enriching them (Quintana 2008, Quintana 2010). The rare stem cell was partly a property of the assay, and non-stem cells revert spontaneously (Chaffer 2011). Ablating Lgr5-positive cells restricted growth without regression in mice; in human organoids the same ablation did produce regression, then regrowth as differentiated cells reverted (de Sousa e Melo 2017, Shimokawa 2017, Lan 2017).

Clinically the compartment has been easy to sell and hard to kill. Two phase 3 trials of napabucasin, marketed on stemness, failed outright: 11.4 months against 11.7 in metastatic pancreatic cancer, 1,134 patients, terminated for futility (Bekaii-Saab 2023); 14.3 against 13.8 months in colorectal cancer, 1,253 patients, with phosphorylated STAT3, the biomarker it was aimed at, marking worse prognosis in the control arm rather than predicting benefit (Shah 2023). The one success is narrower than it sounds: glasdegib with low-dose cytarabine, 8.8 months against 4.9, in a randomized phase 2 of 132 unfit patients (Cortes 2019) — a pathway drug working, not proof a compartment was eradicated. The shared failure mode is the real finding: benzimidazoles shrink Echinococcus without clearing the germinative pool, so growth resumes off therapy, while a cell-division kinase inhibitor taken into human cancer trials kills germinative cells in culture (Schubert 2014). Restricting without eradicating, for the same reason.

Where it is metaphor. Tapeworm genomes have lost 34 homeobox families and several determinants of stem-cell fate, and the differences from planarian neoblasts are called fundamental (Tsai 2013, Koziol 2014): a cancer stem cell subverts intact human machinery, a germinative cell runs a reduced one. The worm's self-renewal is licensed physiology; the tumor's is a failure of control. A 2026 review sharpens it: planarian neoblasts carry constitutive telomerase — the enzyme that rebuilds the protective caps at chromosome ends, so a cell line can divide indefinitely — switched on as ordinary housekeeping, and planarians have no documented spontaneous malignant neoplasms (Diaz 2026). Pushed, they do transform, by chemical cocktail or by knocking down PTEN, a tumor-suppressor gene (Hall 1986, Oviedo 2008). That absence comes from a review, not a histopathology survey of aged planarians, and I am not going to upgrade it.

Two more cautions, then the promises. The New England Journal report never says the malignant Hymenolepis cells were germinative cells (Muehlenbachs 2015); the inference is reasonable, since in cestodes germinative cells are the only proliferating cells, but it is an inference, n of one, with profound immunosuppression attached. And plainly: no published paper draws this neoblast-to-cancer-stem-cell comparison. The synthesis is mine, and only defensible if I say so. The cancer stem cell model is its own Atlas page, and the Hymenolepis case has its own. Now the first promise this show made and did not keep: Strongyloides stercoralis with HTLV-1 and adult T-cell leukemia. HTLV-1 is a retrovirus that causes that leukemia; Strongyloides is a soil nematode whose larvae reinfect the same host indefinitely and, in someone immunosuppressed, multiply out of control.

They travel together: the case report this repository holds is a woman with HTLV-1-induced adult T-cell lymphoma, hypercalcemia and lytic bone lesions, treated with the four-drug CHOP chemotherapy regimen, her strongyloidiasis history in the same paragraph (Altaf 2023). A review reads the parasite as a leukemogenic co-factor, the virus shifting the host toward a type 1 immune response and cutting IgE and eosinophil activation, with no numbers behind it (Pays 2011). What can be measured is co-occurrence: 5 to 10 percent of HTLV-1 carriers go on to the leukemia (Nakahata 2023, review), and in a 30-year Peruvian registry of 116 confirmed cases, Strongyloides hyperinfection was a co-occurring HTLV-1-associated condition in 6.9 percent (Garrido-Pinzás 2026, retrospective registry). A co-occurrence frequency in diagnosed leukemia is not a risk ratio, and the prospective controlled studies the field asks for have not been done (Rosadas 2022). So I am giving no risk figure.

For scale, the current international ledger counts 2.3 million infection-attributable cancers in 2024, 12 percent of all cases, led by Helicobacter pylori at 760,000 and human papillomavirus at 750,000, then hepatitis B, Epstein-Barr, hepatitis C and Kaposi's-sarcoma herpesvirus at 35,000. Below those, in the paper's results rather than its abstract, sits a remainder of 19,000: HTLV-1, Merkel cell polyomavirus, HIV counted for conjunctival cancer alone, and the three parasites the agency classes as human carcinogens — Schistosoma haematobium at 5,900 bladder squamous carcinomas, and the liver flukes at 3,800 bile-duct cancers between them (Rumgay 2026).

The second promise was helminth therapy, the same biology from the other end, and it needs a species correction on air. The real human trial used no Hymenolepis: fifty-four patients with active ulcerative colitis were randomized to pig whipworm eggs, Trichuris suis, or placebo, and improvement came in 13 of 30 against 4 of 24, P equal to .04, while the proportion reaching a disease activity index of 0 to 1 did not differ (Summers 2005). A symptom endpoint moved; remission did not. On cancer, tapeworm infection before an intraperitoneal melanoma challenge moved survival hard: 2 of 10 melanoma-only mice alive at day 26 against 10 of 10 and 9 of 10 in the tapeworm groups, p at or below 0.001. But every mechanism test came back negative, the effect vanished with subcutaneous or intravenous tumor, and the authors call it “more preventative, rather than capable of reducing established melanoma” (Schreiber 2024).

Two corrections belong here, because the tapeworms in that experiment were Taenia crassiceps and Mesocestoides corti. The claim in this corpus is that cestodes other than Hymenolepis nana transform into malignancies in human tissue. No pathology literature describes that. Proliferative larval disease in animals is the worm's own asexual multiplication of whole larvae — canine peritoneal larval cestodiasis is Mesocestoides larvae budding, not transforming (Kashiide 2014) — and the one tapeworm that really does proliferate inside people, Sparganum proliferum, is described as asexual reproduction with sexual maturity lost, not as a neoplasm (Kikuchi 2021). The second is the checkpoint: no helminth is known to encode or secrete PD-L1, the surface protein a tumor uses to switch off an approaching T cell. What is documented is parasites inducing the host's own checkpoint ligands (Terrazas 2005, Narasimhan 2018, that last one in vitro). There is no completed published human trial of a live parasite as a cancer therapy.

The verdict. One species, one site, one lesion, and about a quarter of the dogs — the 32 of 120 referred cases, broken down already (Mukorera 2011). Seventy years after the founding paper's “possible relation” (Seibold 1955), nobody has produced the tumor on purpose: the worm has no laboratory host, and the one controlled infection study stopped at the day-169 ceiling (Kok 2011). That is the best real-world model this season's thesis has, both stronger and narrower than it usually gets told. Strong enough for a parasitology journal to put carcinogenic in a review title (Rojas 2020); narrow enough that the same literature says the triggering factor is unknown and no pro-oncogenic molecule has been detected at any stage (Porras-Silesky 2021), and that the sister worm gave 96.9 percent of 163 red foxes gastric nodules and not one neoplasm (Martín-Pérez 2022).

Now the five scores. Nutrient theft is real, and the auxotrophies are ancestral to nematodes rather than products of parasitism: free-living C. elegans cannot make heme either, these pathways having been lost broadly rather than traded away for a host (Rao 2005, Payne 2006). Cachexia is real in cancer and finally has a phase 2 result that moved weight (Groarke 2024), while the worm parallel fails against a Cochrane review of 51 deworming trials that found no improvement in height, hemoglobin, cognition, school performance or mortality and could not tell whether repeat dosing moves weight at all (Taylor-Robinson 2019) — the stunting figure belonging to the network meta-analysis (Welch 2017). The paraneoplastic parallel is genuine on the hematologic and immune-complex tracks, has one convergent hormone-receptor agonist, all of it rodent, and is rhetorical on inappropriate antidiuretic hormone secretion and on ectopic hormones. The pumps are real shared mechanism, teaching a shared failure. Self-renewal: the architecture is real, the machinery is not.

Change my mind. Experimentally infect dogs in a naturally endemic population, under ethical oversight and with their owners' consent, and keep them past day 169. The single unrun experiment that would turn the best association in veterinary parasitology into a cause. And the comparative studies the field quotes backwards deserve naming. Peto's paradox is the observation that across species cancer risk does not rise with body size, although within a species it does. In 110,148 adult zoo mammals across 191 species, cancer mortality was largely independent of body mass and life expectancy — unequivocal evidence for the paradox, in the authors' words (Vincze 2022). In 16,049 necropsy records across 292 species, neoplasia and malignancy rose with adult mass once gestation time was controlled for, which that abstract calls contrary to the paradox (Compton 2025) — two of its co-authors listing a biotechnology company among their affiliations. Humans are not uniquely cancer-prone: ferrets reach 63 percent at necropsy. Those two disagree in print — a live reversal, and where you should be told the field does not know.

Sources

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