Three Voltages Under a Worm's Skin, and the Decades Without a Controlled Trial

This is the physics of charge, and it opens on a measurement the show once said did not exist. In 1982 Thompson, Pax and Bennett advanced glass microelectrodes through the skin of a living adult male schistosome and found three voltages stacked under the surface; a companion paper in the same issue proved which layer each came from by running dye down the recording electrode. From there we separate the two quantities that are both quoted in millivolts and are not the same thing - the voltage across a living membrane and the charge on a drifting particle - and correct the folklore pair of minus seventy and minus twenty to what Binggeli and Cameron actually recorded in rats and mice in 1980. A healthy Ascaris neuron sits where the folklore puts a tumor; a malaria parasite holds minus ninety-five millivolts. Then the voltage's real job in dividing cells, in frog tadpoles, with the optogenetics experiment the owner's transcripts read backwards; the ion channels proposed as oncogenes and the four corrections that constrain them, including a retracted paper and a germline mutation that gives malformed fingertips instead of tumors, against a 2026 knockout that points the other way; the sugar coat, and every surface-charge number anybody has measured on any parasite, with the conditions attached; waves and gap-junction coupling on both sides, and the one comparison that is the host's own; the single electric-field therapy with randomized trials, two of six pivotal trials negative and no sham ever built; decades of frequency machines with no randomized or controlled trial in either direction, against the independent laboratory that looked for the frequencies in mice and found nothing; and what it looks like when electricity really does kill a parasite - a named channel with a crystal structure, a concentration and a stereochemistry. We end on every number that exists, the four nobody has measured, and three things this part refuses to say.

Show notes

In 1982 somebody pushed a glass needle through the skin of a living worm and wrote down the voltage. The worm was an adult male Schistosoma mansoni, the blood fluke of the veins draining the human bowel. The needle was a microelectrode: a glass pipette drawn fine enough to puncture one cell without killing it. Thompson, Pax and Bennett advanced it inward and found not one voltage but three, stacked under the surface. Not for the first time: the same Michigan State group had characterized the tegument's potential two years earlier (Fetterer, Pax and Bennett 1980), and the 1982 abstract says its values corroborate that paper and a 1981 report it credits to Bricker, Pax and Bennett, which could not be retrieved under that author line and is flagged here rather than cited. The record is forty-six years old; 1982 is the full analysis.

The outermost came from the tegument, the fluke's living outer skin, a continuous sheet of cytoplasm with no cell boundaries inside it: minus 45.9 millivolts, plus or minus 2.5. The subtegumental muscle beneath it sat at minus 22.0, and the extracellular compartment at minus 4.7. Input resistance, how hard a membrane makes it for injected current to flow, was 4.5, 9.2 and 3.5 megaohms; the time constant, how long a membrane takes to charge, was 0.24 and 0.25 milliseconds in tegument and muscle, 0.13 outside (Thompson, Pax and Bennett 1982).

A companion paper in the same issue made them quotable: horseradish peroxidase, which can be stained for afterward, run down the recording electrode itself, so sections show where each tip sat. Minus 51 millivolts arises across the outer tegumental membrane, minus 28 in the muscle beneath, minus 10 at the basal lamina — the tegumental and muscle potentials primarily potassium-dependent, and all three depolarized, pushed toward zero, by altered ion concentrations, by ouabain, which stops the sodium-potassium pump, and by praziquantel (Bricker, Pax and Bennett 1982).

Now the awkward part. This season's first research pass concluded three things: that no measurement of a parasite's membrane voltage or surface charge existed, that the primary papers behind the famous cancer-depolarization number could not be retrieved, and that no paper documents a bioelectric similarity between tumors and parasites. The first two were wrong. This worm's membrane voltage was recorded forty-six years ago, in indexed journals; parasite surface charge has been measured, on eggs and cysts; and the depolarization papers pull up on demand. The last was right: no paper compares the bioelectrics of a tumor and a parasite, where this part ends up. One flag: the indexed abstract prints that third potential as minus 4.7 plus or minus 03 millivolts, a typesetting fault read as 0.3, to be confirmed against the publisher's page before air.

So here is the question for the hour. Do cancer cells and parasites share an electrical identity — a depolarized membrane, a negatively charged surface, a frequency you could tune to — and has anything ever been done with electricity, to a cancer or to a parasite, that actually works? Two refusals up front: this part will not say that cancer has an electrical signature, and it will not say that a frequency cures anything. The third refusal needs the whole hour behind it.

Two quantities are both quoted in millivolts and they are not the same thing. Transmembrane potential is the steady voltage across a living cell's own outer membrane, inside against outside, read by getting an electrode through it; the cell has to be alive for the number to mean anything. Zeta potential is the charge at the slipping plane, the boundary just outside a drifting particle where the fluid stops being dragged along, read by electrophoretic light scattering. It belongs to an object's outside, not a membrane, and it moves with acidity and salt.

Four questions of any claim about charge. Which quantity, membrane voltage or surface charge? Measured how, by electrode, dye or light scattering? Alive, or fixed and dead? In what fluid, at what pH? Almost every version of cancer cells are negatively charged slides between the two: it takes a surface-charge measurement, calls it a membrane voltage, and proposes to change it by drinking something. Alkaline water does not change your cells' resting voltage; the things that do, a dangerous rise in blood potassium or a prescription channel drug, are not wellness products. One honest complication: a 2024 review argues the two are mechanistically linked, because membrane voltage reaches past the membrane and shifts the charge at the slipping plane (Hughes 2024). A review, no new data of its own, and it does not rescue the alkaline-water argument.

The popular form is minus 70 millivolts for a normal cell and minus 20 for a cancer cell, and that is not what the primary paper says. Binggeli and Cameron, in 1980, recorded in vivo with high-impedance micropipettes: Buffalo rat hepatocytes minus 37.1 millivolts against minus 19.8 for Morris 7777 hepatoma, A/J mouse corneal fibroblasts minus 42.5 against minus 14.3 for a fibrosarcoma, both differences at p equals 0.001 (Binggeli and Cameron 1980). A real depolarization of 17 to 28 millivolts, in rats and mice, in 1980 — with the normal cells nowhere near minus 70.

Binggeli came back in 1985 with Weinstein, in culture, where the pairs are unmatched — rat fibrosarcoma at minus 30.7 millivolts against minus 61.9 for normal rat kidney cells, not the same tissue. Recording rat cells every four hours for ninety-six hours, they watched normal cells hold a low potential through several divisions and then jump suddenly to a high one; cancer cells tracked neither density nor time and never jumped (Binggeli and Weinstein 1985). The defect is in the transition, not in a resting value. The one comparison in human tissue this research found is small and unmatched: tissue from nine women with infiltrating ductal carcinoma significantly depolarized against tissue from eight with benign breast disease (Marino and colleagues 1994).

Now set a worm beside it. The first intracellular recordings from nematode neurons, in Ascaris suum, the large roundworm of pigs, gave approximately minus 30 to minus 40 millivolts, tested specifically for electrode injury and found normal (Davis and Stretton 1989). A healthy worm neuron sits where the folklore puts a tumor. At the other end, Plasmodium falciparum stripped out of the red blood cell holds minus 95 millivolts on a V-type proton pump rather than on the potassium gradient that sets voltage in animal cells (Allen and Kirk 2003) — in vitro, from a radiolabeled indicator, and the Atlas page on Plasmodium carries that pump. There is no single parasite voltage to set against a tumor.

Where the voltage comes from, the ion gradients and the pump that maintains them, is derived in full on the Atlas page on tumor bioelectricity, which also carries the channel evidence and the HeLa measurement. Carry one reversal instead. Textbooks said nematode neurons do not fire action potentials; in 2018, under current clamp, the AWA olfactory neurons of Caenorhabditis elegans were shown to fire all-or-none calcium spikes, initiated by the voltage-gated calcium channel egl-19 and terminated by the potassium channel shk-1 (Liu and colleagues 2018). Worms do have regenerative electrical events, and it took direct recording to establish it.

So what does the voltage do in a dividing cell? The owner's intuition is right here, and it is in frogs. Chernet and Levin, in 2013, worked in Xenopus laevis, inducing tumor-like structures with three oncogenes plus one mutated tumor suppressor: Gli1, Kras G12D, Xrel3 and p53-Trp248. Depolarized voltage characterized those structures and was present at precursor sites before any lesion was apparent; it was not only a marker, because overexpressing hyperpolarizing transporters returned the voltage to normal and significantly reduced lesion formation, transduced through SLC5A8 (Chernet and Levin 2013). No effect sizes and no confidence intervals, a real limit.

Lobikin, Chernet, Lobo and Levin pushed it further the year before: depolarizing a strikingly small population of what they call instructor cells induced a metastatic-like phenotype in normal pigment cells, while expressing hyperpolarizing channels beforehand significantly reduced susceptibility to oncogene-induced tumor formation, an effect reproduced by a drug on the animal's own chloride channels (Lobikin and colleagues 2012). Again no numeric effect sizes, and again a frog.

Then the correction, the best teaching moment here. Chernet, Adams, Lobikin and Levin, in 2016, used optogenetics: a light-sensitive ion-moving protein put into a cell so a wavelength of light switches it on. One was Arch, a green-light proton pump; the other ChR2-D156A, a variant of channelrhodopsin-2, the textbook depolarizer. The owner's chatbot transcripts read one as depolarizing and the other as hyperpolarizing. Both hyperpolarized. The reason is the water the embryo grows in: external sodium 9.9 millimolar against 38 inside, potassium 0.2 against 51, chloride 11.1 against 30. Open a cation channel against gradients like that and the cations leave, so channelrhodopsin runs backward in a frog embryo (Chernet and colleagues 2016). One thing from the methods belongs on air: light alone did nothing, tumor incidence not varying among embryos given the oncogene alone in darkness, ambient light or the optogenetic exposures.

The outcome has been overstated too, and the error has a name: a shift in frequency read as a complete response. The paper reports no per-animal response rate. What it reports is a difference in how many embryos normalized — activation delayed until after the lesions had fully formed left 31 percent more embryos normalizing than injected but unstimulated controls, chi-square 8.6, p equals 0.003 — plus prevention of 32 percent for Arch and 31.4 percent for ChR2-D156A. Those are shifts in what fraction of tadpoles were affected, in structures the authors formally name induced tumor-like structures, and they never report metastasis and never reproduce it in a mammal.

The startling part is the one the transcripts missed. In that same assay the authors ran three oncology drugs: selumetinib against MEK at 100 nanomolar, pictilisib against PI3-kinase at 1 micromolar, vemurafenib against mutant B-RAF at 1 micromolar. All three gave the same tumor incidence as the unstimulated control. Light beat the drugs, in tadpoles, by changing a voltage rather than a gene, with the mutant Kras still in the cells — though the authors say these were the first data on those compounds in this assay (Chernet and colleagues 2016). One disclosure: the paper declares no company, but the National Cancer Institute's Office of Cancer Complementary and Alternative Medicine held a first-of-its-kind virtual meeting on cancer bioelectricity in September 2024, fourteen speakers, closing on three of its own funding routes — in a report written by that same laboratory, Michael Levin last author (Mathews and colleagues 2025).

And the best current answer to what the voltage is for comes from neither frogs nor oncology. In a crowded sheet of dog kidney cells, and in living slices of mouse lung, the cell squeezed out to die is the one with the least energy and the least membrane voltage. Crowding opens the epithelial sodium channel, the cell depolarizes; a cell with enough fuel pumps it back out and repolarizes, a cell short of fuel stays depolarized, that opens potassium channels Kv1.1 and Kv1.2 and the chloride channel SWELL1, water leaves, the cell shrinks by about a fifth, and the shrinking amplifies the signal that ejects it (Mitchell and colleagues 2025). Membrane voltage as the deciding variable in a mammalian epithelium, measured rather than asserted — though this is barrier maintenance and ordinary turnover, with tumors named only as what failure of it would mean.

Two of this host's own hypotheses belong here, because both are partly true and neither survives as stated. First: tumors are said to generate endogenous electric fields of 10 to 500 millivolts per millimeter, a range with no source. What has been measured is a wound: skin holds a transepidermal potential of 15 to 50 millivolts, inside positive, and breaching it short-circuits that battery into a lateral field as large as 40 millivolts per millimeter — read with a new instrument, the Dermacorder, by a first author at the company that makes it (Nuccitelli and colleagues 2011). No measured range for a tumor's own field exists. Second: collagen is piezoelectric, and that charge is offered as the source of those fields. The physics is real and smaller than the claim — piezoresponse force microscopy found piezoelectric heterogeneity within a single collagen fibril, in bone and tendon, outside the body (Minary-Jolandan and Yu 2009) — and it needs deformation, so it yields stress transients, not the standing field that drives electrotaxis. The tumor-relevant bridge is a gene instead: deleting the tumor suppressor PTEN enhanced electric-field signaling and electrotactic responses, while disrupting PI3-kinase gamma abolished directed movement of healing epithelium (Zhao and colleagues 2006). And one attribution to fix: the epidermal growth factor receptor gathering on a cell's cathode-facing side is keratinocytes, first author Fang, not a tumor result (Fang and colleagues 1999).

Channels, Coats, and the Sieve in a Worm's Skin

If a voltage matters, the proteins that set it should matter too. An ion channel is a gated hole through a cell's outer membrane that lets one kind of charged atom through; it is how a cell holds a voltage. In 1999 the potassium channel called ether-a-go-go was forced into mammalian cells in a dish and those cells took on a transformed phenotype, the dish behavior of a cancer cell; shutting it down in several human cancer cell lines significantly reduced proliferation, and in immune-deficient mice the transfected cells favored tumor progression (Pardo and colleagues 1999). The strongest version of the claim is cultured cells and a mouse graft.

The human tissue case came seven years later: the channel, by then renamed Eag1, is limited in normal tissue to specific brain areas and restricted cell populations, while tumor samples overexpressed it at high frequency, up to 80% depending on tissue source (Hemmerlein and colleagues 2006). In the cervix it is steeper: across 286 cytology samples plus fifteen biopsies of cervical intraepithelial neoplasia, Eag1 appeared in 27% of normal samples, 67% of low-grade lesions and 92% of high-grade. The authors looked before malignancy because estrogens and human papillomavirus oncogenes regulate this channel's gene — the one place in this hour where an electrical marker and an infectious cause of cancer meet — and morphologically normal cells from dysplastic samples carried the protein too, the specificity problem from the other side (Ortiz and colleagues 2011). The Göttingen group that made those antibodies is on both author lists.

And the strongest version of the Eag1 claim is no longer that 1999 transfection. In 2026 a group deleted the gene outright from a human lung cancer line: growth fell in several assays and in zebrafish carrying grafts of those cells, the cell cycle arrested in its first phase, migration and invasion dropped, the cells' ability to cross a vessel wall was markedly impaired, and the epithelial-to-migratory switch was abolished at its cadherin switch (Berghausen and colleagues 2026, in the Journal of Biological Chemistry). A knockout with a vessel-crossing readout in a living animal is stronger than a transfection — and it points the opposite way from the germline human mutation, so both have to be held at once: delete the channel from a cancer cell and the cancer behaves less like one; be born with a copy that opens too easily and you get fingertips, not tumors.

Four corrections now, in order of force. The first is a strong caution. The gene for that channel is KCNH1, and carrying a single altered copy, present in every cell from conception, gives you not cancer but Zimmermann-Laband syndrome: enlarged gums, intellectual disability, missing or underdeveloped nails and fingertip bones. Heterozygous missense mutations account for a considerable proportion of it, and patch clamp — sealing a glass pipette onto a patch of membrane and reading the current through it, which leaves the cell alive — shows the mutant channels open too easily, gain of function (Kortüm and colleagues 2015). A channel that opens too easily in every tissue that normally makes it gives malformed fingertips rather than tumors, which argues against channel activity alone being oncogenic, though it does not test what the tumor papers report, the channel turning up where it is normally silent. One correction while the method is in hand: the transcripts say patch clamp usually destroys the cell. It does not — the seal breaches one patch, recorded cells are held for long recordings, and the 1989 Ascaris work tested for electrode injury and found none (Davis and Stretton 1989).

Second, from inside that cervical paper: almost 50% of the normal patients taking estrogens also displayed Eag1 (Ortiz and colleagues 2011), proposed as a risk indicator and better read as the specificity problem that sinks screening markers — specificity being the fraction of people without the disease a test correctly calls negative. Third, and say it out loud because reviews still cite it: the 2014 report that Eag1 is aberrantly expressed in human liposarcoma and promotes tumorigenesis is a retracted publication (Wu and colleagues 2014). Fourth kills the general rule. KCNJ5 was significantly upregulated and tied to worse survival in an Indian pancreatic cancer cohort of twenty paired patients, and downregulated in the Cancer Genome Atlas and other Caucasian populations: one gene pointing opposite ways (Bararia and colleagues 2025). A French-language review adds that high TRPC3 correlates with favorable prognosis in lung adenocarcinoma, with no cohort and no effect size (Ouadid-Ahidouch and colleagues 2012). More ion channels does not equal worse cancer.

The best-supported claim here is not potassium but sodium. Voltage-gated sodium channel expression was significantly upregulated in metastatic human breast cancer cells and tissues, identified as Nav1.5 in a neonatal splice form, which correlated strongly with clinically assessed lymph node metastasis (Fraser and colleagues 2005). In colon cancer, blocking the channel with tetrodotoxin or silencing the gene cut invasion either way, with strong Nav1.5 staining in colon cancer specimens and little to none in matched normal colon (House and colleagues 2010). One limit binds the class: the frog work showed prevention, and nobody has halted established metastasis with a channel drug. The cleanest drug result has the genetic controls and still stops short of that: amitriptyline, an antidepressant already given to cancer patients for pain, inhibits Eag1 by binding its PAS domains; it slowed growth and migration in breast and neuroblastoma lines carrying the channel at high levels and did nothing in a melanoma line carrying little, knocking the channel in raised the effect and knocking it out removed it, and in zebrafish grafts the split held (Berghausen and colleagues 2026, in Cancer Gene Therapy).

One more charge-mover, and only as charge. The V-type proton pump sits at the plasma membrane of human breast cancer cells, prominent in the highly metastatic lines and inconspicuous in the lowly metastatic ones, and inhibitors of it cut their invasion and migration (Sennoune and colleagues 2004). The canonical statement of the broader idea, that tumors reverse the normal acid gradient from inside-out to outside-in, is a narrative review in a patents journal by authors advocating drugs against the sodium-hydrogen exchanger, with no pH values in its abstract (Reshkin and colleagues 2013). Primary measurements of pH inside human tumors do exist, by electrode: read before and after local hyperthermia and radiotherapy, tumor pH rose by a mean of 0.23 units across paired readings in 24 tumors while untreated controls did not move (Wike-Hooley and colleagues 1984), and the same group's review reports a wide range, some very low and some where normal tissue falls (Wike-Hooley, Haveman and Reinhold 1984). The proton pumps go as far as charge and no further.

Now the other quantity. The claim that cancer cells are negatively charged comes from nobody's microelectrode. It comes from cell electrophoresis: put cells in fluid, apply a field, time their drift, and you learn the charge on the outside of a cell, not the voltage across its membrane. James, Ambrose and Lowick reported the difference between normal and homologous tumor cells in Nature (James, Ambrose and Lowick 1956). For it: pigmented hamster melanoma held two subpopulations of different mobility, and mild neuraminidase, which shaves off sialic acid — a negatively charged sugar capping many chains on an animal cell's outside — reduced both to one, proving the difference came from that sugar; the faster, more negative fraction rose with tumor size and was remarkably elevated in metastatic nodules (Hyrc, Wilczek and Cieszka 1993).

Against it: in Lewis lung lines selected for high metastatic capacity, total and surface sialic acid were higher, and yet, in the authors' words, the electrophoretic mobility of the cells was unchangeable (Furesz and colleagues 1985). The modern re-reading is better than the charge story anyway: dense tumor sialoglycans are read as a self signal by Siglec receptors, mostly inhibitory receptors on immune cells like the checkpoints antibody drugs already target, and microorganisms use the same trick (Adams and colleagues 2018). The charge was a proxy for a glycan code. And the search that ought to settle the comparison returns nanoparticle characterization rather than cells, which is how this show came to say no modern measurement of a cancer cell's own zeta potential exists — it does, found another way, and the ledger reads it out.

For helminth eggs, exactly one paper gives numbers. In 2026 de Castro Novelli and colleagues read parasite structures on a Zetasizer nano and reported minus 28 millivolts for Giardia duodenalis cysts, minus 21.9 for Ascaris lumbricoides eggs and minus 18.4 for Taenia species, in a spread from minus 14.2 to minus 32.1 (de Castro Novelli and colleagues 2026). The method matters more than the number: material picked out of formalin-preserved human stool, then read in 7.5 percent formalin, with no pH and no ionic strength stated, and zeta potential moves with both. These were eggs and cysts, not living worms, with no normal-tissue and no tumor comparator. Five of eleven authors list a diagnostics company. And the protozoa were measured long before: the same paper cites Giardia cysts at minus 17 millivolts, pH-dependent, and minus 35 at neutral pH, with Cryptosporidium oocysts at minus 38 and minus 40, from the water-treatment literature.

Everything else on the worm side is qualitative cytochemistry: stains that bind charge, not meters. Ruthenium red, a positively charged dye that sticks where a surface is negative, marks a negatively charged glycocalyx — the fuzzy layer of sugar chains standing off the outside of a living surface — on adult Schistosoma mansoni, sialic acid confirmed by neuraminidase, and absent on schistosomula kept 24 hours in vitro (McDiarmid and Podesta 1984).

Then three counterexamples that forbid any generalization. The accessory layer of the cuticle of Trichinella spiralis muscle larvae lacks negative charges and carbohydrates demonstrable in vivo (Wright and Hong 1988). Dirofilaria immitis microfilariae show no negatively charged sugars, only a diffuse layer of sulfated material, where Brugia pahangi sheaths are dense with both (Hammerberg, Rikihisa and King 1984). And within one cercaria, the glycocalyx of the invading body is less negatively charged than that of the tail it throws away (Nanduri and colleagues 1991). When charge was tested as a mechanism it failed: purified human eosinophil and neutrophil cationic proteins bound lung-stage schistosomula independently of the charge on the parasite surface (McLaren, Peterson and Venge 1984). Worms carry negative groups, species by species and layer by layer, in values ordinary for any hydrated, sugar-coated surface, and no paper documents a bioelectric similarity between tumors and parasites.

Here is the hinge that joins this part to the one after it: the worm's negative charge is a collagen sieve. In live, intact Ascaris suum perfused with radiolabeled permeants, the rate-determining barrier was the lipoidal hypocuticle, but only for permeants small enough to cross the aqueous-filled, negatively charged collagen matrix of the cuticle; confirmed as the skin because worms with mouth and anus tied off took up drug as readily (Ho and colleagues 1992). The consequence shows again in a free-living nematode: positively charged iron oxide nanoparticles stuck to the Caenorhabditis elegans cuticle and translocated inward, while negatively charged ones were excreted (Zou and colleagues 2024). And it holds for eggs in a living mouse: positively charged magnetic nanoparticles were internalized by Schistosoma japonicum eggs and accumulated in hepatic granulomas, while carboxyl-terminated negative ones were barely taken up (Mei and colleagues 2026). Uptake, not killing, but it is this hinge in vivo. Charge and collagen are the same fact seen twice; part 008 is that second look, and the Atlas page on collagen invasion carries the matrix side.

The One Device With Trials, and the Decades Without a Controlled One

One electric-field therapy for cancer has randomized trials behind it. Negatives first. Tumor Treating Fields is a wearable device: adhesive skin arrays drive a low-intensity alternating electric field through the tissue beneath, worn at least eighteen hours a day. The band came from a dish. Fields of 100 to 300 kilohertz slowed eleven human and rodent tumor cell lines and tumors in mice, sparing cells that were not dividing, apparently by interfering with the mitotic spindle (Kirson and colleagues 2004) — first author at the manufacturer, NovoCure. Then ten people with recurrent glioblastoma, time to progression 26.1 weeks against historical controls (Kirson and colleagues 2007), a comparison notoriously flattering in glioblastoma.

Then the trial, and it failed. EF-11, 237 people with recurrent glioblastoma, device against physician's-choice chemotherapy: 6.6 against 6.0 months, hazard ratio 0.86, interval 0.66 to 1.12, p equals 0.27 (Stupp and colleagues 2012). A hazard ratio is the rate at which events arrive in one arm against the other, so 0.86 means fourteen percent slower; an interval spanning 1.0 settles nothing. A reanalysis then kept only patients who finished a cycle and got 7.7 against 5.9 months, p equals 0.0093, with wearers of eighteen hours or more living 7.7 against 4.5 months (Kanner and colleagues 2014). That last is circular: a patient well enough to wear a headpiece all day was going to live longer anyway.

EF-14 is the real evidentiary basis. 695 people, newly diagnosed glioblastoma after radiochemotherapy, randomized two to one to 200-kilohertz fields through four scalp arrays plus maintenance temozolomide, or temozolomide alone: progression-free survival 6.7 against 4.0 months, overall survival 20.9 against 16.0, both hazard ratios 0.63, both p below 0.001, with mild-to-moderate skin toxicity under the arrays in 52 percent of device patients and none of the controls (Stupp and colleagues 2017). Nearly five months of median survival in glioblastoma is rare and unambiguous. It was also open-label, and no sham-controlled trial of this device has ever been run.

Then it widened. Lung cancer, 276 people, 13.2 against 9.9 months, p equals 0.035 (Leal and colleagues 2023). Brain metastases after radiosurgery, 298 people, hazard ratio 0.72, p equals 0.044 on a radiographic endpoint rather than survival (Mehta and colleagues 2025). And the trial nobody quotes: 558 women with platinum-resistant ovarian cancer, 12.2 against 11.9 months, hazard ratio for death 1.01, p equals 0.89, its one positive finding a post hoc subgroup (Vergote and colleagues 2025). With EF-11, two of six pivotal trials are negative on their primary endpoint.

Pancreatic cancer needs its own reading. 571 people, device plus gemcitabine and nab-paclitaxel: overall survival 16.2 against 14.2 months, hazard ratio 0.82, p equals 0.039, pain-free survival 15.2 against 9.1; progression-free, local progression-free and response rate unimproved; distant progression-free survival 13.9 against 11.5 months, hazard ratio 0.74, p equals 0.022, analyzed post hoc; device-related skin adverse events in 76.3 percent, grade 3 in 7.7 (Babiker and colleagues 2025). Local control did not improve and distant control did, in an analysis decided after the results were seen, so the local cytoreduction the laboratory work proposed is not what this trial supports, and an unblinded supportive-care effect stays live. A cost-utility model puts the price at 387,300 dollars per quality-adjusted life year, scenarios running to 547,200, two of four authors at NovoCure (Guzauskas and colleagues 2026).

The strongest argument that the field does something is dose-response: in simulated head models of 340 EF-14 patients, survival ran longer above a modeled dose threshold in the tumor bed, 25.2 against 20.4 months, p equals 0.003, and above a modeled average field of 1.06 volts per centimeter (Ballo and colleagues 2019). Flag the unit before air: the indexed abstract prints that threshold as 0.77 milliwatts per centimeter, while the quantity defined is a power density, per cubic centimeter. The dose was simulated, not measured, with thresholds drawn from the same data they were tested on, by authors four of six at the manufacturer.

The criticisms are published and they have names (Ranganathan and colleagues 2026) — one author also works for a Dutch health insurer, CZ, which pays for treatments like this. No sham-controlled trial of this device exists, so the field cannot be separated from the round-the-clock support, home deliveries and regular visits that come with it. Human dosimetry is unvalidated, so nobody can tell a patient how many volts per centimeter reached their tumor. Control arms were suboptimal in half of the six pivotal trials, two of them positive, and there were protocol changes and unusual censoring, with one company generating the main evidence base. What they do not dispute: groups with no manufacturer author have reproduced the antiproliferative effect in temozolomide-sensitive and -resistant glioblastoma cells (Jones and colleagues 2023), in freshly derived colorectal lines (Su and colleagues 2026), on independent hardware at 200 kilohertz (Senturk and colleagues 2026), and in patient-derived glioblastoma organoids, where fields plus temozolomide cut viability, colony formation and migration more than either alone (Zhu and colleagues 2026). The physics is real in a dish; what is disputed is trial design, since in one arm of every pivotal trial the patient wears visible skin arrays eighteen hours a day.

Waves are real on both sides, and the episode should say which half is whose. Many cells in astrocytomas, the family that includes glioblastoma, extend ultra-long membrane protrusions named tumor microtubes and use them to invade, proliferate and interconnect through gap junctions; microtube-connected cells survived radiotherapy that killed the unconnected ones (Osswald and colleagues 2015). Glioblastoma networks also hold a small population of highly connected cells firing rhythmic calcium oscillations, their transients preceding those of connected cells and activating the growth-signaling MAP-kinase relay and nuclear factor kappa B; ablating those periodic cells, or blocking the potassium channel KCa3.1, cut viability across the network, slowed tumor growth in mice and extended survival (Hausmann and colleagues 2022).

Now the parasite half, back at the recording this part opened on. The tegument and the muscle beneath it are both electrical syncytia — tissues so electrically continuous that current spreads as if there were no cell boundaries — joined by low-resistance pathways: a signal started in either loses only 15 to 25 percent crossing into the other (Thompson, Pax and Bennett 1982). The review of that work gives the verdict that matters: despite the schistosome's unique six-layered outer membrane, its biophysical properties are, in its words, 'not markedly different from those of a variety of other multi-dimensional syncytia', and the schistosome's longitudinal contraction waves may be myogenic, arising from the coupled muscle sheet itself (Pax and colleagues 1983).

So here is the only defensible bridge, labeled as what it is: two gap-junction-coupled electrical collectives, one a tumor and one a fluke, measured by two fields that have never cited each other. The similarity is structural, at the level of coupling — not a shared voltage, frequency or molecule, since a tumor's gap junctions are built from connexins and the worm's phylum uses innexins, an unrelated family doing the same job. The comparison is this host's framing. No paper makes it.

One reversal the theme needs: tumors lose their gap junctions was the textbook line and it is half right. In primary breast tumors and the matched lymph-node metastases from those same tumors, connexin 26 and connexin 43 were significantly higher in the metastases, p below 0.00001 and p below 0.001, connexin-negative primaries giving rise to connexin-positive metastases (Kanczuga-Koda and colleagues 2006); primaries often lose junctional communication while metastatic lesions can gain it (Wu and Wang 2019). The further step is a mouse tail-vein result: breast cancer cells overexpressing connexin 43 adhered more to lung endothelium, and a dominant-negative version markedly reduced that adhesion (Elzarrad and colleagues 2008). Same molecule, opposite direction, depending where in the disease you look.

And the schistosome's own phylum has one honest result. In planaria, free-living flatworms that regrow whole bodies from fragments, depolarization by the hydrogen-potassium pump is required for head regeneration: blocking it stopped the front-end genes and the brain, while depolarizing by a route that bypasses the pump drove head formation even at backward-facing wounds (Beane and colleagues 2011). Voltage and coupling genuinely carry pattern information in a flatworm — a free-living one, not a parasite and not a tumor.

Now the frequency machines, the clearest worked example of how a fringe idea acquires a respectable surface. Start with the absence and name it as the finding: the biomedical literature holds no randomized or controlled clinical trial of a Rife-type frequency generator for cancer; what exists is uncontrolled single-arm work from one group. The one authoritative statement is blunt and old — a review of electronic devices found, in its words, no evidence that treatment with the devices mentioned in this review results in objective benefit in the treatment of cancer in human beings, and urged people with cancer not to seek them (American Cancer Society 1994) — and it covers the electronic-device tradition, not the device above.

Walk the respectable version in order. The frequencies came from an unblinded biofeedback examination of 163 patients, yielding 1,524 frequencies from 0.1 hertz to 114 kilohertz, with 13 of 28 compassionate-use patients evaluable and no control group (Barbault and colleagues 2009), registered as NCT00805337, which cannot establish that a frequency is specific to anything. Next, a single-group open-label study: 41 people with advanced hepatocellular carcinoma, three daily sixty-minute intrabuccal treatments, 34.1 percent stable past six months, median overall survival 6.7 months, no control arm (Costa and colleagues 2011). Next, 59 patients pooled, 18 new plus the same 41: among the 32 with mildest liver impairment, median overall survival 10.36 months against 7.74 for comparable historical controls, p equals 0.036 (Blackstock and colleagues 2021) — though only grade 1 mucositis and fatigue, and the Child-Pugh B arm came in at 4.73 months against 4.6 for historical sorafenib controls, no difference at all. Then the objections: one p value, one subgroup, the earlier 41 counted twice, a low-profile journal, and three of fourteen authors declaring TheraBionic affiliations, with Barbault and Pasche running through all three papers.

Then the one independent test. A consortium at French national research laboratories implanted two human liver cancer lines in the flank of immunodeficient mice, exposed them to a 27.12-megahertz carrier modulated by systematic scans and by the specific published frequencies, and read each animal's pulse: no tumor-related frequencies, and no detectable physiological change (Veyret and colleagues 2020). It tested whether the frequencies could be found at all, not whether they shrink tumors, and the authors allow that mice may differ from people.

Two refusals. This episode will not state the regulatory status, clearance number, date or indication of any named frequency device, because the primary documents were never opened. Generically, one pathway asking only for probable benefit in a rare condition asks far less than one asking for demonstrated effectiveness. The Zapper and Syncrometer of the Hulda Clark tradition appear nowhere in the indexed literature, so their descriptions must come from her books or a regulator's published action, never a seller's website. One caution against the skeptical side: the trials above show external arrays can have a clinical effect deep in the body (Stupp and colleagues 2017), though the field that reaches the tumor is modeled rather than measured (Ballo and colleagues 2019). The objection to the Zapper is the absence of a measured field strength, a frequency rationale, an exposure duration and a randomized test.

The only controlled data in the tradition come from the cousin devices. Electrodermal diagnosis, which claims to read illness from the skin's electrical properties, was tested blind against skin prick testing in 30 volunteers, half allergic, across three operators and three sessions — 54 tests each by design, 1,620 in all, of which 1,584 followed the protocol: no correlation with the prick tests, no ability to tell allergic from non-allergic, not one participant consistently right (Lewith and colleagues 2001). The one positive trial in the tradition, a proprietary bioresonance device for smoking cessation, reported a 54.8 percent one-week quit rate in its own placebo arm, which warns about the outcome measure rather than commending the device (Pihtili and colleagues 2014) — and smoking cessation is not cancer treatment.

Then give the resonance idea its due, because the real version exists. Caltech engineers model oncotripsy, tuned low-intensity pulsed ultrasound selectively fatiguing cancer cells by exploiting their size and stiffness (Schibber and colleagues 2020), and a viscoelastic simulation puts the resonant frequency of a cancer cell roughly 50 to 100 kilohertz away from a healthy one's and predicts its vibration amplitude growing about ten times faster (Liaquat and Al-Jumaily 2025). Schibber's model is fitted to that group's own laboratory work; the skin-cell study is simulation only, and its authors ask for the experiment. Neither has been tested in an animal or a person.

Now the contrast, which is the point of this section: what it looks like when electricity really does kill a parasite. Praziquantel was in clinical use for forty years before anyone knew how it worked. Part 003 and the praziquantel Atlas page own the mechanism — a schistosome transient receptor potential channel opened stereoselectively by one of the drug's two mirror images, calcium flooding in, the worm paralyzed (Park and colleagues 2019; Harder 2020). The beat this section owns is the side effect: the same drug is a selective partial agonist of our own cold-and-menthol receptor TRPM8, offered as a possible clue to the nausea, abdominal pain and headaches people report (Babes and colleagues 2017). And the class has company: meclonazepam, an old benzodiazepine that still works on the juvenile worms praziquantel spares, acts on a second, distinct schistosome channel of the same family, its weaker effect on Schistosoma japonicum traced to one polymorphism in the binding pocket (Park and colleagues 2023). Two drug classes, two channels, depolarization the readout for both.

One reversal before the contrast, because the named channel is not fixed. Whole-genome sequencing of 570 schistosome samples from eight countries after decades of mass treatment turned up four naturally occurring variants of that same channel showing reduced praziquantel sensitivity in the laboratory, alongside treatment failure in worms taken from people before and after a dose (Berger and colleagues 2026); a community database now records which variants have been tested, where and how common (Rohr and colleagues 2026). That sharpens the contrast rather than softening it: a mechanism specific enough to fail in a way somebody can measure.

Ivermectin is cleaner still — and the oncology claims made for it, with the arithmetic of a human dose, belong to part 009, not here. Expression cloning from Caenorhabditis elegans produced an avermectin-sensitive glutamate-gated chloride channel, invertebrate while the drug's other target, the GABA receptor, is not, from Merck, which markets the drug (Cully and colleagues 1994); its structure at 3.3 angstroms shows ivermectin wedged in the membrane-spanning domain, holding the pore open (Hibbs and Gouaux 2011). An open chloride pore pins the worm's membrane negative, and a neuron pinned negative cannot fire. It is safe in people partly because we lack that channel and partly because a pump keeps it from the brain: mice lacking the mdr1a P-glycoprotein were a hundredfold more sensitive (Schinkel and colleagues 1994).

Two cautions. The avermectins and milbemycins act on glutamate- and GABA-gated chloride channel subunits, so the single-channel story is a simplification, and ivermectin resistance in ruminant worms may select on ATP-binding cassette transporters (Prichard 2005). The program reaches past ivermectin: emodepside acts on SLO-1 potassium channels and at 0.15 and 0.75 milligrams per kilogram killed adult worms by 18 months, which ivermectin cannot do — eight of its authors at Bayer, in cattle (Bah and colleagues 2021). Then the exception that stops the generalization becoming a slogan: the benzimidazoles, albendazole and mebendazole, have nothing to do with voltage, binding the worm's tubulin (Lacey 1988), while a review reports the majority of antinematodal drugs do target ion channels (Choudhary and colleagues 2021). The structures live on the ivermectin Atlas page. Feel the distance between two things both called frequency medicine: a channel with a name, a crystal structure, a concentration and a stereochemistry, against a number somebody found by watching a biofeedback needle.

Every Number That Exists, and the Four Nobody Has Measured

Before the refusals, the ledger: every surface-charge number this part leans on, set down with the water it was measured in.

Start with the old protozoan numbers, because they are better conditioned than the new worm one. Cryptosporidium parvum is a single-celled gut parasite whose oocysts, the tough-walled stage it sheds, are what water treatment has to catch. In deionized water at pH 6 to 6.5 the oocysts carried a zeta potential close to minus 25 millivolts; at alkaline pH, minus 35; and lowering the pH shrank the charge to zero at pH 2.5, the isoelectric point, the acidity at which a surface carries no net charge (Drozd and Schwartzbrod 1996).

That paper proves the lesson with one organism: the same oocyst reads minus 35, minus 25 or zero depending on the pH you choose, so a zeta potential reported without its water is half a number — which is why a thirty-year-old water-treatment study outranks the 2026 helminth paper here, whose cysts and eggs were read in formalin with neither pH nor salt concentration stated.

Leishmania mexicana amazonensis follows, measured in 1990 as promastigotes, the flagellated form the parasite takes in the sandfly and in culture. Virulent and avirulent alike sat at a zeta potential of about minus 15 millivolts; the abstract does not name the fluid. Trypsin, alkaline phosphatase and phospholipase C each made it less negative; neuraminidase did not change it at all, and cationized ferritin agreed enzyme for enzyme. This parasite's negative surface points to phosphate and protein groups, not to the sialic acid the tumor-charge tradition rests on.

The same experiment pulled two quantities apart: a surface free energy of minus 6.4 millijoules per square meter for avirulent promastigotes against minus 18.1 for virulent ones (Silva Filho and colleagues 1990). Virulence tracked stickiness, not charge — the two were measured in the same cells and did not move together.

Now the one result that is the host's intuition in its strongest form. Costa e Silva Filho and colleagues measured electrophoretic mobility, the raw drift speed a zeta potential is computed from, in three strains of Trichomonas vaginalis, a parasite of the human genital tract, and five of Tritrichomonas fetus, a relative that infects cattle, across pH and ionic strength. The species did not differ. The strains did, significantly, and the strains more pathogenic to mice carried the more negative surface. Trypsin or neuraminidase reduced mobility and raised the isoelectric point, and from the enzyme-recovery pattern the authors suggest sialoglycolipids contribute more to the charge of T. vaginalis than of T. fetus (Costa e Silva Filho and colleagues 1986).

More negative, more pathogenic. Now say how thin it is: eight strains of two protozoa, in 1986, scored against mouse lesions, with no tumor and no human in it, and no mobility value in the abstract to quote. That is the best parallel the literature offers, and it is forty years old.

The cancer side's best-known modern counterpart is not a cell. Exosomes are small membrane packets cells shed into surrounding fluid. Kosutova and colleagues compared exosomes from a prostate carcinoma line, 22Rv1, with those from a benign prostate line, RWPE1: the carcinoma packets were smaller, more negative and produced at about three times the concentration, with alpha-2,3- and alpha-2,6-sialylated glycans more abundant on them (Kosutova and colleagues 2024). The abstract gives no zeta values; the full text does: minus 14.9 millivolts against minus 9.9, single values, in a pH 7.0 buffer. Two cell lines, and packets rather than cells.

Cancer-cell measurements do exist, and this show's search missed them. Hoque and colleagues measured live HeLa cells against primary cervical epithelial cells bought from a supplier rather than taken from the same patients, in a low-salt sugar medium at a stated conductivity of 43 millisiemens per meter. The cancer cells read minus 15.4 millivolts, the normal cells minus 22.3: the cancer cell was the less negative. Membrane voltage derived by dielectrophoresis was minus 49.2 against minus 54.1, calculated rather than recorded, with the equation of Hughes, author of the 2024 review and among these authors too. They call the results preliminary findings (Hoque and colleagues 2025).

A 2019 study built to sort breast cells measured zeta potential on a non-tumorigenic breast line, MCF-10A, the breast cancer line MCF-7 and a tamoxifen-resistant clone of it, in phosphate-buffered saline at pH 7.4 and in a lower-salt phosphate buffer. Every line read slightly more negative in the lower-salt buffer, and the lines were identical within experimental error (Kalasin and colleagues 2019). The salt moved the number. Cancer did not.

Say the consequence out loud. The modern cancer-side measurements do not agree even on direction: a prostate carcinoma's packets more negative than benign ones, a cervical cancer's cells less negative than normal ones, breast cancer cells no different. Hoque and colleagues cite an earlier breast pair that runs the other way from Kalasin, MCF-7 at minus 20.3 millivolts against MCF-10A at minus 31.2 (Hoque and colleagues 2025), so the cell studies lean toward cancer cells being the less negative, and the popular line that cancer cells are more negatively charged finds no support in them. The sentence a worm egg at minus 21.9 millivolts against a cancer cell at X can now be given a number and still cannot be finished: the Ascaris egg was dead and fixed, the HeLa cell alive in a sugar medium.

Most of the worm side gives a sign and no magnitude. The surface coat over the epicuticle on infective larvae of Toxocara canis, the dog roundworm, binds cationized ferritin and ruthenium red, indicating a net negative charge; its major component is the secreted glycoprotein TES-120, ethanol strips the coat, and antibody binding makes the larva shed it, which the authors read as immune evasion (Page and colleagues 1992). An immune-evasion story, not an electrical one.

On the cercaria of Schistosoma mansoni, cationized ferritin stained the cuticular level strongly; neuraminidase did not alter the staining, while trypsin or chondroitinase abolished it (Cavalcanti and colleagues 2008). On the adult, neuraminidase confirmed sialic acid (McDiarmid and Podesta 1984). Within one species the chemistry of the charge changes with the life stage. And the adult's ferritin-binding membranocalyx is continuously rebuilt from vesicles beneath: by four hours in fresh medium most of the label is gone, a half-life of two to three hours (Wilson and Barnes 1977), which the schistosoma Atlas page carries with the rest of the coat. A coat that halves itself roughly three times in an eight-hour shift is not a stable signature of anything.

Now name what the cancer side cites when it says depolarized. It is reviews, which is where this claim lives. The paper most people mean is Yang and Brackenbury's: it states that electrophysiological analyses in many cancer cell types have revealed a depolarized membrane voltage that favors proliferation, and that hyperpolarization is necessary for stem-cell differentiation, osteogenesis and adipogenesis both hindered in human mesenchymal stem cells under depolarizing conditions. From that it proposes depolarization might matter for the emergence and maintenance of cancer stem cells (Yang and Brackenbury 2013). No new measurements, and a full text declaring no commercial relationships.

The 2026 position paper concedes the show's point in the field's own words. Desoyer and colleagues summarize about fifteen years of work and write that malignant cells consistently show depolarized transmembrane potentials correlating with proliferation, stemness, invasion and therapy resistance, suggesting depolarization as a conserved bioelectric hallmark of malignancy. Then they write that a quantitative, translationally actionable framework for membrane potential in cancer is lacking, citing limited comparability across tumor types, incomplete mechanistic integration and no standardization for clinical translation. From the full text, since the abstract carries neither: their structured search screened 358 records and kept 26, and puts cancer cells at roughly minus 10 to minus 50 millivolts and differentiated cells at minus 50 to minus 90, with comparison between studies limited by method, electrodes against voltage-sensitive dyes, and by calibration (Desoyer and colleagues 2026).

Two things the listener should know. It is a review. And most of its authors work at a university institute that houses a European testing center for medical devices, with tumor-treating fields among the paper's keywords; the paper declares no competing interests. The hallmark is asserted; the comparability is admitted to be missing. And on their own band, the schistosome's tegument straddles the boundary: minus 45.9 millivolts in one paper and minus 51 in its companion (Thompson, Pax and Bennett 1982; Bricker, Pax and Bennett 1982), the first inside the cancer range and the second inside the differentiated one, while an Ascaris motor neuron at minus 30 to minus 40 (Davis and Stretton 1989) would be filed with the cancers.

Which leaves four measurements nobody has taken. One: electrophoretic light scattering on live schistosomula and adult worms, in defined buffers across a pH series, with a mammalian cell control run the same day. Two: paired membrane-voltage recording in freshly dissociated human tumor and adjacent normal epithelium from the same patients, with blinded histology. Three: one assay containing both a parasite and a tumor cell, never published. Four: a sham-controlled trial of an alternating-field device, whose control arm wears a convincing dummy array, which nobody has built.

And the searches, named as this show names them. Bioelectric and membrane-potential terms combined with parasite, helminth and tumor return no paper that documents the similarity. A search for a cancer cell's zeta potential since 2015 brings back drug carriers, gold particles and carbon dots, not a cancer cell's surface; the cell measurements in this ledger were found other ways. Everything measurable here has been measured on one side or the other, and never on both at once.

So the verdict on charge, which is all this part claims. One pattern runs through every topic: the physical observation held up and the inference about cancer did not. Cancer cells really are less negative than matched normal tissue, by 17 to 28 millivolts — but the normal cells sat at minus 37.1 and minus 42.5, nowhere near the folklore's minus seventy (Binggeli and Cameron 1980). The channel really is overexpressed in tumor samples — but a stuck-open copy gives malformed fingertips, and almost half the healthy women on estrogens in that cervical series were positive too (Hemmerlein and colleagues 2006; Kortüm and colleagues 2015; Ortiz and colleagues 2011). The sugar coat really is denser on some tumors — but in the metastatic lung lines the drift speed did not budge (Hyrc, Wilczek and Cieszka 1993; Furesz and colleagues 1985). The device really does slow dividing cells and added five months in glioblastoma — open-label, no sham, two of six pivotal trials negative, the dose modeled and never measured (Kirson and colleagues 2004; Stupp and colleagues 2017; Ranganathan and colleagues 2026). That is the shape of the error, four times over.

Three refusals, then. This part does not say cancer has an electrical signature. Dielectrophoretic separation of the NCI-60 panel of sixty tumor cell lines turned out to be reading cell size and plasma membrane folding, not a malignant electrical identity (Gascoyne and colleagues 2013). The microwave contrast that launched an entire detection field runs as high as 10 to 1 only against fat, and no more than about 10 percent against the glandular and fibroconnective tissue tumors actually arise in (Lazebnik and colleagues 2007). An honest electrical test looks like impedance spectroscopy across 1,943 skin lesions including 265 melanomas: sensitivity 96.6 percent, specificity 34.4 percent, positive predictive value 21.1 percent (Malvehy and colleagues 2014). Four in five positives false: high sensitivity with low specificity makes a rule-out test, not a detector. Second, this part does not say a frequency will cure anything. Third, it does not say the link is fringe: three organisms sit in Group 1 of the World Health Organization's cancer agency, the evidence that they cause cancer in people sufficient. And when somebody searched for the rest, leaving those three out, 1,266 records yielded 19 worth detailed evaluation (Machicado and Marcos 2016). That is not evidence against; it is a thin literature.

And the season's honest version. The reason nobody finds parasites in tumors is not that anybody is hiding anything, and it is not that the physics is exotic. The voltage is ordinary. The charge is ordinary. The enzymes, as part 008 shows, are not even the same family — three invasive parasites, three catalytic classes, not one of them a matrix metalloproteinase. What is extraordinary is that the parasite half of this question is on nobody's list — the electrical half reached a National Cancer Institute meeting in 2024 — while the tissue to answer it is already in the building, where part 008 turns it into a protocol somebody could fund.

Sources

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